AMD’s FX-8150 ‘Bulldozer’ processor


Tick-tock. Tick-tock. The sound of Intel’s ongoing CPU development cycle has been constantly in the backdrop for its biggest competitor, AMD, ever since the world’s largest chipmaker set an aggressive cadence for itself more than five years ago. Since then, Intel has turned over new manufacturing technologies followed by extensively revised CPU architectures in relentless succession. The introduction of Sandy Bridge processors at the beginning of this year put Intel firmly in the lead in terms of overall performance, power efficiency, and the value proposition offered to consumers.

Being the perennial number-two CPU maker in such a competitive context can’t be easy, but AMD hasn’t taken the challenge lightly. In fact, the firm has been working for several years on a brand-new breed of PC processors based on a fresh microarchitecture, code-named “Bulldozer,” that aims to restore some competitive balance. Nearly every CPU AMD has made for the past decade-plus (with the exception of the low-power Ontario/Zacate E-series APUs) has been derived from the original K7, the chip first known as Athlon. Bulldozer draws on that tradition in various ways, but it is a novel, clean-sheet design intended to take AMD processors into their next era.

To that end, Bulldozer introduces some unorthodox concepts into the PC processor space. The first of those is a dual-core “module” as a fundamental building block. To date, we’ve seen x86-compatible CPU cores capable of tracking and executing two threads via a feature known as simultaneous multithreading (SMT), better known by its Intel marketing name, Hyper-Threading, and we’ve had a number of chips with multiple cores onboard in a chip-level multiprocessor (CMP) configuration, tracing back to the original Athlon 64 X2. The Bulldozer module is sort of a mid-point between those two familiar arrangements. AMD says the module has “two tightly coupled integer cores” with some sharing of resources—including, notably, the FPU. The idea is to save space on the silicon die by pooling resources where possible while still offering “robust” performance on both threads, with fewer of the performance hazards created by SMT or Hyper-Threading.

At the same time, Bulldozer resurrects a concept that’s fallen out of favor in PC processors in recent years: it’s a “speed demon,” optimized for higher clock frequencies rather than maximum instruction throughput in each clock cycle. The Pentium 4 “Netburst” microarchitecture—particularly in its troubled “Prescott” incarnation—gave frequency-optimized designs a reputation for high power draw and iffy performance. Yet Chief Architect Mike Butler told us the engineering team’s goal with Bulldozer was to “hold the line” on instructions per clock (presumably at about the same rate as the Phenom II) and to “aggressively pursue higher frequencies.” Speed demons have typically reduced the amount of work done at each stage of the pipeline in order to simplify logic and thus enable higher operating frequencies, but this approach can also theoretically help manage power consumption. The rationale, if we understand Butler correctly, is that a design with a relatively low number of gates flipping at each pipeline stage may require less voltage to operate at a given frequency. Chip power consumption has three main determinants: clock speed, the number of transistors flipping, and the square of the voltage. That voltage squared term is, obviously, the single biggest factor in the power equation, so a design capable of keeping voltage in check could make some sense for today’s power-constrained world.

We don’t know precisely how aggressively AMD has pursued the speed-demon approach. When we asked, AMD declined to tell us the number of stages in Bulldozer’s main pipeline. This behavior seems unusually guarded. We’ve been writing about these things for over a decade, and an outright refusal to disclose pipeline depth in a major x86 processor is very rare. Our sense is that it’s somewhere between the 12-14 stages of contemporary Core- and Phenom-branded chips and the astounding 31 stages in Prescott. I expect we’ll learn more about Bulldozer’s inner workings as time passes.

The “Orochi” die with major functional areas highlighted. Source: AMD.

At any rate, here’s the big picture. The first incarnation of the Bulldozer architecture is a formidable chip, with four modules onboard. That gives it a total of eight integer cores, with four floating-point units. Each module has 2MB of L2 cache, and there’s a shared third-level cache of 8MB. This chip retains compatibility with AMD’s existing system architecture, so it has an integrated memory controller with support for dual channels of DDR3. Also present are four HyperTransport links, only one of which will be used in desktop products.

Inside the module

Because Bulldozer is what it is—an all-new, high-performance x86-compatible processor—it’s incredibly complex and difficult to summarize. Nevertheless, we’re going to make a quick attempt, with the assistance of the block diagram below, which provides a high-altitude overview of Bulldozer’s key components.

Block diagram of a Bulldozer module. Source: AMD.

The sharing in a Bulldozer module starts with the front end, where the branch prediction, instruction fetch, and decode units track two threads and service both cores. With two integer cores featuring relatively long pipelines to keep fed, the front-end hardware must be very effective at its job in order for the whole chip to function efficiently.

The decode units dispatch ops, or decoded instructions, to the two integer cores on an interleaved, every-other-cycle basis. Each of those cores has a pair of ALUs, and each ALU has an associated address generation unit. Thus, individual Bulldozer cores have fewer execution resources than those in the preceding Deneb/Thuban architecture. However, instruction scheduling is more flexible, and beyond the obvious increase in integer core counts, Bulldozer seeks to make things up in other ways.

One of those ways is a vastly reworked memory subsystem that looks very different from those in prior AMD chips. Among other things, the memory pipeline can speculatively move loads ahead of stores if doing so won’t cause a problem, a capability Intel has called memory disambiguation. Memory access latencies should be further reduced by the use of multiple data prefetchers that operate according to different rules in order to keep the caches populated with, hopefully, the appropriate data for the cores’ upcoming work. Both the prefetchers and the L2 cache into which they pull data are shared between the two cores in a module, assuming both cores have active threads. If only one thread is active, these resources are fully used by that single thread.

Another major shared resource in the Bulldozer module is the floating-point unit, which has been spun off into a co-processor arrangement in which both integer cores act as clients. This setup is quite different from the intermixed integer and FP execution resources in Sandy Bridge, and AMD has hinted that it may pave the way for a GPU-type shader array to one day take the place of the traditional FPU. For now, though, Bulldozer’s FPU is quite formidable in its own right. The scheduler can track two threads, of course, and the execution units include dual FMAC units capable of processing 128-bit vectors in a single clock cycle, along with dual 128-bit integer units (marked as “MMX” in the diagram above). Yes, that means integer SIMD goodness happens in the FPU, as well as floating-point math.

The fact that both Bulldozer and Sandy Bridge, two substantially new x86 microarchitectures, have hit the streets within the same calendar year isn’t entirely coincidental. The common thread is the advent of the follow-on to SSE, the extended instruction set known as Advanced Vector Extensions, or AVX. AVX increases parallelism by extending the width of vectors from 128 to 256 bits, and supporting those wider datatypes requires the broad reworking of the processor’s execution engine. The result should be much higher peak computational throughput on data-parallel workloads.

However, the path to that destination will have a few twists and turns. After initially proposing its own 256-bit vector extensions known as SSE5, AMD has reversed course and attempted to follow Intel by making Bulldozer compatible with AVX, instead. As that change was happening, Intel apparently was modifying its own course, as well. So Bulldozer catches up with Sandy Bridge on nearly every front, adding support for SSE 4.1 and 4.2 and most of AVX, including the AES instructions for accelerating encryption. It also includes support for AMD’s own XOP extensions, a surviving bit of SSE5 with more of a focus on integer datatypes. Where Bulldozer moves beyond Sandy Bridge, though, is with those two 128-bit FMAC pipes—and there, we get into disputed territory.

The dispute is over the FMAC instruction, which is the key to unlocking AVX’s peak potential. FMAC stands for “fused multiply-accumulate,” an operation that can be described logically as: “d = a + b * c”. Instructions that combine a multiply and an add together tend to map well to multimedia workloads, and they have been a staple of GPU shader cores for quite some. Doing both operations at once has a performance benefit, obviously—the processor is executing two floating-point operations (FLOPS) per clock cycle. The FMAC form of this instruction has a further precision advantage because the results of one operation are fed directly into the other, at the chip’s full internal precision, without being stored. These virtues have made FMAC very popular in other chips, including DirectX 11-class GPUs.

Bulldozer is the first x86 CPU to support FMAC. Sandy Bridge doesn’t, and the upcoming Ivy Bridge won’t, either. Instead, Intel intends to add FMAC support to Haswell, its next architectural refresh, due in 2013. Trouble is, Bulldozer supports a version of FMAC with four operands, while Haswell will support a three-operand variant of FMAC. This sort of incompatibility isn’t a good thing when you’re trying to persuade software developers to use your new instructions. AMD seems to recognize that fact, so it plans to add FMAC3 support in the next version of Bulldozer, code-named Piledriver, alongside FMAC4. The FMAC4-only chip we’re looking at today, though, will always be something of an oddity, as a result.

All of this madness still leaves Bulldozer in decent shape, FPU-wise, but not quite indisputably at the head of the pack. Even without FMAC support, Sandy Bridge still has two 256-bit vector units, so it can produce a 256-bit add and a 256-bit multiply in a single clock cycle. Bulldozer can theoretically match Sandy’s peak throughput, either by processing dual 128-bit FMACs or a single 256-bit FMAC per cycle, but it can’t match Sandy without FMAC.

For a discussion of the Bulldozer microarchitecture in much more depth, let me point you to David Kanter’s excellent piece on the subject, from which I’ve stolen small bits of info here and there.

Power management and Turbo Core

Now that we’ve spent entirely too much time on the FPU, let’s move on to power management, another topic too large to cover in the time and space we have today. Power efficiency has become critically important in modern processors, and any clean-sheet architecture like this one will include a zillion little pockets of logic conceived with power efficiency in mind.

The headliner here, though, is the use of power gates for each of the modules and a fifth power gate for the north bridge and L3 cache. Closing one of these gates shuts off power to the portion of the chip behind it, even leakage power. Intel has used power gates to good effect since Nehalem. AMD first used power gates in the Llano APU, where they are quite effective, but Bulldozer is its first high-end CPU to employ them.

Another feature makes a surprise return: separate clock domains for each of the modules, along with one for the north bridge. (The north bridge and L3 cache run at 2.2GHz in desktop parts and 2-2.2GHz in Bulldozer-derived Opterons.) AMD first instituted separate clock domains per core in Barcelona, the original Phenom chip, but back-tracked in the Phenom II generation and used BIOS code to lock all four cores to a single clock—making the Phenom II operate much like Intel’s recent CPUs do. Turns out threads pinging around from one core to the next in the Windows scheduler sometimes led to performance issues, because threads would be reassigned to cores operating at low frequencies. AMD tells us it has returned to this approach for a simple reason, “because power is important.” Our sense is that Bulldozer should be better equipped to avoid problems on this front. The chip has a higher floor for clock speed (1.4GHz versus 800MHz in the Phenom II), improved latency for clock-speed ramps, and can probe the caches of other modules more quickly. AMD also seems to be banking on smart scheduling in future versions of Windows to accommodate the Bulldozer architecture, a subject we’ll discuss shortly.

Illustration of Bulldozer Turbo Core in action. Source: AMD.

First, though, we should talk about Bulldozer’s version of AMD’s Turbo Core dynamic clock scaling feature, which raises clock speed on all or part of the chip when there’s thermal headroom available to do so. As in other recent AMD CPUs, Turbo Core uses power estimates based on the chip’s internal activity monitoring to determine the extent of that thermal headroom. Bulldozer’s Turbo Core implementation is the most granular one yet, with three P-states possible. P2 is the base clock of the chip, the speed at which it’s guaranteed to run. P1 is an intermediate Turbo clock speed that can apply to all four modules, provided that they’re not too heavily loaded. The third state, P0, is an even higher Turbo clock that comes into use when only two modules are active. As before, Turbo Core seeks to run at the highest possible clock speed for the given conditions, and it dithers between the P-states in order to stay within the chip’s prescribed thermal envelope, or TDP.

Example of AMD’s preferred thread scheduling. Source: AMD.

Now that we have Turbo Core in the picture, we have the context to talk more about thread scheduling. Bulldozer’s unique architecture creates some intriguing questions about how software threads should be distributed across its cores. There are obvious advantages to scheduling one thread per module before doubling up threads on a single module: shared resources like the front end, L2 cache, and FPU will be dedicated to a lone thread, improving performance. However, scheduling two threads per module gets you several nice things, too, including the possibility of data sharing between related threads via the L2 cache. Power efficiency should improve if more inactive modules can be turned off, and Turbo Core can convert that power savings back into performance by raising the clock speed of the active module.

Unfortunately, the Windows 7 scheduler wasn’t built with Bulldozer’s distinctive sharing arrangement in mind, and as far as we call tell, the BIOS doesn’t provide any hints to that OS about how to schedule threads. Win7 simply sees eight equal cores, with no preference between them. AMD claims Windows 8 will be better optimized for the Bulldozer architecture and cites improvements of 2-10% in several recent games with the Windows 8 developer preview. We haven’t been able to squeeze too many details out of AMD about how complex Win8’s understanding of Bulldozer scheduling will be, but we get the sense that the OS may attempt to schedule related threads on the same module when possible. We need to play with the Win8 developer preview on a Bulldozer system in order to learn more.

The chip: Orochi

Like mythical heroes in fantasy novels, modern CPUs are known by many names. We’ve been talking about Bulldozer almost exclusively up to this point, but that code name actually applies to the CPU cores and the microarchitecture inside of them—or something like that. These names are powerful symbols and are often multi-valent. (Yikes, religion major mode OFF. Sorry.) The proper code name for the silicon die that implements the Bulldozer architecture is “Orochi,” and Orochi will be deployed in multiple ways, each with its own name. On the desktop, it’s called “Zambezi.” In 1-2P servers, a single Orochi die will be called “Valencia,” and in 1-4P servers, two dies placed together in a package will be called “Interlagos.” I liked it better when a single name, like K7, could refer to the whole caboodle, before the marketing guys got into the code-name business, but I suppose that horse left the barn long ago.

Whatever you call it, this chip is AMD’s second attempt at a CPU fabricated on GlobalFoundries’ 32-nm process, with high-k metal gates and a silicon-on-insulator substrate. The unnecessarily overpopulated table below shows how Orochi compares to a range of other desktop processors from Intel and AMD.

Code name Key

products

Cores Threads Last-level

cache size

Process node

(Nanometers)

Estimated

transistors

(Millions)

Die

area

(mm²)

Bloomfield Core i7 4 8 8 MB 45 731 263
Lynnfield Core i5, i7 4 8 8 MB 45 774 296
Westmere Core i3, i5 2 4 4 MB 32 383 81
Gulftown Core i7-980X 6 12 12 MB 32 1168 248
Sandy Bridge Core i5, i7 4 8 8 MB 32 995 216
Sandy Bridge Core i3, i5 2 4 4 MB 32 624 149
Sandy Bridge Pentium 2 4 3 MB 32 131
Deneb Phenom II 4 4 6 MB 45 758 258
Propus/Rana Athlon II X4/X3 4 4 512 KB x 4 45 300 169
Regor Athlon II X2 2 2 1 MB x 2 45 234 118
Thuban Phenom II X6 6 6 6 MB 45 904 346
Llano A8, A6, A4 4 4 1MB x 4 32 1450 228
Llano A4 2 2 1MB x 2 32 758
Orochi/Zambezi FX 8 8 8MB 32 1200 315

With roughly 1.2 billion transistors and a die area of 315 mm², Orochi is a very big and complex chip. Sandy Bridge, which has four cores and integrated graphics, is about 100 mm² smaller. Still, Orochi isn’t quite a large as the chip it succeeds, the “Thuban” Phenom II X6, so that’s progress of a sort.

The FX-series processors

Now that we’ve explored Bulldozer’s many code names, we should take a look at the names of the products that, you know, actual people will buy. AMD is introducing a trio of Bulldozer-based products today, and we have their vitals in the table below.

Model Cores Base core

clock speed

Turbo

clock speed

Peak Turbo

clock speed

L3 cache

size

Memory

channels

TDP Price
FX-6100 6 3.3 GHz 3.6 GHz 3.9 GHz 6 MB 2 95 W $165
FX-8120 8 3.1 GHz 3.4 GHz 4.0 GHz 8 MB 2 125 W $205
FX-8150 8 3.6 GHz 3.9 GHz 4.2 GHz 8 MB 2 125 W $245

As you can see, the clock speeds involved aren’t stratospheric, but the FX-8150’s peak of 4.2GHz is a fair bit higher than anything else offered by AMD or Intel these days. These products are targeted directly opposite Intel’s Sandy Bridge parts, so let’s have a look at the competition’s lineup for comparison.

Model Cores Threads Base core

clock speed

Peak Turbo

clock speed

L3 cache

size

Memory

channels

TDP Price
Core i3-2100 2 2 3.1 GHz 3 MB 2 65 W $117
Core i5-2320 4 4 3.0 GHz 3.3 GHz 6 MB 2 95 W $177
Core i5-2400 4 4 3.1 GHz 3.4 GHz 6 MB 2 95 W $184
Core i5-2500 4 4 3.3 GHz 3.7 GHz 6 MB 2 95 W $205
Core i5-2500K 4 4 3.3 GHz 3.7 GHz 6 MB 2 95 W $216
Core i7-2600K 4 8 3.4 GHz 3.8 GHz 8 MB 2 95 W $317

We’ve taken this table almost without modification from our original Sandy Bridge review early this year. Intel hasn’t lowered prices on its key products once since then. The only real changes have been the additions of models to fill gaps in the original lineup, such as the addition of the Core i5-2120 at $177. We’ve not listed every single Sandy Bridge model above, since there are so very many. We think the ones included are the most relevant for our purposes today.

You’ll notice several things about these competing lineups right away. For one, AMD has made no attempt to go after the highest-end Sandy Bridge part, the Core i7-2600K, with a Bulldozer-based offering. We expect AMD would have liked to compete at that level, but doing so apparently wasn’t feasible at present. Similarly, AMD hasn’t attempted to take on Intel’s high-end Core i7-900-series processors. Also, notice that the top two Bulldozer-derived models are rated for 125W power envelopes, while the fastest Sandy Bridge chips have a TDP of 95W. Apparently, AMD needs the extra thermal headroom in order to compete on price and performance with the Intel products it has targeted.

With that said, the competitive match-ups are still reasonably straightforward. At $245, the FX-8150 is priced a bit above the Core i5-2500K, but the two are clearly rivals. The FX-8120 is in an interesting spot, taking on the Core i5-2500, the non-K-series version of that product, with a locked multiplier. Finally, the FX-6100 has one module disabled and a 95W TDP, and it’s priced opposite the Core i5-2320.

Unfortunately, we’ve only managed to get our hands on one of the three initial FX-series products today, the FX-8150. You’ll see a full set of results for it on the following pages, versus a host of other CPUs. We don’t have a real FX-8120, but we did attempt to simulate its clock speeds and performance using AMD’s Overdrive software. We’re only somewhat confident that we’ve managed to do so successfully, but we have provisionally included some performance results for the FX-8120. Take ’em with a grain of salt, and we’ll attempt to replace them with results from the real product when we can.

One more thing. In order to sweeten the pot a little, AMD has decided to unlock the multipliers on all three models of FX-series processors. That should make overclocking relatively easy to do, and it should give AMD a leg up in cases where the FX chips aren’t competing with unlocked K-series parts from Intel.

The platform: Socket AM3+

As we’ve mentioned, Bulldozer-based CPUs should be compatible AMD’s existing socket infrastructure. On the desktop, that’s AMD’s Socket AM3+ platform, which the company introduced back in May alongside its 9-series chipsets. FX-series processors have 942 pins, one more than older Socket AM3 CPUs, and that pin prevents them from dropping into anything but true Socket AM3+ motherboards. On the flip side, Socket AM3+ boards are capable of hosting older Socket AM3 processors like the Phenom II just fine.

The FX-8150 installed in our Asus Crosshair V mobo

If you have an existing Socket AM3+ system with an older CPU and would like to upgrade to an FX CPU, that should be possible after a quick BIOS update. The trick is that users may need to flash their BIOSes to add Bulldozer support using an older CPU before installing the new processor. That requirement generally shouldn’t be a big deal for would-be upgraders, but folks who are buying new motherboards for use with an FX processor will have to hope they receive a board with an FX-capable BIOS. Otherwise, unhappy times may ensue. On that subject, AMD tells us motherboards have been shipping with Bulldozer-ready BIOSes “for some time now,” and it expects any such problems with new mobos to be rare.

The FX-8150’s pins and AM3+ socket

Folks who own a Socket AM3+ system with an older CPU and don’t plan to upgrade will want to be careful about BIOS upgrades, too. A major mobo maker told us recently that BIOS/EFI space for adding Bulldozer support is cramped, so some features aimed at older Athlon II and Phenom II processors, such as core unlockers, may have to be deleted from newer BIOSes in order to make room. Owners of those older CPUs may want to avoid the impulse to update to the latest firmware automatically. They may be better off sticking to an older version with a full feature set for Athlon II and Phenom II CPUs. As always, you’ll want to check with your motherboard maker for the final word on your board’s compatibility story.

Happily, dropping an FX-series processor into a Socket AM3+ motherboard will prompt an upgrade of sorts: the board’s two memory channels will then support the latest in DDR3 memory speeds, up to 1866MHz. Both Bulldozer and Llano officially support those higher memory frequencies. Intel’s K-series Sandy Bridge parts are capable of working with faster RAM, too, but memory speeds above 1333MHz aren’t officially blessed.

Our testing methods

We ran every test at least three times and reported the median of the scores produced.

The test systems were configured like so:

Processor
Athlon II X3 455 3.3GHz

Phenom II X2 565 3.4GHz

Phenom II X4 840 3.2GHz

Phenom II X4 975 3.6GHz

Phenom II X4 980 3.7GHz

Phenom II X6 1075T 3.0GHz

Phenom II X6 1100T 3.3GHz

Pentium
Extreme Edition 840 3.2GHz
Pentium
G6950 2.8GHz
AMD
FX-8120 3.1GHz

AMD FX-8150 3.6GHz

Core
i7-990X 3.46 GHz
Core
2 Duo E6400 2.13GHz
Core
i3-560 3.33 GHz

Core i5-655K 3.2GHz

Core i5-760 2.8GHz

Core i7-875K 2.93GHz

Core
2 Quad Q9400 2.67GHz
Motherboard Gigabyte
890GPA-UD3H
Asus
P5E3 Premium
Asus
P7P55D-E Pro
Asus
Crosshair V Formula 
Intel
DX58SO2
North bridge 890GX X48 P55 990FX X58
South bridge SB850 ICH9R SB850 ICH10R
Memory size 8GB
(4 DIMMs)
8GB
(4 DIMMs)
8GB
(4 DIMMs)
8GB
(2 DIMMs)
12GB
(6 DIMMs)
Memory type Corsair

CMD8GX3M
4A1333C7

DDR3 SDRAM

Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CMZ8GX3M
2A1866C9

DDR3 SDRAM

Corsair

CMP12GX3M
6A1600C8

DDR3 SDRAM

Memory speed 1333 MHz 800
MHz
1066 MHz 1866 MHz 1333 MHz
1066 MHz
1333 MHz
1333 MHz
Memory timings 8-8-8-20 2T 7-7-7-20 2T 7-7-7-20 2T 9-10-0-27 2T 8-8-8-20 2T
7-7-7-20 2T
8-8-8-20 2T
8-8-8-20 2T
Chipset

drivers

AMD
AHCI 1.2.1.263
INF
update 9.1.1.1025

Rapid Storage Technology 9.6.0.1014

INF
update 9.1.1.1025

Rapid Storage Technology 9.6.0.1014

AMD
AHCI 1.2.1.301
INF update 9.1.1.1020

Rapid Storage Technology 9.5.0.1037

Audio Integrated

SB850/ALC892 with

Realtek 6.0.1.6235 drivers

Integrated

ICH9R/AD1988B with 

Microsoft drivers

Integrated

P55/RTL8111B with

Realtek 6.0.1.6235 drivers

Integrated

SB850/ALC889 with

Realtek 6.0.1.6235 drivers

Integrated

ICH10R/ALC892 with

Realtek 6.0.1.6235 drivers

Processor Core
i7-950 3.06 GHz

Core i7-970 3.2 GHz

Core i7-980X Extreme 3.3 GHz

Core
i3-2100 2.93 GHz

Core i5-2400 3.1 GHz

Core i5-2500K 3.3 GHz

Core i7-2600K 3.4 GHz

AMD
A8-3800 2.4GHz

AMD
A8-3850 2.9 GHz

Atom
D525 1.8 GHz
AMD
E-350 1.6GHz
Motherboard Gigabyte
X58A-UD5
Asus
P8P67 Deluxe
Gigabyte
A75M-UD2H
Jetway
NC94FL-525-LF
MSI
E350IA-E45
North bridge X58 P67 A75 NM10 Hudson
M1
South bridge ICH10R
Memory size 12GB
(6 DIMMs)
8GB
(4 DIMMs)
8GB
(4 DIMMs)
4GB (2 DIMMs) 4GB (2 DIMMs)
Memory type Corsair

CMP12GX3M
6A1600C8

DDR3 SDRAM

Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CM2X2048-
8500C5D

DDR2 SDRAM

Corsair

CMD8GX3M
4A1333C7

DDR3 SDRAM

Memory speed 1333 MHz 1333 MHz 1333 MHz 800
MHz
1066 MHz
Memory timings 8-8-8-20 2T 8-8-8-20 2T 8-8-8-20 2T 5-5-5-18
2T
7-7-7-20 2T
Chipset

drivers

INF update 9.1.1.1020

Rapid Storage Technology 9.5.0.1037

INF update
9.2.0.1016

Rapid Storage Technology 10.0.0.1046

AMD
AHCI 1.2.1.296

AMD USB 3.0 1.0.0.52

INF update 9.1.1.1020

Rapid Storage Technology 9.5.0.1037

AMD
AHCI 1.2.1.275
Audio Integrated

ICH10R/ALC889 with

Realtek 6.0.1.6235 drivers

Integrated

P67/ALC889 with

Microsoft drivers

Integrated

A75 FCH/ALC889 with

Realtek 6.0.1.6235 drivers

Integrated

NM10/ALC662 with

Realtek 6.0.1.6235 drivers

Integrated

Hudson M1/ALC887 with

Realtek 6.0.1.6235 drivers

They all shared the following common elements:

Hard drive Corsair
Nova V128 SATA SSD
Discrete graphics Asus
ENGTX460 TOP 1GB (GeForce GTX 460) with ForceWare 260.99 drivers
OS Windows 7 Ultimate x64 Edition
Power supply PC Power & Cooling Silencer 610 Watt

Thanks to Asus, Corsair, Gigabyte, and OCZ for helping to outfit our test rigs with some of the finest hardware available. Thanks to Intel and AMD for providing the processors, as well, of course.

The test systems’ Windows desktops were set at 1900×1200 in 32-bit color. Vertical refresh sync (vsync) was disabled in the graphics driver control panel.

We used the following versions of our test applications:

Some further notes on our testing methods:

  • Many of our performance tests are scripted and repeatable, but for some of the games, including Battlefield: Bad Company 2, we used the Fraps utility to record frame rates while playing a 60-second sequence from the game. Although capturing frame rates while playing isn’t precisely repeatable, we tried to make each run as similar as possible to all of the others. We raised our sample size, testing each Fraps sequence five times per video card, in order to counteract any variability. We’ve included second-by-second frame rate results from Fraps for those games, and in that case, you’re seeing the results from a single, representative pass through the test sequence.
  • We used a Yokogawa WT210 digital power meter to capture power use over a span of time. The meter reads power use at the wall socket, so it incorporates power use from the entire system—the CPU, motherboard, memory, graphics solution, hard drives, and anything else plugged into the power supply unit. (The monitor was plugged into a separate outlet.) We measured how each of our test systems used power across a set time period, during which time we ran Cinebench’s multithreaded rendering test.
  • After consulting with our readers, we’ve decided to enable Windows’ “Balanced” power profile for the bulk of our desktop processor tests, which means power-saving features like SpeedStep and Cool’n’Quiet are operating. (In the past, we only enabled these features for power consumption testing.) Our spot checks demonstrated to us that, typically, there’s no performance penalty for enabling these features on today’s CPUs. If there is a real-world penalty to enabling these features, well, we think that’s worthy of inclusion in our measurements, since the vast majority of desktop processors these days will spend their lives with these features enabled. We did disable these power management features to measure cache latencies, but otherwise, it was unnecessary to do so.

The tests and methods we employ are usually publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory subsystem performance

Our first few tests are synthetic benchmarks that let us inspect the performance of the cache and memory subsystems.

This test is nicely multithreaded, so the caches from all available cores contribute to the throughput measured. You may be surprised to see that the Phenom II X6 1100T achieves higher bandwidth than the FX-8150 at the smaller block sizes, but remember it has six L1 caches where the FX has four. The more apt comparison may be the Phenom II X4 980, with four cores and a 3.7GHz clock frequency. The FX’s L1 caches will cover block sizes up to 64KB, and the FX-8150 is faster than the Phenom II X4 980 at each step from 2KB to 64KB. Then again, with only four cores, Sandy Bridge’s L1 caches are faster still.

The 256KB to 1MB block sizes are L2 cache territory, and the FX’s L2 caches don’t look to be especially fast, either, though they do largely outperform the Phenom II X4 980’s. Bulldozer’s L2 caches may lack for speed, but they’re large. At the 4MB data point, the rest of the CPUs are into their L3 caches. The FX is still in its L2 coverage area. The next step up in block size is 16MB, which is right at the outer edge of the FX’s effective total cache capacity, since its 8MB of L3 cache doesn’t replicate the contents of its 8MB of L2 cache. The FX-8150 again delivers the highest throughput at the 16MB block size, but not by much.

Some of the credit for the FX-8150’s strong showing here no doubt goes to its use of 1866MHz DIMMs. However, we’ve tried 1866MHz memory on the older CPU cores in Llano, and our Stream results topped out at around 15GB/s. Bulldozer’s smart data prefetchers and large L2 caches deserve credit for taking good advantage of the available memory bandwidth.

Measuring memory access latencies has gotten to be tricky with the advent of Turbo-style clock speed ramping, because latencies are reported in the number of CPU cycles. Nevertheless, we’ve chosen to report access latencies with the caveat that our guesses about likely frequencies for these CPUs may be incorrect.

If we’re right, the FX comes out looking pretty good, with access latencies comparable to competing Sandy Bridge parts, despite its larger caches. Again, the use of 1866MHz memory may be helping the FX here.

For what it’s worth, our tool reports Bulldozer’s L1 data cache latency at 3 cycles, L2 at 18 cycles, and L3 at 65 cycles.

Battlefield: Bad Company 2

After a promising start in our synthetic memory tests, we finally get our first look at the FX-8150’s real-world performance—and it’s a bit underwhelming. The FX-8150 is no faster than the Phenom II X4 980, and it’s slower than quite a few Intel processors, even somewhat older models.

Of course, a result like this one should come with a reminder: a frame rate of 75 FPS, with a low of 58 FPS, means the FX-8150 is easily competent to run this game smoothly—at least, as far as we can tell when measuring in frames per second. We have some promising new testing methods we may bring to bear on CPU performance soon, so stay tuned.

Civilization V

The developers of Civ V have cooked up a number of interesting benchmarks, two of which we used here. The first one tests a late-game scenario where the map is richly populated and there’s lots happening at once. As you can see by the setting screen below, we didn’t skimp on our the image quality settings for graphics, either. Doing so wasn’t necessary to tease out clear differences between the CPUs.

Civ V also runs the same test without updating the screen, so we can eliminate any overhead or bottlenecks introduced by the video card and its driver software. Removing those things from the equation reshuffles the order slightly.

In both cases, the FX-8150 slots in just ahead of the Phenom II X6 1100T and just behind an aging Intel CPU, the Core i5-760. All of the Sandy Bridge-based CPUs are faster, including the dual-core Core i3-2100.

The next test populates the screen with a large number of units and animates them all in parallel. It can also run in “no render” mode without updating the screen.

This test is clearly multithreaded—it’s much faster in “no render” mode on the Athlon II X3 455 than on the Phenom II X2 565, for instance, and the 12-threaded Core i7-900-series CPUs capture the top three spots. Still, the FX-8150 and its eight cores end up near the middle of the pack. When the screen is being rendered, a number of Phenom II X6 and X4 models are slightly faster than the FX-8150.

F1 2010

CodeMasters has done a nice job of building benchmarks into its recent games, and F1 2010 is no exception. We scripted up test runs at three different display resolutions, with some very high visual quality settings, to get a sense of how much difference a CPU might make in a real-world gaming scenario where GPU bottlenecks can come into play.

We also went to some lengths to fiddle with the game’s multithreaded CPU support in order to get it to make the most of each CPU type. That effort eventually involved grabbing a couple of updated config files posted on the CodeMasters forum, one from the developers and another from a user, to get an optimal threading map for the Phenom II X6. What you see below should be the best possible performance out of each processor.

The results at the two higher resolutions underscore a dynamic that AMD and Nvidia have been banging on about: if your GPU is the primary performance limiter, a CPU upgrade may not do you much good. That’s a noteworthy practical point, but we’ll still want to focus on the lower-resolution results in order to compare CPU performance.

At 1280×800, both the FX-8150 and our “pretend” FX-8120 are faster than any other processor AMD has fielded to date, and the FX-8150 isn’t far from hitting that GPU wall at around 65 FPS.

Metro 2033
Metro 2033 also offers a nicely scriptable benchmark, and we took advantage by testing at several different combinations of resolution and visual quality.

Again, the lower resolution tests lets us see the impact of CPU performance more clearly. The FX-8150 doesn’t look half bad in those tests, either, delivering slightly higher frame rates than the Phenom II X6 1100T and mixing it up with the Core i5-2400.

Source engine particle simulation

Next up is a test we picked up during a visit to Valve Software, the developers of the Half-Life games. They had been working to incorporate support for multi-core processors into their Source game engine, and they cooked up some benchmarks to demonstrate the benefits of multithreading.

This test runs a particle simulation inside of the Source engine. Most games today use particle systems to create effects like smoke, steam, and fire, but the realism and interactivity of those effects are limited by the available computing horsepower. Valve’s particle system distributes the load across multiple CPU cores.

Honestly, I expected FX processors to perform well in this test. This program is widely multithreaded, and Intel CPUs seem to benefit greatly from Hyper-Threading when running it. Seems a likely target for Bulldozer, no? However, the FX-8150 once more trails the Phenom II X6 1100T, which itself is well behind the FX-8150’s ostensible competitor, the Core i5-2500K.

Productivity

SunSpider JavaScript performance

Several of AMD’s older Phenom II processors outperform the FX processors in this test, as do over half of the CPUs we’ve tested. The FX-8150 is only slightly quicker than the lowly Athlon II X3 455.

7-Zip file compression and decompression

Here’s a nice bright spot where the FX-8150 runs with and even defeats the top Sandy Bridge, the Core i7-2600K. AMD’s older processors also perform fairly well in this test, interestingly enough.

TrueCrypt disk encryption

This full-disk encryption suite includes a performance test, for obvious reasons. We tested with a 500MB buffer size and, because the benchmark spits out a lot of data, averaged and summarized the results in a couple of different ways.

TrueCrypt has added support for Intel’s custom-tailored AES-NI instructions since we last visited it, so the encoding of the AES algorithm, in particular, should be very fast on the CPUs that support those instructions. Those CPUs include the six-core Gulftowns, the dual-core Clarkdales, Sandy Bridge, and of course Bulldozer.

Bulldozer’s support of AES-related instructions pays off nicely for it in TrueCrypt. At these encryption rates, few storage devices will be able to keep up.

Image processing

The Panorama Factory photo stitching
The Panorama Factory handles an increasingly popular image processing task: joining together multiple images to create a wide-aspect panorama. This task can require lots of memory and can be computationally intensive, so The Panorama Factory comes in a 64-bit version that’s widely multithreaded. I asked it to join four pictures, each eight megapixels, into a glorious panorama of the interior of Damage Labs.

In the past, we’ve added up the time taken by all of the different elements of the panorama creation wizard and reported that number, along with detailed results for each operation. However, doing so is incredibly data-input-intensive, and the process tends to be dominated by a single, long operation: the stitch. Thus, we’ve simply decided to report the stitch time, which saves us a lot of work and still gets at the heart of the matter.

Image stitching would seem to be a natural fit for Bulldozer’s eight integer cores and high memory throughput, and the FX processors do improve on the performance of the Phenom II X6. Still, the FX-8150 can’t catch the Core i5-2400, which is a $184 product. The FX-8150 costs $245.

picCOLOR image processing and analysis

picCOLOR was created by Dr. Reinert H. G. Müller of the FIBUS Institute. This isn’t Photoshop; picCOLOR’s image analysis capabilities can be used for scientific applications like particle flow analysis. Dr. Müller has supplied us with new revisions of his program for some time now, all the while optimizing picCOLOR for new advances in CPU technology, including SSE extensions, multiple cores, and Hyper-Threading. Many of its individual functions are multithreaded.

At our request, Dr. Müller graciously agreed to re-tool his picCOLOR benchmark to incorporate some real-world usage scenarios. As a result, we now have four tests that employ picCOLOR for image analysis: particle image velocimetry, real-time object tracking, a bar-code search, and label recognition and rotation.

The FX-8150 delivers measurable gains over the Phenom II X6 1100T in picCOLOR’s real-world tests, but it’s still not as quick as the Core i5-2500K.

picCOLOR also includes some synthetic tests of common image processing functions, and the FX-8150 proves to be substantially faster than older AMD CPUs in these tests. Again, though, the slowest Sandy Bridge quad-core is faster.

Video encoding

x264 HD benchmark

This benchmark tests one of the most popular H.264 video encoders, the open-source x264. The results come in two parts, for the two passes the encoder makes through the video file. I’ve chosen to report them separately, since that’s typically how the results are reported in the public database of results for this benchmark.

Another ray of light for Bulldozer here, in the more multithreaded second pass of the x264 encoding process. The FX-8150 matches the pricier Core i7-2600K—and clearly outruns the i5-2500K. Too bad about that first pass, which is also part of the overall picture.

Windows Live Movie Maker 14 video encoding

For this test, we used Windows Live Movie Maker to transcode a 30-minute TV show, recorded in 720p .wtv format on my Windows 7 Media Center system, into a 320×240 WMV-format video format appropriate for mobile devices.

Live Movie Maker combines both passes into a single encoding time metric, and it places the FX-8150 just a few seconds ahead of the Phenom II X4 980.

3D modeling and rendering

Cinebench rendering

The Cinebench benchmark is based on Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. This test runs with just a single thread and then with as many threads as CPU cores (or threads, in CPUs with multiple hardware threads per core) are available.

Here’s another one of those tests where the FX-8150 improves on the Phenom II X6, but then the X6’s performance was fairly strong compared to the Core i5-2500K already. The most interesting result, in my view, is the FX-8150’s fairly low score of 1.03 with a single thread, a little less than the score of the Phenom II X6 1100T, which runs at a lower clock frequency. Bulldozer hasn’t entirely held the line on instructions per clock. Also, the Core i5-2500K is nearly 50% faster than the FX-8150 with a single thread. Bulldozer simply makes up the difference with more threads—at least, in this case it does.

POV-Ray rendering

We’re using the latest beta version of POV-Ray 3.7 that includes native multithreading and 64-bit support.

Another solid showing from the FX-8150 in the “chess2” test, which is more widely multithreaded than POV-Ray’s benchmark scene.

Valve VRAD map compilation

This next test processes a map from Half-Life 2 using Valve’s VRAD lighting tool. Valve uses VRAD to pre-compute lighting that goes into games like Half-Life 2.

The FX slips a little here, falling behind the Phenom II X6 yet again.

Scientific computing

MyriMatch proteomics

Our benchmarks sometimes come from unexpected places, and such is the case with this one. David Tabb is a friend of mine from high school and a long-time TR reader. He has provided us with an intriguing new benchmark based on an application he’s developed for use in his research work. The application is called MyriMatch, and it’s intended for use in proteomics, or the large-scale study of protein. I’ll stop right here and let him explain what MyriMatch does:

In shotgun proteomics, researchers digest complex mixtures of proteins into peptides, separate them by liquid chromatography, and analyze them by tandem mass spectrometers. This creates data sets containing tens of thousands of spectra that can be identified to peptide sequences drawn from the known genomes for most lab organisms. The first software for this purpose was Sequest, created by John Yates and Jimmy Eng at the University of Washington. Recently, David Tabb and Matthew Chambers at Vanderbilt University developed MyriMatch, an algorithm that can exploit multiple cores and multiple computers for this matching. Source code and binaries of MyriMatch are publicly available.
In this test, 5555 tandem mass spectra from a Thermo LTQ mass spectrometer are identified to peptides generated from the 6714 proteins of S. cerevisiae (baker’s yeast). The data set was provided by Andy Link at Vanderbilt University. The FASTA protein sequence database was provided by the Saccharomyces Genome Database.

MyriMatch uses threading to accelerate the handling of protein sequences. The database (read into memory) is separated into a number of jobs, typically the number of threads multiplied by 10. If four threads are used in the above database, for example, each job consists of 168 protein sequences (1/40th of the database). When a thread finishes handling all proteins in the current job, it accepts another job from the queue. This technique is intended to minimize synchronization overhead between threads and minimize CPU idle time.

The most important news for us is that MyriMatch is a widely multithreaded real-world application that we can use with a relevant data set. I should mention that performance scaling in MyriMatch tends to be limited by several factors, including memory bandwidth, as David explains:

Inefficiencies in scaling occur from a variety of sources. First, each thread is comparing to a common collection of tandem mass spectra in memory. Although most peptides will be compared to different spectra within the collection, sometimes multiple threads attempt to compare to the same spectra simultaneously, necessitating a mutex mechanism for each spectrum. Second, the number of spectra in memory far exceeds the capacity of processor caches, and so the memory controller gets a fair workout during execution.

Here’s how the processors performed.

This is another case where I definitely expected the FX-8150 to improve on the performance of the Phenom II X6. Alas, it wasn’t to be.

STARS Euler3d computational fluid dynamics

Charles O’Neill works in the Computational Aeroservoelasticity Laboratory at Oklahoma State University, and he contacted us to suggest we try the computational fluid dynamics (CFD) benchmark based on the STARS Euler3D structural analysis routines developed at CASELab. This benchmark has been available to the public for some time in single-threaded form, but Charles was kind enough to put together a multithreaded version of the benchmark for us with a larger data set. He has also put a web page online with a downloadable version of the multithreaded benchmark, a description, and some results here.

In this test, the application is basically doing analysis of airflow over an aircraft wing. I will step out of the way and let Charles explain the rest:

The benchmark testcase is the AGARD 445.6 aeroelastic test wing. The wing uses a NACA 65A004 airfoil section and has a panel aspect ratio of 1.65, taper ratio of 0.66, and a quarter-chord sweep angle of 45º. This AGARD wing was tested at the NASA Langley Research Center in the 16-foot Transonic Dynamics Tunnel and is a standard aeroelastic test case used for validation of unsteady, compressible CFD codes.
The CFD grid contains 1.23 million tetrahedral elements and 223 thousand nodes . . . . The benchmark executable advances the Mach 0.50 AGARD flow solution. A benchmark score is reported as a CFD cycle frequency in Hertz.

So the higher the score, the faster the computer. Charles tells me these CFD solvers are very floating-point intensive, but they’re oftentimes limited primarily by memory bandwidth. He has modified the benchmark for us in order to enable control over the number of threads used. Here’s how our contenders handled the test with optimal thread counts for each processor.

The FX-8150’s gains over the Phenom II X6 here could easily be explained by the move to faster memory alone. Unfortunately for AMD, there’s a whole range of Intel CPUs from various generations that outperform the FX, even with slower DRAM.

Power consumption and efficiency

We used a Yokogawa WT210 digital power meter to capture power use over a span of time. The meter reads power use at the wall socket, so it incorporates power use from the entire system—the CPU, motherboard, memory, graphics solution, hard drives, and anything else plugged into the power supply unit. (The monitor was plugged into a separate outlet.) We measured how each of our test systems used power across a set time period, during which time we ran Cinebench’s multithreaded rendering test.

Please note that, although we tested a range of AMD processors, only the FX-8150 and the Phenom II X6 (the results marked “990FX”) were tested on the same motherboard. The others were tested on an 890GX-based board from Gigabyte whose power consumption characteristics differ. Oh, and we tested the FX-8150 with four DIMMs here, since that’s the config all of the other dual-channel systems shared, and it only seemed fair to match the DIMM count for power testing. Fortunately, the move to lower memory clocks didn’t impact rendering completion times.

We’ll start with the show-your-work stuff, plots of the raw power consumption readings.

We can slice up these raw data in various ways in order to better understand them. We’ll start with a look at idle power, taken from the trailing edge of our test period, after all CPUs have completed the render. Next, we can look at peak power draw by taking an average from the ten-second span from 15 to 25 seconds into our test period, when the processors were rendering.

At idle, the 990FX-based sytem with a Phenom II X6 processor draws 94W at the wall socket. Simply as a result of dropping in an FX-8150 instead, idle power consumption plummets to 76W. That’s the impact of Bulldozer’s ability to gate off power to its idle cores and north bridge. Also, notice that the Phenom II X6 draws even less power at idle on the 890GX board: 82W. Our 990FX may be a bit of a power hog at idle. We’ll have to swap the FX-8150 into a different Socket AM3+ board soon and see if we can’t get it down into the mid-60-watt range, like the Sandy Bridge systems. Seems possible.

At peak, the FX-8150 system draws about 22W more than the same system equipped with a Phenom II X6 1100T. That’s not an entirely shocking result, even though the chips have the same TDP rating. Bulldozer’s improved Turbo Core is more effective at wringing every last ounce out of a given thermal envelope, and that translates into higher measured power draw at peak. With a more modest 95W thermal envelope, the Sandy Bridge-based competition delivers its superior performance at much lower total system power levels.

We can highlight power efficiency by looking at total energy use over our time span. This method takes into account power use both during the render and during the idle time. We can express the result in terms of watt-seconds, also known as joules. (In this case, to keep things manageable, we’re using kilojoules.) Note that since we had to expand the duration of the test periods for the Pentium EE 840 and Core 2 Duo E6400, we’re including data from a longer period of time for those two.

We can pinpoint efficiency more effectively by considering the amount of energy used for the task alone. Since the different systems completed the render at different speeds, we’ve isolated the render period for each system. We’ve then computed the amount of energy used by each system to render the scene. This method should account for both power use and, to some degree, performance, because shorter render times may lead to less energy consumption.

Even though it ostensibly benefits from finer process technology, the FX-8150 actually uses slightly more power than the Phenom II X6 1100T does in rendering this scene. That’s not the sort of progress we had hoped to see.

AVX performance

None of the benchmarks you’ve seen on the preceding pages make of use AVX instructions, with the exception of the AES subset used in TrueCrypt. At this point in time, finding applications or benchmarks that make use of AVX isn’t easy. Fortunately, I was able to find several, and at least one program looks to be reasonably well optimized for Bulldozer: AIDA64 from FinalWire. AIDA64 includes several small, synthetic tests that can be accelerated with Bulldozer’s new instructions.

In order to measure the impact of those new instructions on performance, I tested both the FX-8150 and the Core i5-2500K in two configurations: with and without Windows 7 SP1 installed. Service Pack 1 is required for AVX support, so testing without it has the impact of disabling AVX. The results marked “No AVX” below are those without SP1.

The first test, CPU Hash, uses AMD’s XOP instructions on Bulldozer. The next two, FPU Julia and FPU Mandel, make use of FMA4. Tamas Miklos of FinalWire, maker of AIDA64, tells us these benchmarks were developed using pre-release Bulldozer systems. He further says:

Our code is 100% optimized for Bulldozer. We don’t see much room for improvement. We’ve had a chance to talk to AMD, and we’ve explained (in details) how our benchmarks work, and what tricks we use on Bulldozer. They seemed to be content about what we do and how we do on Bulldozer, and they didn’t tell us any hints on possible improvements.

So we should have a resonable opportunity to see Bulldozer’s full potential with the latest instructions.

Here’s another one of those occasional instances where the Phenom II X6 was already faster than Sandy Bridge, and again, the FX-8150 is a little faster still. Looks like there’s a roughly 10% gain with AVX/XOP instructions in use.

The two tests above make use of FMA4, and these really aren’t the sort of results we were anticipating. In both cases, the Phenom II X6 110T is faster than the FX-8150 with AVX and FMA4 enabled. Hrmph.

We do have another round of AVX tests, from the latest version of SiSoft’s Sandra. We don’t have any word on how well these tests are optimized for Bulldozer or whether they use XOP and FMA4. They do appear to make use of AVX on the FX-8150, though.

Although these quick benchmarks are labeled “Multimedia” in Sandra, in truth they’re simply fractal computations like the AIDA64 Julia and Mandel tests.

Well, Bulldozer looks great in the integer test, but the FPU results don’t look much different that they did in AIDA64. At least the FX-8150 is faster than the Phenom II X6, I guess.

We’re disheartened by these results, but AVX is still early in its life, so we’re hesitant to draw any definitive lessons from them. AMD did supply us with custom-built versions of the x264 video encoder that use XOP and FMA4 late in the review process. We’ll have to try those out soon.

Overclocking

Given that this is a “speed demon” architecture and that AMD managed a Guinness world record frequency of 8.4GHz using an FX-8150 cooled with liquid helium, we were really looking forward to doing a little overclocking with our copy of the FX-8150. We didn’t have any dangerous liquids in play, but we did have a fairly beefy tower cooler, at least. We set our mobo to use its most aggressive fan speed profiles and fired up AMD’s Overdrive software, which makes dialing up new speeds on an unlocked CPU like this one a snap, to see what we could do.

Our starting point was the stock operation of the chip. Our FX-8150 runs at 3.6GHz and 1.2625V by default. When Turbo Core kicks in, the CPU ranges up to 1.4V and 4.2GHz. We figured we’d begin at just 200MHz beyond that top Turbo speed, 4.4GHz, at 1.4V. Seems like an easy first step, right? When we fired off Overdrive’s CPU stability test, however, it quickly came back with an error. We had to raise the voltage to 1.425V in order to get the chip to pass just three minutes in that stability test. The rest of our overclocking work log looked like so:

4.6GHz, 1.425V – BSOD

4.6GHz, 1.4375V – Error

4.6GHz, 1.45V – AOD crash

4.6GHz, 1.4625V – AOD crash

4.6GHz, 1.4625V – AOD crash

4.5GHz, 1.4375V – AOD crash

4.5GHz, 1.45V – Error

4.5GHz, 1.4625V – AOD crash

Yep, 4.4GHz was about it. Perhaps we were a little timid, but raising the voltage beyond 1.465V on a brand-new, pre-release 32-nm processor felt like asking for trouble to us, especially with just air cooling. Then again, the crashes and errors we were seeing came quickly, well before the chip had a chance to heat up beyond the capacity of our cooler. We’d almost surely run up against a frequency limitation in the chip—and that’s unusual. CPU these days tend to be primarily thermally constrained, and you can usually push them a fair ways past their default frequencies before running into stability issues.

Worried that we weren’t reaching our chip’s full potential, we pinged AMD PR on the matter, who pointed us to a section in the reviewer’s guide (a document we shamlessly ignore after extracting any useful info) that suggests 4.5GHz is a reasonable expectation for FX-8150 overclocking with an air cooler. We also discovered, at the same time, that AMD had disabled three of the chip’s four modules during the Guinness World Record run. That’s not something we’d expect, you know, real users who care about performance to do.

The FX-8150’s performance does appear to scale well with clock frequency increases. Had AMD been able to ship a Bulldozer-based part at close to 4.4GHz, the story told on the preceding pages might have had a different feel to it.

Conclusions

You’ve seen the performance results and probably have a sense of where things stand, but we can summarize them with more precision using one of our world-famous price-performance scatter plots. (We’ve omitted a few really expensive CPUs, because they tend to squeeze everything together, making the plot hard to read.)

The more desirable mixes of price and performance should be close to the upper left corner of the plot area, while the least desirable will be toward the bottom right. As you can see, the FX-8150 can’t quite match the overall performance of the Core i5-2500K, and it costs $29 more, as well. That puts the FX-8150 in a difficult position purely from a value standpoint—without even considering that the Core i5-2500K fits into a smaller 95W power envelope. That leaves us pondering what exactly AMD was thinking when it priced the FX-8150 above the 2500K. Perhaps it’s hoping to get some marketing traction from the FX’s inherently bigger numbers—Eight cores! Higher clocks! Larger caches!—but those things don’t count for much when it comes to delivered, real-world performance.

I’m also not entirely thrilled with the way the FX-8150 managed to edge so close to the Core i5-2500K in our overall performance index, with strong performances in certain types of widely parallel tests and some really rather weak showings in more typical desktop applications that use one to four threads. Bulldozer’s performance characteristics could make a fair amount of sense for server-class workloads, but desktop users will probably always have to contend with some applications dominated by a nasty, branchy single main thread. In such cases, the FX chips aren’t horribly weak, but they’re sometimes no faster than a relatively cheap CPU like the Athlon II X3 455.

Faced with such results, AMD likes to talk about how Bulldozer is a “forward-looking” architecture aimed at “tomorrow’s workloads” that will benefit greatly from future compiler and OS optimizations. That is, I am sure, to some degree true. Yet when I hear those words, I can’t help but get flashbacks to the days of the Pentium 4, when Intel said almost the exact same things—right up until it gave up on the architecture and went back the P6. I’m not suggesting AMD will need to make such a massive course correction, but I am not easily sold on CPUs that don’t run today’s existing code well, especially the nicely multithreaded and optimized code in a test suite like ours. The reality on most user desktops is likely to be much harsher.

Speaking of harsh realities, the fact that such a large chip, at 315 mm², can’t manage to keep up with Intel’s much smaller Sandy Bridge silicon is really quite unfortunate. The hope for the Bulldozer architecture’s success in desktop PCs now rests on a series of future possibilities, starting with the maturation of GlobalFoundries’ 32-nm manufacturing process. We know AMD is having trouble shipping enough Llano chips to meet demand, and thanks to our overclocking exploits, we have a sense that FX-series processors may be up against process-related challenges, too. If those get fixed and AMD is able to squeeze several hundred megahertz or more into the same power window, maybe the FX series can improve its value proposition.

Beyond that, we know AMD already has working examples of the Trinity APU, whose “Piledriver” core includes improvements for both instruction throughput and power savings. The plan of record is for Trinity to be on the market before Intel’s Ivy Bridge arrives next spring. An updated server and desktop chip based on Piledriver—a true replacement for Orochi/Zambezi—is slated for some time next year, as well. If AMD can deliver those chips in timely fashion and stick to its projected yearly release cadence, with the mysterious “Steamroller” and “Exacavator” scheduled for 2013 and 2014, perhaps this architecture will progress toward its true potential.

Comments closed
    • Arag0n
    • 8 years ago

    [url<]https://techreport.com/articles.x/15818/1[/url<] I remember a similar bank of results when we first saw the benchmarks from intel i7 new architechture.... however, I'm pretty sure everyone is happy intel is i7's and not core2 anymore... it's going to take time for compilers, software and everyone to take advantage of what bulldozer places over the table, until then, AMD outlook will be pretty bad.

    • drbaltazar
    • 8 years ago

    i be average user dont care too much about fpu number lol .fx 8150 fma4 optimised test show that with prper software this proc crush intel whre people use their computer and i bet you hadnt even optimised threading.so in the end we still have much to do to comprehendto full power of thr fx 8150 and has you have shown fma4 is the main thing that is underestimated and yet show us huge number when compared to 2600.what most will wonder tho is a 2600 using ms compiler instead of intel compiler .since ms has been fma4 compliant since sp1.for me decision is made i ll go with the un understood fx 8150

      • sschaem
      • 8 years ago

      “The two tests above make use of FMA4, and these really aren’t the sort of results we were anticipating. In both cases, the Phenom II X6 110T is faster than the FX-8150 with AVX and FMA4 enabled. Hrmph.”

    • Draphius
    • 8 years ago

    seems like amd really screwed the pooch on this one. i just dont see what market this fits into. i would gladly pay less for a 2500K if im on a budget and if not id definetly go with the 2600K or just wait till the 2700K is out and who knows maybe this will prompt intel to just drop the 2600K price to the same price as the 8150 and kill amds whole market

    • gc9
    • 8 years ago

    [quote<][url=https://techreport.com/articles.x/21813/6<]Memory System Performance[/url<]: You may be surprised to see that the Phenom II X6 1100T achieves higher bandwidth than the FX-8150 at the smaller block sizes, but remember it has six L1 caches where the FX has four. [/quote<] I think you mean L2 caches. The Phenom II X6 has six 512kB L2 caches, and six 64kB L1 data caches. The FX-8150 has four 2MB L2 caches (one per module), and [b<]eight[/b<] 16kB L1 data caches (one per core). What is this memory test doing? I will guess may be it is a block copy. (There are also read-only and write-only tests.) The L1 cache is write-through, so for writing a block smaller than 2MB, or copying a block smaller than 1MB, the FX might be limited by the L2 cache write bandwidth. I see the FX crosses over somewhere near 1MB in your graph. (The Software Optimization guide for Family 15h also notes the FX has worse performance than previous generations in cases involving concurrent streaming stores through the write-combiner, but that may not be the problem here, especially if this is a single threaded test.)

    • ronch
    • 8 years ago

    Here’s my belief on why AMD settled for this architecture, why it’s really an 8-core processor, and why its performance is the way it is.

    Zambezi is truly an 8-core processor. Each integer cluster is independent of the other integer cluster and each minds its own business. It also doesn’t matter much whether they share the fetch and decode stages either or being called a module or whatever. As long as the fetch/decode can feed both integer clusters fast enough, it doesn’t matter if it’s split or unified. What [u<]really matters[/u<] is what happens [u<]inside[/u<] the integer clusters. Each integer cluster only has two ALUs and two AGUs. The K10 had more than that. To make up for that narrow path, AMD, instead of putting in more ALUs and AGUs per cluster which would have required a smarter dispatch unit capable of extracting every bit of parallelism from a thread, decided to ramp up clock speeds instead. This not only allows AMD to settle for dispatch logic that only has to feed two ALUs, but having to feed a total of four ALUs across two integer clusters only requires the decoder to put out 4 RISC ops (ROPs) per cycle. Whether each integer cluster gets 2 ROPs per cycle or whether each integer cluster gets 4 ROPs every other cycle doesn't really matter in the grand scheme of things because the dispatch unit queues the ROPs so it could look for parallellizeable instructions to dispatch later. What AMD lacks in width, it wanted to compensate for with clock speed. But what if AMD decided to put 4 ALUs per integer cluster, thereby more closely matching SB's wider architecture? Would there still be as much debate whether this is a 4-core or an 8-core? The thing is, AMD didn't get that clock speed. And that's why we're here today, crying out loud. As it is now, I believe AMD's insistence that this is an 8-core product is valid. But I also said before that whether people will accept its claim that it's an 8-core processor would hinge solely on its performance. If Zambezi managed to clock at 6+GHz and delivered as much single-thread performance and twice the multi-threaded performance of a 4-core SB, the debate would've died much sooner.

      • sparkman
      • 8 years ago

      If we are going to call BD an octo-core, then we have to admit that while each core may have good integer performance, the same core has only crippled, half-speed floating point performance.

      If we are going to call BD a quad-core, then we only have normal floating point performance, but WOW double integer performance per core!!!!

      Either way it comes down to semantics and only matter for marketing.

      To be most accurate maybe we should call this BD chip a 12-core chip. Eight of the cores being integer-only and four being floating-point-only. But then we have to talk about why the 12-core chip can only run up to 8 threads simultaneously, most of the time.

      In short, microprocessors are complicated. What’s more interesting is how fast it runs your favorite game etc.

        • travbrad
        • 8 years ago

        Don’t give the marketing team any ideas. 🙂

        • ronch
        • 8 years ago

        It is an 8-core processor, but without the clock speed AMD wanted, each core is pretty weak and you’d have to harness each and every one to rival SB. As for the FPU, there are 2 x 128-bit FPUs that you can merge into one 256-bit FPU. You can look at this as though each integer cluster (core) has its own 128-bit FPU, but can use the other core’s FPU if it needs to. AMD just decided to adopt a co-processor layout that decouples the FPUs to the integer clusters. Also, the 128-bit FPU may be ample if this was launched in 2007 (Barcelona had a 128-bit FPU per core), but in light of the competition having 256-bit FPUs, BD’s 256-bit FPU per module looks like each core only gets half of it. Not pretty if both cores are running FPU-intensive code, but I suppose FMAC is what will give BD an edge in supercomputers and enterprise usage.

    • mark625
    • 8 years ago

    After letting these test results stew for a few days, I have come to the conclusion that AMD really screwed up everyone’s expectations by calling Bulldozer an eight-core processor. What AMD has done is to take a slightly different approach to multi-threading than Intel. They put two full integer execution units per core, instead of just duplicating the per-thread registers like Intel does. But the end result is still two active threads per core (nee module), just with a higher level of concurrency.

    If AMD had been telling world+dog for the past four years that Bulldozer would be a quad-core processor with concurrent multi-threading, expectations might have been more in-line with reality. We can see the results right there in the benchmark scores, with the quad-core Zambezi just inching out the six-core Phenom 1100T. Thus each Bulldozer core is about 1.5 x K10.5 core, for a 50%+ improvement over the previous generation.

    Looked at in that light, AMD has done a very good job of improving their core CPU technology, with many additional improvements to come. But when touted as an eight-core monster CPU, the results do not look quite so good. Once again, AMD’s marketing department has been its own worst enemy.

      • NeelyCam
      • 8 years ago

      Yeah; It probably should’ve been called a quad-core.. then AMD could claim a significant IPC improvement..

      But whatever it’s been called or marketed as, the bottom line is that its performance/power and performance/cost (and here I mean cost for AMD – not price for a customer) metrics aren’t very good, and that’s ultimately what really matters.

        • luisnhamue
        • 8 years ago

        I still can’t understand how they come up with a whopping 2billion, 300+mm^2 die are cpu, and still sell it for 200 and some bucks.

        Or intel is selling their CPU at a high price (what I hard believe) or AMD and GF have discovered a way to make CPU really cheap.

          • sweatshopking
          • 8 years ago

          intel charges more for their cpu’s. it’s not news. look at the recent quarterly earnings.

            • travbrad
            • 8 years ago

            I think that’s an extreme oversimplification to just look at earnings. Intel sells a lot more CPUs so even if their margins were the same Intel would be making a lot more money. Intel/AMD don’t exclusively produce CPUs either. For example AMD sells a lot of GPUs, and Intel sells a lot of SSDs.

            Obviously Intel is getting much better margins on a more mature architecture that has half the transistors (using their own fabs) though.

            • sweatshopking
            • 8 years ago

            it’s not just volume, they charge more per cpu. something like 10-15% more than amd per chip.

          • Lazier_Said
          • 8 years ago

          Price isn’t cost. They’re cheap because that’s what they’re worth.

          Read the review, who would pay $300 for one of these duds? Even $200 is a sucker’s buy.

          During the Prescott debacle 6 years ago Intel was deeply undercutting 140 mm2 Manchesters with their 200 mm2 Smithfields. Bad products get discounted.

          • shank15217
          • 8 years ago

          Its not that hard, 16MB of cache, 4 HT links large xbar updated memory controller.. out of 4 HT links only 1 is used, out lf 16MB of cache most desktop loads don’t benefit from such a large number. BD for desktop is trinity and it will be a lot smaller.

      • luisnhamue
      • 8 years ago

      totally right. this is a quad core CPU. But they already messed up the things with customers.

      It’s like, for everybody AMD is putting an 8core CPU against a 4core intel CPU, and even this way they loose.
      For customer it means that they CPU’s just suck. And it also means that they need at least an 12 cores CPU to outperform an intel 4core CPU.

      Bulldozer FAILED to DELIVER EVERYTHING.

        • sweatshopking
        • 8 years ago

        LOSE. THE WORD YOU WANT IT IS LOSE, NOT LOOSE!!!!!!

          • Bensam123
          • 8 years ago

          What was the word on the forums someone brought up… pedantic?

            • yogibbear
            • 8 years ago

            pedagogic.

            • sweatshopking
            • 8 years ago

            erotic

            • yogibbear
            • 8 years ago

            pr0n?

          • NeelyCam
          • 8 years ago

          Where’s the link to that funny cartoon, addressing a bunch more (like it’s/its, your/you’re…)?

            • Meadows
            • 8 years ago

            [url<]http://theoatmeal.com/comics/misspelling[/url<]

            • NeelyCam
            • 8 years ago

            Thx!

      • swaaye
      • 8 years ago

      I haven’t really been paying attention to what AMD mouthpieces are spouting. What I saw in reviews was this newfangled architecture being soundly defeated by quad cores in almost all cases. It even does poorly in cases where its 8 threads should be doing well, so I think it’s undeniably a wreck no matter how you view it.

      • Bensam123
      • 8 years ago

      I agree with this. It’s quite misleading.

      • Lazier_Said
      • 8 years ago

      OK, call it a quad core. Now you have an explanation to enthusiasts asking, “Why so slow?”

      But if it’s just a quad core, then why so hot? Why so big? Why so expensive?

      The bottom line is desktop BD is not a competitive product. It isn’t going to sell to anyone who looks closely enough to ask questions. People who look closely ignore marketing anyway. Marketing’s job is to sell to the ignorant.

      And with nothing else going for the product, [b<]8[/b<]*[b<] cores![/b<] is about the only angle they've got. If you call it a 4 core, where does it stand out from AMD's existing products let alone the real competition? AMD already has a 6 core chip and a flagship with just 4 is a downgrade as far as the Best Buy crowd is concerned.

        • Bensam123
        • 8 years ago

        You know, it’s interesting… I was asking the same questions. Then it came to the point of ‘Why did they even make this chip?’. Assuming the guys at AMD are relatively smart and discarding all the BS like saying it’s a ‘roadster’, what did they actually see in the chip that made them continue to persue it for years? Assuming early in their prototypes they could’ve seen relative performance, they could’ve discarded it at a relatively early stage.

        In this day and age, like you’re saying, it’s hard to pass something off as a PoS in the computer hardware industry unless there is no alternative. The Phenom 2 runs almost just as fast in most tests. They could’ve dumped BD and tried optimizing the Phenom 2 more. So why would they pursue such a bomb of a processor? Why would really smart people want to do this to themselves? Is it just ego? Is it getting caught up in something you’ll thinks really awesome and not having it manifest? It’s hard to believe a company like AMD, which is one of the forerunners to running the world of computers would make such a huge mistake.

        At least part of me was wondering this. Maybe I’m giving them too much credit, but there has to be SOMETHING that makes this chip worth spending billions on actually bringing to a level where people can consume it. Optimizations around an instruction set or an increase in memory speed is hard to believe (that’s something they could’ve added to a Phenom revision). It may have something to do with how it executes threads…

        I don’t know, I think there is part of the picture we aren’t seeing yet, but that could just be really hopeful thinking.

          • sparkman
          • 8 years ago

          Firstly, Bulldozer was super late. Had BD hit the shelves a year ago or earlier, then it might have been impressive. The engineers worked on a promising chip for many years then on a late chip for only a much shorter time.

          Secondly, Bulldozer is designed for the future. It has a more modern architecture with interesting design choices that could be great. Poor execution of details probably resulted in lower-than-planned clock speed, holding back the total potential of the chip. Plus the poor marketing choices already mentioned did not help. Had this first supposedly-octo-core BD chip been marketed as a “fat” super-quad-core, and arrived for Christmas 2010, maybe we would have all been excited.

          But as things stand, BD is only slightly better than a flop. AMD needs to fix several problems if we are ever to see what this radical new modular design is capable of.

            • Bensam123
            • 8 years ago

            You seem to be under the impression that they never tested Bulldozer performance wise till the last year or so and that they couldn’t cancel it. I’m going to assume here, but it’s safe to say that one department in AMD isn’t completely oblivious to the workings of another and even if it was amazing years ago, it wasn’t near close enough to an actual production model. They could’ve still canceled it back then just like they can see now that Phenom 2 performs just as well.

            Don’t mistake what I’m saying as either I’m for or against AMD, I’m merely questioning WHY they moved on with the model when you take everything else into consideration. Even if it had an extra ghz tacked onto the clock speed it would still be slower then Intels chips and faster then AMDs current lineup, but definitely more power hungry. I think in order to understand my point of view you have to completely discard all the ‘common’ answers they spit out concerning this chip… it’s power hungry, late to market, it didn’t hit the target for speed, etc.

            Even now, after the launch, they can STILL cancel the chip. They can still discard it and move on with Phenoms, which appear to be a better chip at this point in time. Heck, they don’t need to completely discard it when canceling it either, they could roll what makes the chip good into the Phenom series, same as Intel did with early P4s before switching to Core2s. So there has to be something that is keeping it going.

            • sparkman
            • 8 years ago

            The answer to your WHY question is that BD is a more advanced design with great potential if AMD can get their act together. I.E. it could be better than Phenom etc. Don’t confuse the crappy performance of this current chip with the performance of the overall long-term architecture that might still turn out to be great in future chips. (Or might not — but the point is we don’t know yet.)

      • sschaem
      • 8 years ago

      The biggest issue is that AMD just cant articulate its design choice and share its vision.
      Zambezi seem to also suffer from a few problems, so we are looking at a stepping stone and not the architecture fully unlocked.
      Bulldozer will NEVER achieve SandyBridge performance and IPC with any workload using 6 or less thread, thats by design.
      What is open is power efficiency, and heavy threading compute.

      Look at llano, they have product future prof product (Flash11 ready to invade the web), yet AMD is doing near nothing to promote it.
      [url<]http://www.nissan-stagejuk3d.com/[/url<] They have a perfect CPU/GPU combo: get a llano laptop today and your web experience will get better overtime. Not the case if you get a SandyBridge based laptop. AMD cant create a sense of excitement for its product even so they have a winner product. I'm taking this stand after I got the opportunity to play with a A8-3500M laptop, and I was amazed at how quiet/barely warm it got running at full load (4core + GPU maxed out) over one hour of intense multiplayer gameplay. Way, way quieter and cooler then a i7-2630qm.

    • stmok
    • 8 years ago

    [u<][b<]A[/b<][/u<] [u<][b<]M[/b<][/u<]arketing [u<][b<]D[/b<][/u<]isaster => Over-hyped by Marketing Dept. This silicon does [b<][u<]not[/u<][/b<] deserve the FX title. Architecture is probably great for specialised workstation/server or supercomputing work, but doesn't belong in the desktop-consumer market at this time. Especially when the last generation product can compete with it in various situations for a lower price! ...I'm having doubts as to whether there's any point in considering the Piledriver-based Trinity APU and Vishera CPU for 2012. (That's when I plan my next major upgrade...Currently on 2005 desktop gear while running a low-end Core i3 Sandy Bridge-based Lenovo ThinkPad as my interim main system.) AMD says they're going to aim for 10% to 15% increase in performance per year, but I don't think that's aggressive enough to start giving Intel a run for their money. (Intel is focusing on making up their deficient IGP side with Ivy Bridge and later on, Haswell in 2013.) Sure, consumers need AMD for competition reasons. But there's really nothing you can do when they're fumbling the ball by letting their Marketing Dept run their game...Over-promise, under-deliver. 🙁 Marketing Dept should be punished with budget cuts and those savings re-directed to Engineering; so they can actually come up with something worthy to talk about. Right now, the FX-series can be summed up in one word: [b<][u<]Disappointment[/u<][/b<]. Screw brand loyalty. If one cannot deliver what I'm looking for in products, service, or solutions; I'm taking my money elsewhere. No company deserves my sympathy if they cannot deliver what I want.

      • MrDigi
      • 8 years ago

      No matter the extra time taken to fix BD it still falls short. Given the power levels it does not leave room for APU graphic cores. I would count on PD making enough difference given the 10-15% performance projection, unless AMD is now under hyping their future. AMD can only continue to drop prices to compete at the expense of making money.

      Intel’s future competition will come from ARM.

    • ub3r
    • 8 years ago

    Hold ‘Shift’ and click ‘Thumbs up’ on my comment and see what happens.. 😀

    • ish718
    • 8 years ago

    Interesting read: [url<]http://www.xbitlabs.com/news/cpu/display/20111013232215_Ex_AMD_Engineer_Explains_Bulldozer_Fiasco.html[/url<] That could explain the lower than expected performance and high amount of transistors... AMD FX 8120 Bulldozer for $220 on newegg BTW...

    • shaq_mobile
    • 8 years ago

    clock for clock, it doesnt do too shabby when its at 4.4ghz. the performance seems to increase significantly. is that just my imagination? if not, why would it seem to gain such a bonus in clock speed vs a similarly overclocked 2600?

    • ronch
    • 8 years ago

    Wow, 500 comments here! Ok, make that 501.

    Bulldozer is undisputably the most anticipated, most hyped up, and most tragic CPU launch in the history of the entire universe.

    Oops, looks like Meadows commented a little sooner than I did. Make that 502 comments.

    • Meadows
    • 8 years ago

    Can we reach 500 comments?

      • Bensam123
      • 8 years ago

      no

        • Bensam123
        • 8 years ago

        yes 500

      • ronch
      • 8 years ago

      I originally expected this topic to reach 700 comments, if not more. We’ll see if I’m right.

        • Anonymous Coward
        • 8 years ago

        We’ll need someone to start trolling.

          • Meadows
          • 8 years ago

          Hold on, I’m calling NeelyCam.

            • NeelyCam
            • 8 years ago

            ……hello…?…..zzzzzz…

            • Palek
            • 8 years ago

            Your powers are needed! Flame on!

            • NeelyCam
            • 8 years ago

            I’m tired, and I have work to do. Isn’t 500 comments enough..?

            • ronch
            • 8 years ago

            We need a thousand more. So, will you help us?

            • yogibbear
            • 8 years ago

            Challenge accepted?

      • ssidbroadcast
      • 8 years ago

      Well I would comment five hundred times…

    • ub3r
    • 8 years ago

    Let down after let down, fans can only take so much.

    • clone
    • 8 years ago

    a long time spent waiting for weaksauce.

    I held off buying a Thuban for this?……why AMD even bothered to hype this product is the mystery.

    not sure if I’m going to go Intel or AMD but I will be spending time comparing prices and perf not on Bulldozer but on Thuban and Intel’s latest.

    wish AMD would produce Thuban using 32nm.

    • MrDigi
    • 8 years ago

    Lower module performance verses 2 full cores was expected, but what is surprising is the single thread performance appears much lower than Phenom II. The Cinebench rendering single thread score, adjusted for clock speed (.98/1.11) * (3.6/4.0) equates to 79.5% the performance for BD.

    The power consumption graph looks much worse for BD than Ph2. Not sure how the task energy values ended up so close. Intel results for 45 to 32 nm was 33% lower task power (875K vs 2600K).

    So is the new architecture or the process the bigger issue to solve?

      • swampfox
      • 8 years ago

      BD has a longer pipeline, so clock speeds need to be increased in order to get equal throughput. It was designed to have higher clocks, so adjusting for the clock speed just takes this intention out of the equation.

        • MrDigi
        • 8 years ago

        So like Netburst, an architecture without the necessary process.

          • Krogoth
          • 8 years ago

          Bulldozer has little in common with Netburst.

          The clockspeed comparison is tenuous at best.

          The current problem with first iteration of Bulldozer is that it is meant to be a server chip.

      • Theolendras
      • 8 years ago

      Maybe some of both, many cues seems to point at the process technology node not very conclusive for now. At least Llano apparently is not having good yields.

      • OneArmedScissor
      • 8 years ago

      I’m going to defy the knee jerk logic and say neither.

      The impact of the “architecture” is not yet a known quantity. That would be the “module” itself. This thing has 16MB of cache in a completely outlandish configuration that was never meant for desktops.

      People are always clamoring for an option of X CPU, plus more cache. Well now you know why nobody does that. If you want to judge the architecture itself in a PC context, you at least need an example designed with PCs first and foremost, like Trinity.

      If you blame the manufacturing process, you’re just buying into AMD’s fud. While they may very well have yield issues, it doesn’t seem to have a power efficiency fault. The idle power did drop quite a bit, and Llano turned out well enough.

      Also keep in mind that veryone’s alleged “efficiency” tests are always completely unrealistic scenarios where the CPU is left running a synthetic benchmark flat out, not performing a legitimate task that a person uses a PC for. Considering idle power dropped, it really should use less in a lot of situations.

      • sschaem
      • 8 years ago

      This is what the architecture disclosure predicted.

      Bulldozer start with 50% less execution unit per thread. At the same clock speed its hard to compensate for this.
      AMD decided to spread the units across more logical core. So you need more threads to utilize whats available.

      For power consumption, a smart review site in germany decided to measure BD vs Phenom at a lower P state.
      Guess what, complete reversal. at 1.4ghz Bulldozer is almost twice as efficient per what as a Phenom in Cinebench.
      I would like to see more confirmation of this, but does indicate BD is a huge step forward.
      We just dont see this when the chip is using ***1.4volts***

      The issue we have is : bulldozer is clock limited because AMD need to juice bulldozer to insane levels on 32nm
      This cause massive power consumption and ceil the clock rate.

      The architecture can clock to 8ghz, the architecture at low voltage is very power efficient
      so, AMd/GF got room to tweak… The key is lower the voltage.

      I still beleive that the cache on BD B2 is busted.

        • khands
        • 8 years ago

        It’s strange, the desktop is like the perfect place for bulldozer to [i<]not be[/i<].

        • MrDigi
        • 8 years ago

        Seems like the module level instruction dispatch could share the integer resources like the FPU resources for single thread, then again you couldn’t count it as being two cores. Intel’s approach of up to 3 integer instructions per clock (single thread) or split for dual threads is a more effective approach. So we will have to see if BD can hold its own in servers. My sense is if that is where it shines they would have kept it for that market, especially to cash in on higher prices, and worked on a specific PC version.

    • chuckula
    • 8 years ago

    I just made a post in the forums but: Has anyone checked on BD’s actual availability online? Aside from Newegg & Tigerdirect that have the 8120 & 6100 I haven’t found a single online outlet that will actually ship you a Bulldozer without a wait. Also, I can’t find the 8150 in stock anywhere.

    I know NCIX is popular here, and neither the US nor Canadian branches have any models of Bulldozer in stock at all.

    I’m not suprised that the 8150 might sell out in the first day, but the very low number of online retailers that actually appear to have gotten *any* Bulldozers is implying that this was *not* a hard launch.

      • smilingcrow
      • 8 years ago

      They do seem to be suffering from processor erector dysfunction so more of a soft launch.

    • ish718
    • 8 years ago

    I guess when AMD named the architecture Bulldozer they were being sarcastic. O_O

    This is supposed to be a new architecture but it sometimes( a lil too often) performs slower than the previous generation architecture, epic fail? -_-

    • bwcbiz
    • 8 years ago

    Score one for all the Intel fans who were predicting disappointment. The overall performance results were only slightly off what I expected, but I expected them to at least hit the 95W TDP point for a 4 module/8 core chip. And the single-threaded performance is just awful. Clock for clock, it doesn’t even beat a Phenom II. Either Intel has patents on all the good stuff, or AMD just doesn’t have the right microcode designers. Or both.

      • OneArmedScissor
      • 8 years ago

      Yeah, woohoo, go Intel fans! Cheers to no faster CPUs for like two years! Score! Epic win!

      • A_Pickle
      • 8 years ago

      [quote=”bwcbiz”<]Either Intel has patents on all the good stuff, or AMD just doesn't have the right microcode designers. Or both.[/quote<] Or Intel has $11 billion more net income than AMD. That might have something to do with it... nawwww.

    • xiaomimm
    • 8 years ago
    • HisDivineOrder
    • 8 years ago

    The first thing that struck me about these CPU’s is how they just seem fabrication-stricken. The next thing that occurred to me is to call them, “like Prescott” or the initial P4 launch, too. Then I asked myself, “Why do I think AMD can beat Intel?” They did it for a period there when Intel stumbled. And that’s when I remembered…

    The only time AMD beats Intel is when Intel miscalculates. It isn’t AMD that beats Intel; it’s Intel that makes a mistake. When AMD stomped on Intel, it was because they spent too long not upgrading the P3. AMD had time to eventually show up with an Athlon that was just better than the Pentium 3 and took so long in making the P4 they tried to release a P3 that too far exceeded manufacturing capability (to reliably hit the clockspeed) that they had to recall it. By the time Intel FINALLY got around to releasing the P4, AMD had become quite good at making Athlons, making only minor changes and supporting the industry-wide progression from SDRAM to DDRAM while Intel tried to lock AMD out through the tri-fecta of super speeds, new architecture/instruction set/chipset, and new RAM technology. Unfortunately, none of that worked out well at all. Much as it does now, the NEED for SPEED led to a tech that wasn’t great IPC and the speed they sought didn’t materialize. Moreover, RDRAM was expensive and the expense did not afford any advantage over DDR-equipped AMD systems. Intel miscalculated, leading to the golden era of AMD CPU’s that has never truly been replicated since. Eventually, Intel switched to DDR and reduced the cost on the chipset, leading to a resurgence toward Intel that has not abated since. AMD can’t beat Intel; only Intel can beat Intel.

    So expecting AMD to out-architect Intel is silly. They’ll never do it. They never have. Intel has to trip and fall on their face, much as they have with Atom. That was another miscalculation on their part. They thought Atom would never amount to anything, so they didn’t bother to engineer more advancements beyond die shrinks, never expecting AMD to drop Bobcat and it become so popular. AMD might do the same with Trinity, but it seems unlikely.

    I’ve been saying for a while now that I felt like Llano and Bulldozer are prototypes for the future, not really meant to be anything but a canary in the coal mine to see if the different ideas to be integrated into Trinity (Fusion and the new module architecture, respectively) have problems and how those problems might be resolved. I also think that the potentially good aspect of decoupling the FP from the dual INT aspects is that they can then more easily migrate the FPU over to the same area/share the same area as the GPU functions they’ll eventually shift to (versus having a GPU grafted into a shared core as they do currently). Bulldozer looks like a work in progress to me. How many modules can you have if you start combining the space in the total CPU that uses FPU with the space that also has the GPU functions? The first step toward that would be to begin disentangling the traditional parts of the CPU from one another that don’t absolutely have to go together to later integrate them with the GPU functions to come…

    That said, it’s not really an excuse for why anyone should buy Bulldozer or even for why anyone would buy an 8150 or 8120. I don’t see any good reason to. The high end of the mainstream (SB vs enthusiast’s SB-E) is just not a great place for AMD and hasn’t been for quite some time. Perhaps they should have focused on their value proposition toward the middle or low end…

    I read an article that clearly lamented the fact that AMD settled only on the 81xx series and didn’t send out samples of the 61xx or 41xx series. I’d like to see those CPU’s in practice against their chief opposition because I have to wonder if those chips won’t fare much better against their opponents. They’ll have much the same clockspeed (and OC’ed be especially capable of it with less cores to worry about going wrong), have the same amount of L3 cache as the 4 module, 8-INT core versions, and they’ll be against CPU’s with less cores than the Bulldozer variants have modules.

    TLDR; this is exactly what I expected and I don’t think anyone should expect Piledriver to put AMD on top. If you want ultra performance, go SB-E or possibly IB. If you want a value in gaming machines, you might wait for FX-41xx (especially 4170) or FX-61xx series benchmarks/pricing to solidify or you can be safe and go something SB. Piledriver probably won’t improve performance higher than SB, to say nothing of IB. Trinity will continue Llano’s argument of, “Good enough CPU, great enough GPU” at a value price but now with tighter integration of CPU and GPU along with higher performance, lower power usage, and more cores in the laptop variants.

      • travbrad
      • 8 years ago

      Great post and I hope people don’t skip it just because of the length. 🙂

      Techspot did some benchmarks of the 2/3 module versions. I’m not sure I trust them completely (never read their site before), but it’s the only thing we have to go on right now. The “6 core” version does decently, but it’s priced too high, and the “4 core” falls flat on it’s face pretty much. The 4170 does okay in gaming (considering the price), but even there it’s struggling to beat similarly priced Phenom IIs.

      Link:
      [url<]http://www.techspot.com/review/452-amd-bulldozer-fx-cpus/page9.html[/url<]

    • anotherengineer
    • 8 years ago

    That’s to bad for everyone.

    I wonder what the load voltage is? I know my 45nm 955BE will run 1.35 to 1.40V which is too much in my opinion and I lowered it to 1.200V

    I wonder if Bulldozer load voltage is also higher than it should/needs to be?

    Here’s hoping that the next revision will bring gains like the phen1 to phen2.

    In theory it looks good. I wonder if its the silicon, (SOI vs HKMG), the architecture itself or other issues that is hindering or capping its potential.

    I find it funny that so many people are booing it, I mean AMD shelled out 5 billion for ATI which it could have put into CPU R&D if they didn’t buy ATI. Then Intel came out with core duo, then PC sales slowed, then the economy tanked in 2008, and PC parts started falling dramatically in price (lower margins) and PC sales slowed more, and Intel which has been raking in about 10 billion a year for the past several years. And people expect AMD to best Intel in overall performance with a fraction of the resources………………..

    I am not in the market for a CPU so it doesn’t matter too much to me, I am still saving for an SSD, which would probably make the most noticable system performance boost over a cpu or gpu.

    • tbone8ty
    • 8 years ago

    [url<]http://www.xtremesystems.org/forums/showthread.php?275873-AMD-FX-quot-Bulldozer-quot-Review-%284%29-!exclusive!-Excuse-for-1-Threaded-Perf.[/url<] interesting...this needs to be tested more, disabling a combination of cluster units and threads in bios and how it performs.... just curious how it performs vs phenom I 9850/9950 vs phenom II how much did Phenom II increase over Phenom I? was it the 10-15% AMD is talking about with Piledriver?

    • Buzzard44
    • 8 years ago

    Awesome review, great work! Really appreciate it.

    One nit – page 4, first paragraph under the second table, you reference the 2120 when you mean the 2320.

    • Aussienerd
    • 8 years ago

    Great review Scott, very balanced. Unlike the comments

    It is good to see AMD now have a processor that is at lease in line with Intel in the mass market.

    I dont know what people expect from company that is only a fraction the size of Intel.

    This at least provides an alternative for people at a compeditive level that is until Ivy Bridge.

    • ssidbroadcast
    • 8 years ago

    I can’t wait for the next podcast. You guys will be talking about this and Steve Jobs. It’ll be a really good episode!!

    • jamsbong
    • 8 years ago

    I’m so glad I’ve upgraded my old Q6600 to 2500K a few months back! After all the rumours and hype, this is as disappointing as Duke Nukem Forever. Both product has given me the same disappointing feel.

    Lets hope ATI can save AMD’s arse again with their new graphic core.

    • LoneWolf15
    • 8 years ago

    For those of you who enjoy a good deal, high-end six-core and four-core Phenom IIs may be an excellent choice as they get pushed out of the channel. If I had an AMD-based system, this is where I’d go.

    I’m as disappointed as everyone else. I didn’t think AMD would beat Intel, but the fact that the Core i3-2100 can compete so well with the top-end Bulldozer in so many situations is highly disappointing.

    Add that to the poor power consumption, and I have trouble seeing why anyone with a Phenom II X4 or X6 would even consider upgrading –and I certainly wouldn’t expect to see anyone with a Nehalem or Sandy-Bridge Intel CPU to be a switcher, either.

    • Bensam123
    • 8 years ago

    Sad… I was hoping AMD had another AMD64 up their sleeve. Guess we’ll have to wait and see. Like Scott said in the article, this is quite reminiscent of the netburst P4s and all they failed to be.

    Nice article though. Consider adding Hard Reset to the gaming benchmarks. I’m pretty sure it has the best visuals on the market right now or it’s in a very close headbutt with Metro.

    • Krogoth
    • 8 years ago

    Call this a long-shot hunch.

    I suspect the very reason why Bulldozer’s FP performance is so lackluster to begin with is because CPU architects were anticipating that GPGPUs are going to pick-up it. FYI, GPGPUs have far superior FP performance then any the current x86 chips on the market. However, their integer performance is very weak.

    AMD’s real mainstream product has always been “Llano” and soon the upcoming “Trinity” chips. Where it is quite possible that the integrated GPU can be used as a GPGPU (FPU) to help address Bulldozer’s lackluster FP performance.

      • OneArmedScissor
      • 8 years ago

      I’ll call that a given.

      However, it’s not an excuse for cranking the core clock speed of what is effectively a low power server chip and pawning it off as the successor to the Athlon 64 FX line for desktops. An overclocked, low power server chip just makes for an incomprehensible hodge podge of a server chip.

      The core clock speed is double that of the already high latency L3 cache. I think this would be a very different story if there were a separate version configured as the PC iteration of Sandy Bridge is.

      This was straight up marketing nonsense, not some sort of too forward thinking design or technical failure. They ought to at least be slapped on the wrist for it.

        • Krogoth
        • 8 years ago

        The current FX line is just another way for AMD to recoup their R&D costs, while bring another product refresh to the existing Phenom line.

        It is more of a less, AMD’s Willamette. A new CPU architecture that was inferior to its predecessor (Pentium III) with most applications at the time. It was only faster in applications that took advantage of its new architecture. 😉

          • OneArmedScissor
          • 8 years ago

          I understand that, but there’s dropping your worn out thing and moving on to make a buck…and then there’s going to the lengths of releasing a new socket type, months in advance, for a coming “high end” CPU, which is likely going to be trounced by their own mid to low end CPU in very short order. It’s a trap!

          The architecture isn’t inferior. The stupid way they set up the desktop chips is. Since a lot of these are going to be the rejected server chips, anyways, they could have disabled some L3 cache, turned its speed up significantly, and then lowered the base clock speed a little.

          But then they wouldn’t be able to show bigger numbers in the specs on Newegg, and thus, AMD chose to go full retard and even branded them FX. Ugh…pant suit.

          • NeelyCam
          • 8 years ago

          [quote<]The current FX line is just another way for AMD to recoup their R&D costs, while bring another product refresh to the existing Phenom line.[/quote<] How do you recoup R&D costs when the chip is [i<]huge[/i<], underperforming, and yields are in the toilet? Cost/performance is how you make profits. They would be better off selling X6 1100Ts, with good yields and the cost of 45nm fabs pretty much amortized right about now

            • Krogoth
            • 8 years ago

            A sale is sale. AMD has to make use of their yields, otherwise they go to waste.

            Intel did the same thing in their worse moments.

            • NeelyCam
            • 8 years ago

            Intel has fabs, so they have to do that.

            AMD does [i<]not[/i<] have fabs. Moreover, they have that sweet deal with GloFo where bad wafers cost nothing.

            • shank15217
            • 8 years ago

            Lol if bad wafers cos nothing then why does it matter how big the die is? AMD is selling all of their FX chips and will be selling all of their Interlagos chips

            • NeelyCam
            • 8 years ago

            ???

            Big die costs more to make than a small die – that’s why it matters. Low yields and bad wafer deals are a separate issue.

            • Anonymous Coward
            • 8 years ago

            I presume they expect to get a lot better margin by selling these things for servers, with desktop sales as a bonus. They certainly didn’t make this chip for the desktop market with servers as a bonus.

      • sschaem
      • 8 years ago

      Cinebench and povray seem to show decent results.

      The issue might be the writes… take a closer look at L1, L2, L3 and memory writes on their own.

      23BG for L1 cache writes is abysmal, not even half what a core2 , phenom gets. No CPu in the past 5 years get L1 and L2 cache performance that low.

      For a shocker.. 11GB L3 write bandwidth! 11GB. That nearly half what sandy bridge get out of ddr3.
      AMD L3 cache is TWICE SLOWER then dual channel ddr3-1333 memory. How come no review made that observation?!
      Even more mind boggling, you can write data to main memory on an i7-2600k as fast as you can write it to the L2 cache on bulldozer. ~20GB

      This is borderline insane. To me this scream BUG. AMD must have done some hack to make sure cache coherency wasn’t broken and the chip invalidate all cache levels propagating the data all the way to the dead slow L3 cache.
      I dont see how else to explain this ?

      How can running write operation at half speed be good for any architecture?

      I mean getting 10GB out of DDR3-1600 is not normal. no for a desktop chip, specially not for a server chip. If AMD designed bulldozer purposefully with a 10GB memory write bus, everyone at AMD should be fired 🙂

      To recap:

      SandyBridge i7-2600k memory writes bandwidth to DDR3-1333: 19Gigabytes/s
      Bulldozer fx-8150 memory writes to bandwidth to DDR3-1600: 10Gigabytes/s

      20% faster ram, half the bandwidth… this is not a design choice. We are seeing the result of a hack to work around a cache bug.

      This need to be investigated…

      BTW, read performance is decent. So when you measure bandwidth via copy the half speed writes are less obvious. that 50% loss become ‘only’ a 25% loss.

        • just brew it!
        • 8 years ago

        Something is borked with those AIDA64 numbers you keep coming back to. They are inconsistent with the graph on the “Memory subsystem performance” page of this review. Until you can explain that discrepancy, it is not unreasonable to assume that there’s been some sort of “measurement error”.

        It seems to me that if cache performance was truly as bad as your numbers imply, performance of this chip wouldn’t just be mediocre; it would be abysmally bad. Something doesn’t add up.

          • sschaem
          • 8 years ago

          This review did not benchmark read and write in isolation. The issue is with writes.
          And thats when the B2 stepping cache coherency bug manifest itself and AMD had implement this slow work around.

          And you dont think the numbers from the Fx-8150 are absolutely abysmal?
          [url<]https://techreport.com/articles.x/21813/15[/url<] The FX is 30 to 40% slower compared to a i7-2600k for task its supposed to excel at. B2 stepping seem to have decent read performance, the issue is with writes. Check ALL review that include read/write and cache/memory test. ***And to date AMD as not spoken about AIDA reporting false result for its FX chips.***

    • Welch
    • 8 years ago

    Why do I get the feeling this is like Nvidia and its release of Fermi… More Power consumption, less or almost on par only in some games (applications)… Hmmmm.

    I’d bet that this architecture will become pretty damn good in time. Seems like they took way to long getting it out and couldn’t ask anyone to wait any longer. They will probably release their updates to it in about a year or so and it will make sense then. /sigh, disappointing 😐

    • blaydes99
    • 8 years ago

    Scott (Wasson), could TR sort the game benchmarks that have a maximum + minimum framerate according not just to max frames, but then by min frames? It gives a strange result when you have 5 cpus all at 65 fps, but then the one on top has a minimum of 21 and the one below it has 30 fps.

    Ideally, it would actually be the minimum fps that decides who’s on top of each category in which the max fps matches among CPUs. Make sense to you?

      • Damage
      • 8 years ago

      Nah, FPS minimums are a nice sentiment, but a poor measure of performance. Here’s a better way:

      [url<]https://techreport.com/articles.x/21516[/url<] Also, the averages are higher precision beneath the covers, so we're sorting by 89.1, 89.3, 89.6, etc. Adding a secondary sort for minimum values won't change the final sort order as a result.

    • NeelyCam
    • 8 years ago

    -deleted-

    • flip-mode
    • 8 years ago

    333

    YEAY BABY! Fourth time! I’m the king of the world!

      • dpaus
      • 8 years ago

      flip-mode is not on-point; flip-mode is point-less…

        • flip-mode
        • 8 years ago

        Being 333 doesn’t require any other point – it IS the point.

    • bcronce
    • 8 years ago

    This is how I look at it. I really like what AMD stands for. Lots of opensource support and I loved their Keynote about their vision for the future.

    Personally, I think they have the talent, but some manager botched the entire thing. Even if the idea is good, the implementation seems to be horrible. Twice the transistor count of Intel chips, while getting schooled isn’t “competitive”.

    My hope is that the design actually works and they can fix the “problems”, whatever they are. I really want AMD to do well.

    For now, I will have to go with Intel because that’s the only logical choice, unless Piledriver really turns things around. They have until the release of Ivy Bridge to convince me, before I am forced to upgrade.

    I’ll probably be sticking with their GPUs, unless nVidia makes huge leaps, which I don’t see it happening.

    • BlackStar
    • 8 years ago

    Is it true that the Win7 scheduler misbehaves on Bulldozer, reducing performance? If so, is it possible to run a few tests on Win8 for comparison?

      • NeelyCam
      • 8 years ago

      Yes, and Tom’s Hardware had some:

      [url<]http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-23.html[/url<]

      • khands
      • 8 years ago

      I know AMD has mentioned that (anandtech.com mentions it in their review), the difference seemed to average 8-12%, significant, but not quite enough.

      Oh, and that was from AMD’s own slides on an (obviously) dev build of Windows 8, Anandtech hasn’t tested that yet themselves

    • geekl33tgamer
    • 8 years ago

    AMD CPU’s are slower than their Phenom II X4/6 CPU’s clock for clock – Check
    AMD CPU’s are still slower than Intel’s clock for clock – Check
    AMD GPU’s are still slower than Nvidia’s clock for clock – Check
    AMD 7xx/8xx/9xx Chipsets still perform slower than Intel’s – Check

    Last time I checked, this is not what progress looked like.

      • geekl33tgamer
      • 8 years ago

      Must have picked up the AMD fanboy’s with this one. Last time I scanned over the benchmarks, all of the above is correct. Take the rose tinted glasses off sometimes who ever keeps down voting.

      I also expect this to get downvoted into oblivion too.

        • NeelyCam
        • 8 years ago

        You forgot a few.

        AMD IGPs are still faster than NVidia’s.
        AMD IGPs are still faster than Intel’s.
        AMD netbook chips are still faster than Intel’s.
        AMD chipsets are the only ones that support USB3.0.
        AMD’s chipset market share is growing, while NVidia’s is shrinking.

        Meadows: could you please take care of the grammar part?

          • khands
          • 8 years ago

          They’re the only ones that support USB 3 [i<]natively[/i<], though I would gather it's more because of Thunderbolt than anything else. I would argue with rcs2k4 about the GPUs clock for clock as well, on GPUs clock speed is not nearly as important as it is on CPUs. And AMD GPUs are still more power efficient and generally priced better than their Nvidia counterparts (not that there aren't reasons still to go with Nvidia, just that the GPU race isn't nearly as cut and dry as the CPU race).

          • dpaus
          • 8 years ago

          Neely, are you feeling OK? Maybe you should sit down…

            • NeelyCam
            • 8 years ago

            Actually, I have this weird sense of impending doom.. just like when I first saw that Steve Jobs was dead.

            I have a feeling this will affect the prices of my future Intel CPUs somehow… I just can’t put my finger on it.

    • RtFusion
    • 8 years ago

    Obligatory:

    [url<]http://www.youtube.com/watch?v=SArxcnpXStE[/url<] *sorry if this was posted already here.

    • luisnhamue
    • 8 years ago

    Yes. this is a joke. I prepared a great dinner, just to get delighted while reading what turned out to be a piece of crap, that somehow is giving me stomachache.

    First pages, it sounded strange, but still interesting, but I soon realized that this is not a real 8 core CPU, they did a trick that resembles GPU architectures, and wow, its just compatible with some OS’s that simple don’t even exist

    And then, why so many transistors in a Xtra Large die area, I say more than doubling the competition, and High TDP… all that with 0 benefits over the previous architecture.

    I don’t know whether they are getting mad, but there must be something wrong.

    People are saying that this is a Fermi, on d CPU side, but NO. Fermi at least was fast even though it was power hungry, hot, inefficient…it was FAST enough to beat the competition, and AMD just failed to put a bullet on Core i5 2500K.

    Game Over for AMD. I repeat, they’re joking with people now.

    • Meadows
    • 8 years ago

    [quote<]However, we've tried 1866MHz memory on the older CPU cores in Llano, and our Stream results topped out at around 15MB/s.[/quote<] Have you tried upgrading to an 80386 DX to lift that limitation?

      • ShadowTiger
      • 8 years ago

      Also, isn’t 15 MB/s a typo? The chart goes past 15GB/s and the lowest thing on the chart is 3.6 GB/s for the atom…

      I think it would be ridiculous to have memory going slower than my HDD in this day and age.

        • derFunkenstein
        • 8 years ago

        That was his point.

      • Damage
      • 8 years ago

      Nice, smartass. Typo fixed.

        • Meadows
        • 8 years ago

        Thanks. 😉

          • phez
          • 8 years ago

          i lol’d, seriously

    • chrone
    • 8 years ago

    wake us up when 2013-2014 is here. oh wait, hope amd is still exist. 😀

    • OneArmedScissor
    • 8 years ago

    On the bright side, we don’t have to wait years for new relevant things from AMD anymore. They should have something new showing up on a 3-4 month cycle. We have this in October, likely the new Bobcat in January, then Trinity, Vishera…and so on. At least it’s not so boring as one hit or miss thing from Intel each year, and Intel might even pick up their pace with more variations of their own.

    • faramir
    • 8 years ago

    Wow, it is interesting to reminisce in how incredibly full of fail Pentium 4 (represented in this test by Pentium EE 840) was … which makes Bulldozer AMD’s Pentium 4. There’s no node shrink in sight to save this Phenom-all-over-again 🙁

      • WillBach
      • 8 years ago

      AMD has demoed Trinity running with updated Bulldozer cores (Piledriver) at 28nm…

        • chuckula
        • 8 years ago

        No it hasn’t since Trinity is a 32 nm part. AMD has demoed 28nm pure-GPU parts that will be used in laptops though. Edit: I didn’t mean to imply that AMD hasn’t demoed Trinity.. it has… just that Trinity is definitely a 32nm part.

          • faramir
          • 8 years ago

          I’m afraid chuckula is correct on this one. Maybe this iteration of Bulldozer is just plagued with bugs that are going to get ironed out when AMD familiarizes itself with the concept of multiple threads running on a single “module” but until they do …

          I’m sure they can cut down the power consumption once Global Foundries figures out how to make 32 nm chips properly. When this is done they should be able to achieve higher frequency, giving them advantage over Phenom II (afterall Bulldozer was built for higher frequencies and it is painfully obvious that the foundry didn’t quite deliver what the engineers have expected) and perhaps even reaching parity in more than just a handful of tests with Sandy Bridge just before Ivy Bridge is out.

          At that point AMD is likely to revert to its second fiddle status, until they figure out how to make their own “Core2” – low power and high IPC chip that doesn’t just throw more transistors, more cores, more volts and more watts at the problem of beating the other team. They are using SOI, they should have some process node advantage (especially regarding pwoer consumption) over Intel, right ?

    • kamikaziechameleon
    • 8 years ago

    Well so much for getting one of these. The performance gains over something like a 1090T(my current processor) are so slight its pointless, not to mention how poorly it fairs price/performance wise against intel. This has pretty much insured I only invest in a new GPU for BF3. How can they price it in that range and expect anything good??? Are they trying to burn early adopters???

    I expect a major price cut in the next month. Seeing how the new architecture fairs vs the x6 of last gen I’m inclined to expect a 190 or 180 dollar FX-8150 by the end of next month.

    • dpaus
    • 8 years ago

    Ouch… AMD, I feel for ya!

    [EternalOptimist] The ‘Piledriver’ version is supposed to bring 10-15% performance improvement, no? That – and a $46 price reduction – would make an 8150 a lot more competitive[/EternalOptimist]

      • ermo
      • 8 years ago

      Here’s hoping!

      • sschaem
      • 8 years ago

      If the new sheduler does give ~5% speedup (seem to be 2 to 10%), and AMD lower the voltage on this 32nm process I would pay $195

      My issue is still how can they have botched the DDR3 memory controller.. over 2 time slower then intel using 20% faster ram?
      How can that be ever explained… (beside the L1 cache being 4x slower)

        • derFunkenstein
        • 8 years ago

        Part of that latency, if I understand it correctly, is because it has to pass through a 3rd layer of cache. More L2 and no L3 would have made for better access times.

          • sschaem
          • 8 years ago

          I’m not talking about random access, but streaming.

          Check AIDA64 memory write performance. How can AMD only get 10GB out of DDR3-1600 and Intel get 19.4GB with DDR3-1333

          Amazingly the i7-2600k DDR3-1333 memory performance is almost as fast at the FX-8150 L3 cache!!!!!! How is this even possible?

          It seem AMD invested 4 years of R&D in the wrong places. the cache and memory controller seem to be the worse part of the design.

            • just brew it!
            • 8 years ago

            [quote<] How can AMD only get 10GB out of DDR3-1600 and Intel get 19.4GB with DDR3-1333[/quote<] Is that an apples-to-apples comparison? (I.e. was the Intel test done on a dual-channel memory configuration?)

            • sschaem
            • 8 years ago

            Have you seen any review yet that show the FX-8150 getting more then 10GB write to its DDR3-1600 memory?
            And have you seen any review showing the i7-2600K getting anything less then 19GB write using DDR3-1333 ?

            So for in AIDA64 the difference is clear. AIDA was updated in June and in August with optimization for Bulldozer.


            Memory Tests

            Memory bandwidth benchmarks (Memory Read, Memory Write, Memory Copy) measure the maximum achiveable memory data transfer bandwidth. The code behind these benchmark methods are written in Assembly and they are extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x86, MMX, 3DNow!, SSE, SSE2 or SSE4.1 instruction set extension.

            The Memory Latency benchmark measures the typical delay when the CPU reads data from system memory. Memory latency time means the penalty measured from the issuing of the read command until the data arrives to the integer registers of the CPU.

            • just brew it!
            • 8 years ago

            OK, got it. For some reason I thought the 2600K was capable of triple channel. Never mind…

          • bcronce
          • 8 years ago

          My i7-920 has 8MB of L3 cache. I’m not sure how this is much different than AMD’s current cache.

            • khands
            • 8 years ago

            It’s not so much the size of the cache as it is the speed.

        • just brew it!
        • 8 years ago

        [quote<]My issue is still how can they have botched the DDR3 memory controller.. over 2 time slower then intel using 20% faster ram?[/quote<] Umm... what?

      • derFunkenstein
      • 8 years ago

      Incremental improvements like that are only useful if Intel stands still. Not likely.

        • OneArmedScissor
        • 8 years ago

        Ivy Bridge and Sandy Bridge E are just about the equivalent of standing still in the desktop space. That’s not what either of those things are really for, of course, but it seems to be all anyone here cares about.

          • paulWTAMU
          • 8 years ago

          Because that’s what I care about? I don’t do servers, etc. I use desktop PCs.

            • OneArmedScissor
            • 8 years ago

            Case in point. How about laptops, far and away the majority of PCs? As I was saying, this site is ever so slightly biased towards one thing.

    • Rza79
    • 8 years ago

    Bulldozer was originally scheduled for a 2009 launch. Well, performance wise, it most certainly feels like a 2009 CPU. But of course in 2009, on a 45nm process, it wouldn’t even perform this ‘good’. What makes matters worse is that Intel could launch an ‘i7 3000K’ in a 125W power envelope any time and just completely destroy AMD. That is, if they want to destroy AMD, which they don’t want.

      • sschaem
      • 8 years ago

      Are you referring to Sandy Bridge -E ? The 130w version start at $560

      LGA2011 got my interested… but not ready to upgrade yet.

        • Rza79
        • 8 years ago

        No a SB @ 3.8Ghz with a turbo of 4.2Ghz.

      • OneArmedScissor
      • 8 years ago

      They’re already “destroying” AMD in desktops. What will a higher power Sandy Bridge change there? It’s still Sandy Bridge, still stuck at the same turbo speeds. Been there, done that.

      For far and away the vast majority of people, that just means more cores to power gate, and more cache those cores don’t need – arguably some of the very same issues Bulldozer is facing here.

      People need to ask for logical improvements that actually make your computer work better, instead of just waste. The high end desktop market may be small, but that line of thinking just further marginalizes it.

        • NeelyCam
        • 8 years ago

        [quote<]What will a higher power Sandy Bridge change there? It's still Sandy Bridge, still stuck at the same turbo speeds.[/quote<] More power could enable higher turbo for a longer period of time. That said, I agree - there is no point in Intel releasing a 125W enthusiast desktop CPU right now.

    • tfp
    • 8 years ago

    Maybe someone can explain this.

    Bulldozer can only feed one core at a time per module with it’s current decoders. Wouldn’t this cause pipeline latency or a bubbles switching back and forth every clock? Does this still happen when only one of the cores is active and the other doesn’t have a thread to process?

    There are so many things about Bulldozer that is just off I’m not even sure it will make a good server chip. The scaling from 1 to 8 threads is better than 1 to 8 for intel with HT but the cores are so under performing it just doesn’t catch up in most cases. I guess we’ll see with server benches but even on integer benchmarks it is not impressive.

    I hope they improve in a number of areas on the next chip.

      • Damage
      • 8 years ago

      As I understand it, the decoder can dispatch 4 ops/clk in interleaved fashion to one of the two cores. Those ops then go into the scheduler queues to be scheduled for execution, so generally they’re not acted upon immediately, anyhow. AMD said it expected this sharing arrangement to work well, because decoding shouldn’t be particularly latency-sensitive. With only two ALU/AGU pairs per integer pipe, a four-instruction decoder hitting on every other clock should probably get the job done, rate-wise. I doubt BD’s issues are as simple as a missed estimation on this front.

      Sorry these uarch issues weren’t covered in more detail in the text. Too much info to process and convey, too little time.

        • sschaem
        • 8 years ago

        Did AMD disclose anything to you guys ? (mainly about the cache/memory operation at half speed)

        My impression is that the chip also had to be overvolted… can you try to undervolt your ‘sample 8150 ?

        The hope is that at some point AMD will be able to get to use much lower voltage and bring the power consumption during load, much, much lower.

        Also check other review using a Radeon 6950 vs nVidia GPU. The delta in games seem much lower.

        x264 AVX seem to give a 10% boost to both platform… pass1 seem to be limited to AMD horrid cache/DDR3 performance.

        Thanks for hammering on the chip for us.

          • OneArmedScissor
          • 8 years ago

          They’re going to “overvolt” it no matter what. See: upcoming FX-8170, and every single flagship Phenom ever made.

          While the new model at 125w TDP often ends up bumping the previous down to 95w, nobody ever cares. People are happy to buy 130w TDP Intel CPUs, too. AMD probably feels like it’s ok, so long as they avoid the old 140w tier, which seemed to get them some bad press.

            • bcronce
            • 8 years ago

            Intel’s Ivy Bridge 6core is suppose to be about 60 watts. 1/2 the power is a big difference, especially with servers. Or even look at corps with 1000 computers. Shave off an average of 20 watts would save both on power and cooling.

            • OneArmedScissor
            • 8 years ago

            That’s quite possibly [i<]over[/i<] a year away. Just the standard laptop/desktop Ivy Bridge is still about 6 months out. You're comparing apples to oranges, anyways. AMD has also had very low TDP, bazillion core server CPUs for years. The TDP levels aren't changing for the PC iteration of Ivy Bridge, aside from quad-cores being offered at 35w. It's all wait and see. AMD has several different versions of Bulldozer coming next year, not even just a sole "fixed" one like this.

          • Damage
          • 8 years ago

          AMD never said anything about the caches running at “half speed,” and I’ve not heard that elsewhere, either. The caches didn’t measure out particularly fast for us, but we’d expect relatively lax timings on a design intended to run at high frequencies. I think the whole chip is running slower than expected, not that any one subsystem is.

            • sschaem
            • 8 years ago

            The latency numbers seem within acceptable range, but the throughput is questionable:

            2600K vs 8150 cache write performance:

            L3 : 24.7BG vs 12.5GB – 2x slower
            L2: 37.6GB vs 20.1GB – 1.9x slower
            L1: 60.9GB vs 23.4GB – 2.6x slower

            MEM: 19.3GB (ddr3-1333) vs 10.2GB (ddr3-1600) – 1.9x slower (while using 20% faster ram)

            Getting 10GB out of DDR3-1600 while Intel get 24.3GB out of DDR3-1333 seem like something is broken.

            The L1 cache performance being almost 4 time slower is also of concern. I’m not sure how to explain both ?

            Edit: numbers are now both for stock clock.

            • chuckula
            • 8 years ago

            I hate to downthumb the master, but the L3 cache does run at a slower speed similar to how the L3 in Nehalem was on a different (slower) clock plane than the CPU cores. It’s one reason why Bulldozer with 2 Billion transistors has a (relatively speaking) less-than-expected increase in power consumption compared to a Sandy Bridge where the L3 ring bus runs at full speed with the CPU cores. I’m not exactly sure what the clocks of the L3 cache in BD are though (more info needed).

            • Damage
            • 8 years ago

            Ah, I assumed we were talking about L1/L2 caches, since the article was quite clear the L3/NB are on a separate clock domain, at 2.2GHz in desktop parts. AMD never mentioned anything about a “half speed’ arrangement for the L1/L2. Also, L3 cache contributes to latency but is otherwise barely even measured in the Sandra test we showed, which skips from 4MB to 16MB block sizes.

            • chuckula
            • 8 years ago

            I 100% agree for L1 and L2 caches. Those are considered part of each module and are clocked 1:1 with the cores.

            • sschaem
            • 8 years ago

            All numbers L1, L2, L3, mem writes seem very low… is it just me that notice this ? I use AIDA as reference.

            Unless AIDA is busted, all site report the i5-2500K to have 100+ GB in L1 cache copy performance, ~60GB write
            But the FX-8150 show < 50GB copy and < 23GB write.

            Its like L1 cache write are automatically propagated to the L2 cache and are atomic (stall execution) ?!?!

            Reads are amazingly fast… so its it seem to be the writes that are having a problem ?
            Makes sense as read dont cause synchronization problems. (note: AIDA only use 1 thread of its test)

            Maybe a bug forced AMD to propagate all writes to all cache levels ? I mean 23GB to the L1 cache, thats half phenom performance.

            Again, is this totally expected or am I wrong to raise a big red flag ?!

        • ermo
        • 8 years ago

        Are you likely to write a follow up to this article as e.g. the scheduler issues and various packages are optimized for BD?

        Wouldn’t it make sense to discuss the underlying microarchitectural details that necessitate these optimizations in such an article when more becomes known about Zambesi?

          • Damage
          • 8 years ago

          I think we need more code before anything other than a very short article becomes possible about Bulldozer-optimized applications. Also, remember, the AIDA64 tests in this review are purportedly well optimized and use XOP and FMA4. I didn’t want to make too much of it, but those results look pretty dire. I wouldn’t expect miracles from optimized code, given that.

          Oh, and I actually wrote about the Bulldozer architecture last year:

          [url<]https://techreport.com/articles.x/19514[/url<] And then you have Mr. Kanter's article, linked in the review, which provides even more detail. There may be some opportunity to look at things like branch prediction accuracy and misprediction penalties (crucial with a longer pipeline), comparative IPC, and the like, now that we have a chip in hand. I'm also excited to try out my new game benchmarking methods on the latest CPU. So... we'll see what's next.

            • sschaem
            • 8 years ago

            More benchmark idea ?

            – Use a CUDA card and run their memory benchmarks. This will isolate the PCIe implementation

            – OS/Application boot time would be of interest

            – Go the other way, dont do an overclocked test but an undervolting test.

            – Check Battlefield3 on windows7 vs windows8, Radeon vs geforce, using i7-2600k/i5-2500k

            • ermo
            • 8 years ago

            Yes, you wrote about it last year. And yes, Mr. Kanter wrote about it last year. And yes, I’ve read both more than once. My question was not meant as a dig at this article and your previous coverage. My apologies if it could be read like that.

            However, a few details were left out of both those reviews, partly because no realization of the architecture was out and about and ostensibly also for competitive reasons (as in, these details were not disclosed to the public).

            Those are the details that I’m inquiring about, along with what happens on the optimization front and how software vendors feel about optimizing for the minor player and the politics involved in this.

            • Damage
            • 8 years ago

            AMD still hasn’t disclosed a number of things, including basic info like the pipeline depth. They’ve been pretty guarded leading up to the product introduction. Still, what other info were you wanting to know?

            Seems to me like the inclusion of FMA3 in Piledriver resolves a lot of the optimization questions, too. People will use FMA3, and that will be that.

            • OneArmedScissor
            • 8 years ago

            I’m curious about the cache clock speeds. With 16MB in a 4+ GHz CPU, they surely didn’t go the Sandy Bridge route of using the same transistors for the L3 and syncing the clocks, but it appears that even the L2 is something else.

            I ask because without someone saying, you don’t know if that varies from one SKU to another, as was the case with all the Core iXs so far. You need a chart for those stupid things.

            • Damage
            • 8 years ago

            AMD told me the NB/L3 is 2.2GHz in desktop parts and 2.0/2.2GHz in the Opterons. Don’t have specifics on the Opteron models, but the desktop bit seems straightfoward.

            • OneArmedScissor
            • 8 years ago

            Thank you. I have not seen anyone else mention this.

            • Waco
            • 8 years ago

            So cranking the NB clock increases the L3 cache speed?

            Seems like an easy way for a lot more performance assuming it scales well…

            EDIT: Looks like that is indeed the key – [url<]http://www.rage3d.com/reviews/cpu/amd_fx_8150/index.php?p=9[/url<] Look at the minimum framerates jump with the "OC2" score. That's with the northbridge clock cranked up a bit. Damn.

            • OneArmedScissor
            • 8 years ago

            This is what I was getting at. Somebody really needs to test this more extensively. If you take the core clock down a notch and the voltage drop that goes with it, the northbridge should be able to go significantly faster without driving up power use. 2.2 GHz is just ridiculously slow. It’s being compared to the SB i7 that runs it up to about 3.7 GHz. Darn AMD to heck for this numbers game flop for the sake of marketing.

            • Waco
            • 8 years ago

            Well my 8120 came in a few minutes ago. I’ll probably have preliminary benches tonight of the CPU stock versus the stock core speed with the highest NB OC I can push it to. Hopefully I’ll remember to post them up here, if not, they’ll be on OCC.

            • khands
            • 8 years ago

            Please do, I’m very interested in your results.

            • Waco
            • 8 years ago

            I wish I could post results right now – but I’ve been fighting with my motherboard for the past 4 hours trying to get it to boot anything but the stock settings. I’ve tried various BIOS versions and not one of them will boot with the CPU settings set to anything except auto.

            It’s an ASRock 890FX Deluxe5 if anyone cares. BIOS 1.60. Boots and runs just fine at stock. Shows error code 46 when booting with the stock CPU settings set manually (boot code 46 is in the middle of RAM initialization).

            I’ll post here if I find a solution and get some benches run…

            • khands
            • 8 years ago

            Well that’s frustrating, hopefully they fix that soon.

            • Waco
            • 8 years ago

            The chip didn’t work properly in my fiancee’s 990FXA-UD3 either (BIOS is the newest version too). It wouldn’t boot with memory speeds set at 1866 or 1600 (she has DDR3-2133…so it’s not the RAM). At 1333 with the timings set VERY conservatively it wasn’t stable AT ALL. BSODs all over the place just trying to boot Windows.

            I was planning on having fun overclocking this chip but these crap BIOS updates are taking all the fun out of it.

            • khands
            • 8 years ago

            Makes me wonder if it’s the particular chip you got if it’s failing like that on both of them or if both those boards compatible BIOS updates simply suck for Bulldozer.

            • Waco
            • 8 years ago

            I’ve seen mention of other UD3 boards not playing nice as well…so I’m hoping it’s just a BIOS bug. If not, Newegg is always great about returns/replacements. 🙂

            • kamikaziechameleon
            • 8 years ago

            WOW!!! interesting. It would be really funny if OCers unlocked these processors potential and not AMD’s engineers, lol.

            • sschaem
            • 8 years ago

            I think FMA4 will be mainly used in their OpenCL compiler, and possibly Radeon driver.

        • tfp
        • 8 years ago

        Thanks for the response.

        So there are 4 decoders feeding the 2 ALUs, and the 2 other Agen pipes with in a core as well as the FPU that includes the 2 MMX pipes (for more integer calc within the FPU if needed) and the 2 FMACs per clock cycle. From the little I understand this is all pipe lined with items moving 1 space per clock down the pipe within each “core” include the FPU it seems like there would be gaps in the pipe until the pipe stalls. If the pipeline stalls I would expect all of the other items in the pipe would keep moving forward and file in the “holes”? With out a clear definition of the pipeline and how it works it’s hard to say if they have stages built in for this alternating decode as well to mask the latency just like intel with the P4 back in the day. Just asking because I think it’s interesting.

        But I agree I don’t think there is just one specific item slowing down bulldozer (for example memory bandwidth seems off as well) I just hadn’t seen this info so I wanted to ask. It just seems like the front end is weak for all of the resources behind if you have to feed the Int core + FPU each clock. I know they had to pick and choose what to put in because of die space/time and really might not be an issue at all.

          • Damage
          • 8 years ago

          I think the scheduling queue does the job of decoupling the decoders from the execution units. With enough entries in the scheduler and a sufficient aggregate rate of ops coming in from the decoders, the latency from having to wait every other cycle for delivery of ops should be effectively masked. When the scheduler wishes to send ops to the execution units, there generally should be plenty of ’em waiting.

          Bulldozer has queues between other parts of the pipeline, like between prediction and fetch, to facilitate sharing, mask latency, and make logic suited for high-frequency operation workable.

          Now, if the decode rate is too low to supply the hardware behind it, then yes, you might have issues. But I doubt that’s a major issue. It is something we could measure with the right thread affinitization. 🙂

            • tfp
            • 8 years ago

            Well if you can measure and publish the results I will defiantly read it. 🙂

            Thanks for the discussion.

      • WillBach
      • 8 years ago

      Damage did a fairly good job of explaining it but one thing he left out is that some modern CPUs cache the decoded results (I’ll dig around for references if I have time later today). When such a CPU is in a tight loop or executing the same function repeatedly, it will need to decode most instructions less than once for each time that instruction is executed.

        • srg86
        • 8 years ago

        Indeed, especially in the case of Sandy Bridge, if the code you’re looping through fits inside this cache, the decoders are turned off completely and it runs purely from that cache. Thus saving power.

    • ClickClick5
    • 8 years ago

    Just finished reading word-for-word…

    Bulldozer II perhaps?

    • NeelyCam
    • 8 years ago

    I know AMD copies a lot of things from Intel, but did they [i<]really[/i<] have to copy Netburst...?

      • geekl33tgamer
      • 8 years ago

      Looks like it, including hyperthreading too! The FX-8150 masquerading as an 8-core CPU’s fooling no one…

        • yogibbear
        • 8 years ago

        Intel may have been the first to implement HT in an x86 cpu, that doesn’t mean they invented it.

      • NeelyCam
      • 8 years ago

      The funny thing is, when I posted pretty much the same comment a month ago, my comment rating was around -10.

        • Bensam123
        • 8 years ago

        You didn’t have test results a month ago. Difference between being uninformed fud and a joke based on information that backs it up.

          • Lazier_Said
          • 8 years ago

          Not like it was any kind of secret that BD had a much longer pipeline and 2 BD cores share a single FP unit. We knew IPC would not be good.

          It was no secret that BD was very late due to manufacturing problems. Which suggested target clocks were being missed. Therefore the product would have to be overvolted to the edge to make up for it. We knew it was going to be hot.

          Most pointedly of all, it was no secret that AMD leaked exactly zero benchmarks because the results were and are embarassing. We knew it was going to be slow.

          The only thing left for benchmarks to confirm was how hot and how slow – which it is and they did.

          The only uninformed fud is coming from AMD true believers who refused to read the writing on the wall in front of them.

            • Bensam123
            • 8 years ago

            Those are some pretty spotty assumptions. Intel is having manufacturing issues as well, does that mean their next chip is going to perform horribly?

            It’s a false premise.

            That leads it back to being FUD.

            • NeelyCam
            • 8 years ago

            [quote<]Intel is having manufacturing issues as well[/quote<] You missed the Haswell demo on IDF, huh..?

          • NeelyCam
          • 8 years ago

          [quote<]...uninformed fud...[/quote<] How many times does it have to be correct for you to consider re-classifying it from "uniformed fud" to "informed analysis"..? Although, I'm still working on the joke part... I would get killed on an open-mic night; I just need more practice.

            • Bensam123
            • 8 years ago

            There weren’t facts available when you made the comment you referenced, that would make it uninformed. It would be FUD because it served no real purpose at the time besides functioning as a fanboi stomping on the other side.

            Tons of people make completely random comments based around speculation and then say ‘I told you so’ afterwards IF they’re correct, even though at the time they had relatively nothing to back up their claim. Just the same as some people thought Bulldozer was the second coming of Jesus before benchmarks hit the street.

            Unless you had some sort of super secret inside source that whispered to you the lottery numbers from the future, it would be uninformed fud at the time it was originally written.

      • tejas84
      • 8 years ago

      lol! BD is really a terrible CPU. Sandy/Ivy Bridge Xeons will eat this alive. Faildozer more like.

        • swaaye
        • 8 years ago

        Blewdozer. 😉

          • NeelyCam
          • 8 years ago

          Dozing Bull?

            • Bensam123
            • 8 years ago

            Dozing Bull, Bridein Sandy.

          • phez
          • 8 years ago

          blowdozer

        • Krogoth
        • 8 years ago

        >mfw calling a server chip “terrible” when it outside of its intended element.

        That’s just like saying Atom and SB are terrible because they can’t effectively handle vitualization and datacenter workloads that are commonplace in the server arena.

        The era of one CPU design to rule them all is over. We are moving into the era of specialized CPU architectures.

      • NeelyCam
      • 8 years ago

      Test failure.

        • NeelyCam
        • 8 years ago

        (-1), and the fun is over. *sad*

        How about a new game? Let’s see how long we can keep the score at 0..?

          • Meadows
          • 8 years ago

          You can’t possibly think I’ll let you get away with that.

            • NeelyCam
            • 8 years ago

            I was hoping you wouldn’t see it for an hour or two..

            EDIT: I guess I should’ve known you were here, as all my posts started getting voted down: I have no excuse for my poor timing.

            BTW, you missed a couple on the second page. Yes – there is a second page.

            • NeelyCam
            • 8 years ago

            Oh that’s right – I remember now that you got an AM3+ board, and were hoping to drop in a BD to replace your POS Phenom. How’s that working out for ya?

            • NeelyCam
            • 8 years ago

            I guess the -1 means you’re a bit pissed off. I can’t blame ya. I would, too, if I were you, but then again I wouldn’t go out and buy an AM3+ board before seeing BD benchies.

            • Bensam123
            • 8 years ago

            They whisper to you too, huh?

            • Meadows
            • 8 years ago

            I didn’t want an FX CPU before seeing benchmarks. Learn to read.

            I also didn’t give you a minus until you already had -2 (so yes, now I did give one). Some people have day jobs, not everyone lives with their parents and watches comments like hawks.

      • Krogoth
      • 8 years ago

      What a gross oversimplification.

      Bulldozer has about as much in common with Netburst as Sandy Bridge has in common with the original P6 architecture.

        • NeelyCam
        • 8 years ago

        Let’s see… (supposedly) high clocks, horrendously high power consumption, slower than what came before it…

        Yep. Netburst.

          • Krogoth
          • 8 years ago

          [Facepalm]

          Bulldozer isn’t clock that much higher then the current competition. Did you quickly forget that SBs are also clocked well over 3.0Ghz? The power consumption of Bulldozer is comparable to Bloomfield and Gulftown chips when loaded. The only thing that made Sandy Bridge stand out was its incredible power efficiency, not its performance.

          If anything, Bulldozer was meant to go out against Bloomfield. It came too late to the party.

          Bulldozer at best is a mediocre desktop CPU and a potentially potent server CPU.

          It is not even in the same league as Netburst.

            • Lazier_Said
            • 8 years ago

            “Bulldozer isn’t clock that much higher then the current competition. ”

            Because manufacturing and thermal problems prevent it from being clocked any higher.

            The same manufacturing and thermal problems that have it a year late as an 125W part just to compete, barely, with Intel’s 65 watt chips from almost a year ago.

            • A_Pickle
            • 8 years ago

            [quote=”Krogoth”<]It is not even in the same league as Netburst.[/quote<] Yep... Netburst wasn't even a "potent server CPU."

          • Anonymous Coward
          • 8 years ago

          This has a lot more to do with Phenom #1 than with anything Intel ever made. I’m not speaking about architecture of course, I’m speaking about at-launch performance. It’s over-promised, late, and not especially competitive but we’ve been there before. They’ll improve it. If it was a [i<]terrible[/i<] design it wouldn't have even been this competitive so soon!

            • NeelyCam
            • 8 years ago

            I guess “competitive” means different things in different circles……

            My bad for thinking that 3rd party benchmarks have value. Apologies.

            • Anonymous Coward
            • 8 years ago

            I used the word “competitive” and so did you, but otherwise there doesn’t seem to be a lot in your post that had to do with mine.

            • smilingcrow
            • 8 years ago

            The competition has a platform that is being replaced in roughly 6 months time that is 50+ to 100+ percent more power efficient so how can BD be even faintly classed as competitive? BD’s nearest rival timescale wise is Ivy Bridge and that’s gonna be even more of a humiliation.
            If BD was a car it would have to compete in the 6L class versus SB being in the 4L class. I think if AMD had slapped a 170W TDP on BD to boost performance the fanboys would say it was a success.

            [url<]http://www.pcper.com/reviews/Processors/AMD-FX-Processor-Review-Can-Bulldozer-Unearth-AMD-Victory/Power-Consumption-and-P[/url<]

            • Anonymous Coward
            • 8 years ago

            There is a difference between “competitive”, which I [b<]did not[/b<] use to describe BD, and "this competitive", which is what I said. As in, if the design of BD was utter trash, given AMD's usual launch problems and the general swirling rumors of problems with their 32nm production, I do not think that it would be [i<]as competitive as it is[/i<].

            • kamikaziechameleon
            • 8 years ago

            Its bad that they get worse every gen with launches it seems. Sure this is not as bad as the phenom launch but its still horrible.

            • Anonymous Coward
            • 8 years ago

            The new 32nm 4-module product more or less matches the old mature 45nm 6-core product, you can’t say the same about the 4-core 32nm K10 vs the 4-core 45nm K10. I think “horrible” is too strong a word. Give AMD time.

            • smilingcrow
            • 8 years ago

            You must have had incredibly low expectations of BD and AMD to use the phrase ‘it wouldn’t even have been this competitive so soon!’

            So soon, really! AMD give the impression that they sync their desktop PC launches with Halley’s comet.

            And ‘this competitive’, really! Intel’s 9 month old platform is offering 50 to 100+ percent more power efficiency. I can’t see why the use of the word competitive even comes to mind at all when a perfectly good word such as uncompetitive fits the bill perfectly.

            • Anonymous Coward
            • 8 years ago

            You’re being dense.

      • ronch
      • 8 years ago

      5 years from now I wouldn’t be surprised if AMD realizes that BD was a big miscalculation on their part and go back and revive the K7/K8/K10 architecture just the way Intel realized Netbust was a mistake and went back to the P6.

        • Anonymous Coward
        • 8 years ago

        We’ll see. I note that neither Atom nor Bobcat is an especially short pipe design despite their tiny power targets, IBM has been screwing around at high clockspeeds for both Power6 and Power7, and even the latest SPARC T4 chip from Sun clocks pretty high (3.0ghz, OoOE with 64 threads). The Sun chip especially is big on work-per-watt yet they dramatically boosted the clock speed from the previous version. It may be that high clockspeeds are not quite done yet.

    • indeego
    • 8 years ago

    And for the 6th year I pass on AMD. They had a nice run in the 00’s. Shame.

    • NeelyCam
    • 8 years ago

    Those Newegg coupon codes for AMD 45nm chips expire today… I guess they were meant to clear the stock for Bulldozer…

    Now, I think today is the [b<]Bargain Day[/b<]; those discounts will disappear tomorrow for good since the old CPUs are still competitive. PII X4 965BE for $115, X6 1100T for $175... Get your Phenom now!! Don't wait! This offer won't be repeated!!

    • smilingcrow
    • 8 years ago

    VR-Zone measured the power consumption when over-clocked which is painfully high. Haven’t seen other reviews giving such data so hard to know how indicative it is. Idle and Load (Prime95) wattages:

    i7-2600k (4.7GHz, 1.47V) – 155W / 286W
    FX-8150 (4.7GHz, 1.47V) – 182W / 440W

    Viewed as the absolute difference between the two platforms: idle / load
    27W / 154W

    Viewed as the relative difference between the two platforms as in AMD consumes x% more than Intel: idle / load
    17.4% / 53.8%

    Maybe the figures are exaggerated because Prime95 is stressing all 8 integer cores!

      • yogibbear
      • 8 years ago

      They don’t perform the same clock for clock either.

    • OneArmedScissor
    • 8 years ago

    Bulldozer looks to be an even better candidate for the Athlon II treatment than Phenom II. Lose the L3 and get back to us with Trinity, please.

    Considering that Bulldozer could act as a quad-core CPU with 2MB of 16 way L2 cache per core, very much like the i7 Sandy Bridge’s L3 “slices,” I have to wonder if the gob of L3 on top of that really does [i<]anything[/i<] but add latency in a PC context. And now I see this line in the Anandtech article, "However AMD indicated that the L3 cache was only really useful in server workloads." *double face palm*

      • khands
      • 8 years ago

      Yeah, they should’ve worked on better L1/L2 latency and gotten rid of L3 altogether.

        • bcronce
        • 8 years ago

        Lack of L3 would be horrible. It would increase inter-module latency and increase required bandwidth as each module would have to snoop the L2 cache of all the other modules for updates.

        L3 cache is extremely important for multicore CPUs, at least with any form of cache-coherency in play.

        When we finally remove cache-coherency, both Intel and AMD have stated this, then we may not need it or at least reduce L3’s usefulness.

          • OneArmedScissor
          • 8 years ago

          The trouble is that we don’t need a squillion cores for PCs. Trinity will only have 4…or 2, as some people prefer to look at it.

            • bcronce
            • 8 years ago

            The other problem is there is an inherent limit on how fast you can calculate a single thread of instructions. You can double the transistor count of a core and make it 20% faster, or you can double the transistor count, keep the same cores, and almost double the core count.

            We’re almost at the point where we can’t milk much more single threaded performance be throwing transistors at it. We can shrink the transistors and make the CPU more power efficient, but we can’t make them much *faster*

            If you want faster, you almost have to add more cores.

            • OneArmedScissor
            • 8 years ago

            [quote<]We're almost at the point where we can't milk much more single threaded performance be throwing transistors at it.[/quote<] I'd argue we're actually past that point. Adding more transistors is making them slower, which is what I was getting at. All the recent CPUs traded being "faster" in some way for more of something else that most people really don't get anything out of on a PC. It's not that we could really push clock speeds any further, as turbo boost and power gating take care of that, but that there's all this extra whiz bang stuff piled on for the sake of further expanding multi-threading. Yes, you need ring busses, many large cache levels, hyper threading, and unconventional core configurations to make that work, but you don't need it to do the work that the vast majority of us are doing, and it's increasing memory latency and wasting power at all times. Much as some people here are asking why they didn't just shrink the Phenom II, I'm still wondering why Intel didn't just shrink the Core 2, which was tiny, and just turn it into a system on a chip. And so they are, as it turns out, for the 22nm version of Atom, but it's silly to have to wait around for these things while more silliness and wastefulness is thrown at us in the mean time.

            • NeelyCam
            • 8 years ago

            I finally qualify as “people”! Happy day!!!

            • OneArmedScissor
            • 8 years ago

            Oh, I’m sure there are several people in your head.

            • NeelyCam
            • 8 years ago

            That’s a +1 – flawless delivery.

    • StuG
    • 8 years ago

    *Sigh* Was really planning on upgrading my Phenom II 940 @ 3.8Ghz to one of these. I’m not even sure it would be worth it though. Still kinda contemplating getting one…if someone wants to persuade me otherwise.

      • NeelyCam
      • 8 years ago

      It’s not worth the extra money and work/time. Keep your dollars and save for Trinity.

      • Arclight
      • 8 years ago

      If your upgrading from a 940 you will need a CPU, mobo and RAM, so why not buy a 2500K and Z68 mobo and some cheap 1333 Mhz 1,5V 8 Gb RAM? It’s available now, no waiting or doubts involved.

      • Kaleid
      • 8 years ago

      Keep your CPU

      • A_Pickle
      • 8 years ago

      Well, I’m underwhelmed like the rest of the world, but then… here I was hoping for another Athlon 64-like upset. To say that that [i<]can't[/i<] happen these days is probably disingenuous and would portray me as having a low expectation of human capability, but... I certainly shouldn't have [i<]expected[/i<] that. Intel is competing strongly now, they aren't opening a giant, competitive vulnerability open for their opponents to exploit (like performance-per-watt). AMD siezed the performance-per-watt market when they realized what Netburst's undoing was... what's the weak spot of Intel's current CPU lineup? I'd argue that they don't really have one (they certainly don't have a weak spot quite as significant as Prescott's performance-per-watt was, either), other than graphics. That said, Bulldozer doesn't look like a bad chip. On the whole, I'd argue it's an improvement over the Phenom II. A little rough at the edges, yes, but so was the original Phenom -- whose design was still the basis of the Phenom II that we all know and love. It seems to perform better overall (especially with that AES block), despite consuming a miniscule extra bit of power. I like it, but I don't like it enough to invest in this initial generation. I was hoping for Bulldozer to be epic, but I had a pretty damn good feeling that it wasn't going to be the Athlon 64 all over again. I'd like to see some architectural improvements and, perhaps, a die shrink before I jump in. Those AES extensions are a carrot dangling on a string, though, lemme tell ya. I was hoping to upgrade to Bulldozer if it was good enough. I think the FX-8150 puts on a respectable showing! It's competitive with Phenom II at worst, and just behind Sandy Bridge at best. That isn't bad at all!

    • tbone8ty
    • 8 years ago

    $279 @ the egg

    [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16819103960&Tpk=fx%208150[/url<] hahaha

      • entropy13
      • 8 years ago

      FX-8150 at $280 is 355% ([b<]$0.78[/b<] per percentage point) Core i7 2600K at $317 is 425% ([b<]$0.74[/b<] per percentage point) FAIL

        • travbrad
        • 8 years ago

        If you place a lot of value in synthetic tests and niche applications like 3d modeling, then yes it can come close. For gaming or regular desktop apps that the majority of people use, not so much.

        You are also ignoring the $180-220 i5-2500K, which trounces the FX8150 in gaming, and even in the highly-threaded applications it’s about even. So the 2500K is cheaper, has better performance, overclocks better, and it uses HALF the power while doing it: [url<]http://i.imgur.com/k6rhe.png[/url<] I know which one I'd rather buy. 😉

          • entropy13
          • 8 years ago

          If you have just scrolled down a bit more (especially if you’re viewing comments style is based on the timestamp of the initial posts) you would have seen my post wherein the 2500K was not ignored.

          • A_Pickle
          • 8 years ago

          I think Intel engages in dickish business practices. I’d like a more power efficient processor, but Intel needs competition, so I’ll gladly buy AMD. Probably not Bulldozer, but whatever ends up being the Phenom II to the original Phenom with respect to Bulldozer… I’ll probably upgrade to.

            • smilingcrow
            • 8 years ago

            AMD fanboys simultaneously live in the past and future:

            Ghost of launches past. (use sulky teenager voice). I don’t support Intel because they didn’t RSVOP me about my eighteenth birthday party.
            Translation: they done very bad things to me in the past and I will never forgive them because they now date the hottest chick from school and have a degree from Harvard.

            Ghost of launches future. (use battered housewife voice). He sweet talks me and has a nice high revving V8 in the garage and I just luvs him; V8 is better than 4 cylinder junk right!
            Translation: I know he’s been bad to me in the past and has been promising to change but I keep on believing even though I have the emotional bruises from all the previous let downs. Am I stupid?

      • NeelyCam
      • 8 years ago

      And they sold all five they had in stock!

    • MadManOriginal
    • 8 years ago

    Typo? Page 17, first paragraph:

    [quote<]...finding applications or benchmarks that make use of AES isn't easy.[/quote<] I think you mean AVX in that sentence.

      • UberGerbil
      • 8 years ago

      Yeah, I flagged that earlier but it got lost in all the rending of garments and gnashing of teeth,

    • yogibbear
    • 8 years ago

    So a Bulldozer is big and slow. Who woulda thunk?

    [url<]http://www.youtube.com/watch?v=Q1gCqhNF4Ss[/url<]

    • entropy13
    • 8 years ago

    Using TR’s scatter plot…

    FX-8150 at $245 is 355% ($0.69 per percentage point)
    FX-8120 at $205 is 330% ($0.62 per percentage point)

    Core i7 2600K at $317 is 425% ($0.74 per percentage point)
    Core i5 2500K at $216 is 360% ($0.60 per percentage point)

    Looking at them in terms of price per performance, they’re not that bad.

    However, the 8150 can by no means match the 2600K In pure overall performance (425% v. 355% = 70% difference), and is actually behind by 5% to the 2500K while being a bit more expensive ($0.69 v. $0.6).

    The 8120 is even worse, just add $11 and you get 30% more performance (at $0.37 per percentage point) with the 2500K.

    And there is still the power consumption to talk about. Core i7 2600K and i5 2500K both idles at 64W. Peak power consumption is 144W and 132W respectively. The FX-8150 has an idle power consumption at 76W and peaks at 209W. There is also a “task energy” graph for them; 8.5W and 9.9W respectively for the two Intel CPUs while it’s 14.4W for the FX-8150. Comparing the 2600K with the 8150, $317:$245 means you save $72, but you end up using more power (12W more at idle, 65W more peak, 5.9W more task energy).

    If the 8150 and 8120 is priced at $200 and $165 respectively, then it would be $0.56 per percentage point for the 8150 and $0.50 per percentage point for the 8120. This would offset, at least, the raw performance advantages of the 2600K and 2500K respectively.

    Take note that those percentages are as “percentage points.” Meaning, “behind by 5%” doesn’t mean that it (8150) is 95% the performance of the 2500K. 355/360 = 98.61%

    • revcrisis
    • 8 years ago

    Disappointing. For the consumer, competition is always good. If Bulldozer was a resounding success and actually outperformed Intel’s current offering, Intel would start seriously thinking about price slashes, sandy bridge-e, and ivy bridge. We might see Ivy come out sooner, we might see cheaper 25/600k’s, and we might see more competitive SB-E parts to fill the gap. Instead, Intel will look at the results, laugh, and continue on with their existing plan/roadmap. There is simply no need for them to adjust their strategy as a simple 2600k price cut and the release of a 2700k will crush the FX. We’ve seen the back and forth between Nvidia and ATI/AMD result in lower prices and higher performance for graphic cards. I think this applies here as well. I really wish AMD had stepped it up this round, better performance for both teams is a win-win for us.

      • mboza
      • 8 years ago

      Intel still need to convince us to buy new PCs to replace the current P4s and Conroes. If Bulldozer was faster, AMD would probably charge more (with a 2600K price level part) rather than force Intel into a price war.

    • TurtlePerson2
    • 8 years ago

    I can assure you that we’re not going back to the Netburst days. As a current computer architecture student, the only time that I hear about Netburst is to hear about the terrible mistakes that Intel made. Meanwhile P6 is presented as a good modern architecture.

    • ermo
    • 8 years ago

    Now imagine being an AMD motherboard partner having invested in releasing AM3+ boards with UEFI back in July. Of course, the chipsets are more or less the same as the past generation, which cuts down on R&D, but still:

    How many of those do you think you’ll get to sell at a nice fat margin now?

    Ouch.

      • FuturePastNow
      • 8 years ago

      It gets better: people who bought those AM3+ motherboards back in July, along with a “placeholder” Phenom X3 or X4, are BETTER OFF keeping the cheapo processor they have, especially if they’re comfortable doing a little overclocking.

        • NeelyCam
        • 8 years ago

        As I said earlier, I think AMD was being very dishonest suggesting people should by AM3+ boards in advance, for “smooth transition”. And I feel bad for people who fell for it.

          • kamikaziechameleon
          • 8 years ago

          I held out some hope but this is a horrible launch. My 1090T is better than what I’d get for the dollar upgrading. Maybe when they discount the 8150 to 180 or 160 dollars it will be worth something.

          • swaaye
          • 8 years ago

          There was no chance of them saying anything but positive comments regarding their upcoming CPU.

            • NeelyCam
            • 8 years ago

            This is true.

            But the way I see it, they were cashing in on the trust of their customers… short term profits, sure, but trust is difficult to regain. There might be some long-term AMD fans out there who bought AM3+ mobos in anticipation, and seeing these results they feel screwed, and switch to Intel vowing never to touch AMD again.

            This is how these things go. Some people will never again buy IBM hard drives, go to best buy, sign up with AOL…

    • wierdo
    • 8 years ago

    Hmm, that’s generally disappointing.

    The only product that interests me personally is the FX-4100 model, four cores at 3.6ghz for $115 doesn’t sound like a bad deal for me personally.

    I don’t need more than 4 cores, 2 is good, 4 is luxury for my needs, the per core performance is my concern at the moment, it needs to be acceptable, I hope the numbers wont be too different from those in the article with only 2 modules instead of 4 (I’m hoping the extra 2 modules are usually idle or lightly utilized).

    I just need to see what the competing models from AMD and Intel have to offer compared to it if I’m in the market for an upgrade this coming year.

    Previously I was looking into the Athlon II X3 line, maybe the 455 for about $80. We’ll see if the new FX-4100 will be worth the extra premium – not a fan of the Phenom II line, Athlon II looked more attractive price/performance wise for me personally.

      • FuturePastNow
      • 8 years ago

      The $150 Phenom II X6 1055T outperforms the $165 FX-6100 in almost every way. That doesn’t bode well for the FX-4100.

      You’d be better off spending that $115 on an i3-2100.

        • yogibbear
        • 8 years ago

        My HTPC looks like it outperforms this new Bulldozer thing.

          • NeelyCam
          • 8 years ago

          My passively cooled HTPC outperforms this thing.

      • FuturePastNow
      • 8 years ago

      Here, Techspot reviewed the FX-4170 alongside the other processors: [url<]http://www.techspot.com/review/452-amd-bulldozer-fx-cpus/[/url<] All I can say about the 2 module/4 core Bulldozer is that at least its power consumption isn't awful.

      • NeelyCam
      • 8 years ago

      [quote<]The only product that interests me personally is the FX-4100 model, four cores at 3.6ghz for $115 doesn't sound like a bad deal for me personally. [/quote<] Seeing how FX-8150 is doing, I would rather go with this $115 quad-core: [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16819103727[/url<] And that's a [i<]real[/i<] quad-core.

        • wierdo
        • 8 years ago

        I didn’t catch the $15 promo code, not a bad idea, thanks.

    • flip-mode
    • 8 years ago

    The longer the results of the review sink in the more disappointing this situation gets.

    • Cranx
    • 8 years ago

    I would be interested to see a benchmark based on compile time of a complex c application. Also, how long does it take to start an android emulator.

    • juampa_valve_rde
    • 8 years ago

    amd, hint! phenom III, 32nm, 8 core, 8 mb l3, powergated, with avx, new divider (the one from llano), turbo core working, and then BAMM at 4 ghz it would be competitive with sandy bridge, with a smaller die than bulldozer and higher performance.

    Oh, and the task of mix and match to get it could be done by the guys that made the bobcat! 😉

    • derFunkenstein
    • 8 years ago

    Wow. I mean, wow. I had a feeling it wasn’t going to live up to what people were publicly hoping for, but to be beaten soundly in several tests by the Phenom II X6 is pretty bad.

    It’s no wonder AMD said they were going to discontinue 45nm Phenom IIs so soon – why would you buy these if you could still buy a Deneb or Thuban and get similar performance? I <3 my $90 unlocked quad. 😀

      • sweatshopking
      • 8 years ago

      what die is smaller? does it make sense to discontinue the phenom ii if it’s as fast/faster AND a smaller die?

      edit nm at 32 nm, bulldozer is likely to be smaller

        • derFunkenstein
        • 8 years ago

        Die sizes are in the article, dude. And BD is smaller than Thuban. :p

          • sweatshopking
          • 8 years ago

          yeah. noticed that

            • OneArmedScissor
            • 8 years ago

            Die size isn’t an issue for “high end” desktop CPUs, anyways. In both Intel and AMD’s case, they’re not high volume parts, but left over scraps from server manufacturing.

            Even if people here would like to believe otherwise, they already have the high volume desktop replacement, which is Llano. Trinity will follow shortly.

            With Opteron sales dwindling and Bulldozer being socket compatible, it was time to let the entire Phenom II line go before it even showed up.

          • bcronce
          • 8 years ago

          I would hope the 32nm BD is smaller than the 45nm Thuban. 32nm transistors are about 1/2 the size of 45nm(32^2/45^2)

    • FuturePastNow
    • 8 years ago

    I’m guessing JF_AMD is going to make himself scarce on forums for a while.

      • sweatshopking
      • 8 years ago

      I hope not. I like him.

    • maroon1
    • 8 years ago

    Bulldozer should be called Pentium 4 X8

    • flip-mode
    • 8 years ago

    Seems to me that AMD is perhaps intentionally abandoning the high performance desktop space. These chips might be great for servers and AMD’s lower performance chips are fine for typical desktop and notebook applications but as for high performance desktop applications AMD just throws a server chip at them and says “good luck”.

    There’s nothing in the design of this chip that shows any intention of targeting the high performance desktop space. This chip is huge and expensive and highly multi-threaded – all features that seem to aim for the server space bull’s eye. AMD’s strategy for this chip was to get back into the server space, I think. Perhaps with later chips they will try to redress the desktop space.

    But I wonder how much this processor design can really be tuned, or would AMD really need a design based on a whole different philosophy to make a desktop comeback.

    And none of that even touches on the fact that GloFo can barely manufacture anything these days!

      • shank15217
      • 8 years ago

      Lets not be so hard on glofo, this chip is a 2B transistor monster on a new process with a new micro architecture.

      • yogibbear
      • 8 years ago

      How does the fact that it took MORE energy for the FX to do the cinebench render than the 1100T demonstrate that this is a good chip for the server space? Surely servers are where you start to care about power and heat even more?

        • khands
        • 8 years ago

        Servers do more than scene rendering, which this chip obviously fails at.

        • Krogoth
        • 8 years ago

        Cinebench = Workstation application, not server.

        Wait until real-world server tasks get thrown at Bulldozer.

        • flip-mode
        • 8 years ago

        Honestly I don’t know how good of a server chip this is. All I know is that it seems to be a pretty unsatisfactory desktop chip.

        Perhaps it will be a crummy server chip too, or perhaps it will shine. If it has some kind of surprisingly good Virtual Machine capabilities then it might sell by the truckload.

        • just brew it!
        • 8 years ago

        Cinebench is probably very floating-point intensive, and floating point is Bulldozer’s architectural weak spot. Server workloads tend to be more integer oriented.

        • OneArmedScissor
        • 8 years ago

        Think of the desktop version as being pre-overclocked. It appears to require 1.3v to hit those ridiculous speeds.

        The server configurations take the opposite approach.

        The 8 core version won’t push the clock speed through the roof. Instead of maximum voltage, they’ll run closer to minimum, while still retaining the large memory bandwidth advantage. The power use goes way, way down from what you see here.

        There’s also a dual-die version with 16 cores and four memory channels, which will have an even more extreme emphasis on power efficiency.

        This is exactly how things have been with the Phenom IIs and their Opteron counterparts for some time. For example, Phenom IIs run 1.4v, hence the clock speeds being even higher than 32nm Sandy Bridge CPUs, but most Magny-Cours are around 1.0v, and still considerably more powerful overall.

        They could push Bulldozer down very, very low. Some of the later laptop iterations of the Athlon II managed 2 GHz at 0.9v.

          • yogibbear
          • 8 years ago

          Thankyou. This fully explained why I was struggling to see how they would apply to servers. And now I can see that there is some hope for BD.

    • Chrispy_
    • 8 years ago

    Tomorrow’s architecture, today!

    “Tomorrow”, this chip will be obsolete. AMD needed today’s architecture, “yesterday” 🙁

    • luisnhamue
    • 8 years ago

    I didnt even read it yet. Let me just prepare my popcorn and beer, to digest all this.

    Wow…I’m so excited! I just hope this story end up well

      • bcronce
      • 8 years ago

      I feel sorry for this guy.

    • ModernPrimitive
    • 8 years ago

    Well AMD, you sure soured my coffee this morning…..

    • Peldor
    • 8 years ago

    Does anybody else feel like AMD is getting an undeserved free pass on the “8-core” label? This is a quad-core chip with double integer units. It’s no real surprise this thing isn’t beating a quad-core with hyperthreading (Sandy Bridge) or even generally a 6-core (Thuban).

      • axeman
      • 8 years ago

      Yeah, that’s a bummer. Apparently halving the number of FP units does not come for free. >:(

    • chuckula
    • 8 years ago

    I feel bad for the engineers at AMD. There are some truly talented people there who don’t deserve what’s been done to them starting with that insider-trading criminal Hector Ruiz. I don’t feel bad for AMD’s marketing or upper management, many of whom are already out of a job due to this mess.

    I’m still baffled as to how a 2 *billion* transistor chip can do so poorly against chips that are less than half its size even in benchmarks that are fully multi-threaded and even with clockspeed advantages.

    I’ve said it before and I’ll say it again: There would be nothing wrong if AMD had come out with a chip that had Bulldozer’s performance characteristics but was of a comparable size and transistor count to Sandy Bridge (or even Gulftown), that was easy to produce and that had good power consumption characteristics. Instead we get this monster that is likely going to be bigger than all of the Sandy Bridge-E chips except for the 8-core server parts.

    If you want to make a 2 billion transistor chip with a large power envelope and super-high clock speeds, you are playing with fire if you can’t get it right. It’s *not* a miracle that Bulldozer can keep up with a 2600K in some benchmarks… it’s a disaster that Bulldozer doesn’t *annihilate* the 2600K in practically *every* benchmark.

      • NeelyCam
      • 8 years ago

      My thoughts exactly.

      • Arclight
      • 8 years ago

      [quote<]I feel bad for the engineers at AMD. There are some truly talented people there who don't deserve what's been done to them starting with that insider-trading criminal Hector Ruiz. I don't feel bad for AMD's marketing or upper management, many of whom are already out of a job due to this mess.[/quote<] Don't feel bad for them, they are to blame!

      • sschaem
      • 8 years ago

      Some possibilities.

      – Idle power match Intel
      – load power is caused by the higher voltage AMD had to use with the current state of GF 32nm process

      If AMD can drop the voltage, its possible future revision will show a dramaticaly lower load power consumption.

      TR could have fun trying to undervolt the process to prove this ?

      The other issue that seem to plague the design… the cache & memory controller.
      At this time it look like AMD is having problems clocking them correctly. having a L1 cache, L2, L3 cache 2 to 3 time slower then a i5-2500k is like putting the parking brake on.

      And WHAT?! 10GB of bandwidth for stream write on DDR3 1600, thats half what Intel get on DDR2 1333.
      AMD seem to have borked the memory controler… again.

      You can see that with benchmark that are not very bandwdith hungry, like x264. Pass2 is a match with the 2600K.

      I have a feeling that if AMD can lower their voltage, make the cache work at expected speed, and tweak the ddr3 controler, update the scheduler in windows, update software to use FMA or tweak for the architecture, the FX-8150 will be worth $250.
      Right now its $195 processor.

      From what I gathered it does seem like the FX perform much worse (relatively to the 2600k) using an nvidia graphic card.
      F1 2011 on a 6950 for example run faster on an Fx-8150 then on i7-2600k, TR show a dramatic reverse situation on a 460 GTX.
      BattleField3 beta also show to run better on the FX processor.

      AMD will most likely survive despite its marketing…. But AMD is doing itself allot of bad by having this main message for BD.

      “Perform mega-tasking and get pure core performance with new “Bulldozer” architecture.”

      • phez
      • 8 years ago

      Don’t blame the people who engineered the chip .. wha?

        • chuckula
        • 8 years ago

        Individual engineers do not throw the entire chip together. They each get stuck with very small pieces of a much larger puzzle. I’m sure in isolation the engineers at AMD did a great job, but the overall vision (coming from upper management) forced the engineering of the chip in directions that were not good for the overall finished product. The exact same thing could be said for the P4 and other chips that have flubbed. There’s brilliant engineering, but that can’t fix larger issues.

          • bcronce
          • 8 years ago

          This ^^

    • ronch
    • 8 years ago

    I sincerely hope AMD doesn’t let Mike Butler go. I mean, even if BD isn’t running as we expected it to, Mike is still a kick-ass engineer. Microprocessors are as complex as hell and there are only a small bunch of these great engineers alive. AMD should keep the talent they have, but also look for other great engineers to augment their existing ones.

    Go AMD! The forums may be filled with cries of disappointment today, but you did what no one else could! Only real men have fabs, and you recently lost your fabs, but you’re the Last Man Standing! The future of x86 doesn’t belong to Intel — it in fact belongs to you because you keep the industry in balance! Get up and fight!

    • [+Duracell-]
    • 8 years ago

    I was hoping that Bulldozer would be able to beat the 1100T in most, if not all, scenarios, but I guess I was overly optimistic. I’m really disappointed with the power consumption. I thought it would have been able to at least beat the 1100T, but it ended up using more. I wonder how much it consumes during typical workloads and games, though. Cinebench seems like a worst-case scenario as far as loading the CPU goes.

    Dropping in an 1100T into my AM2+ system sounds mighty tempting right now…

    • Coran Fixx
    • 8 years ago

    I wonder if there will ever be a VH1s Behind the Music on what went wrong with Bulldozer process. How many of you guys will buy if it is a drop in replacement for you when the price drops 10-20%? 0?

    Intel will have less of an incentive to release Ivy Bridge on time or at a reasonable price.

    AMD/Intel could pay any average Gerbil here onetenth of a marketing execs salary to sit on a rock until they came up with better chip names then Ouchie and Evacuator (As in evacuate your AMD stock options on your way out the door)

    The worst scenario I can think of is that new AMD leadership decides to abandon the enthusiast segment completely to focus on tablet/cellphone hardware.

      • Krogoth
      • 8 years ago

      Nothing went wrong with Bulldozer.

      The problem is that some people were expecting to be a powerful desktop chip, when the underlying design made it quite evident that it was meant for server and workstation related tasks.

      The days of a single CPU architecture to rule them all are over. We are going into the age of specialized CPU architectures gear towards certain tasks and niches.

        • Kretschmer
        • 8 years ago

        Krogoth,

        i5 is a powerful desktop chip that still destroys Bulldozer in most server and workstation tasks. What is BD specialized for…heating your data center?

          • Anonymous Coward
          • 8 years ago

          [quote<]What is BD specialized for...heating your data center?[/quote<] I think Krogoth is just making it up. Oops, somehow this reply ended up on the wrong post.

          • axeman
          • 8 years ago

          You have some i5 benchmarks of server workloads handy? And some Interlagos benchmarks for comparison?

          • Krogoth
          • 8 years ago

          Where are these server and workstation related tasks?

          TR’s benchmark suite only covers mostly gaming/desktop stuff with a few workstation stuff. Bulldozer manages to get up the heels of its SB counterparts in those workstation related stuff.

          People forget that the X6-1100T is a capable workstation chip in its own right.

        • travbrad
        • 8 years ago

        [quote<]The problem is that some people were expecting to be a powerful desktop chip[/quote<] Well it's easy to see why people might have thought that, with the way it was being marketed by AMD. 8ghz overclocking runs are clearly aimed at "impressing" enthusiasts, not server admins. That 8ghz had worse performance than a stock 2500K of course (1 core vs 4), but they left that part out. 😉 Everyone should take marketing/PR with a large heap of salt, but when AMD repeatedly targets so much of their marketing at gamers/enthusiasts you can't completely blame people for expecting a desktop cpu. It's not that great even in multi-threaded stuff though..certainly not for $250

        • axeman
        • 8 years ago

        Wow, downvotes. Boo to you just downvoting because it’s Krogoth. He’s right.

        • TravelMug
        • 8 years ago

        “Nothing went wrong with Bulldozer.”

        A few things did, in fact. If I only have to pick one it’s that they didn’t reach the projected clock speeds, hence the abysmal performance results compared to even their own X6. Then you have the laughable cache/memory subsystem performance as a “bonus” to that.

          • khands
          • 8 years ago

          If this architecture had come out back in 2009 (or whenever the initial release date was) we’d be singing a whole different tune. C’est la vie.

            • paulWTAMU
            • 8 years ago

            ….duh? But it didn’t, and it’s competing against what intel has in late 2011, not what intel had in 2009.

            My last 2-3 CPUs have been AMD, but at this rate? If I build in the next year or two it’ll be intel. This is just a major bust as far as desktop performance goes.

            • khands
            • 8 years ago

            I’m just saying they floundered for too many years, hopefully they actually start meeting their deadlines for upcoming architectures else we’ll have an Intel monopoly in 4-5 years.

            • paulWTAMU
            • 8 years ago

            Gotcha. Yeah, I dont’ want an Intel only state of affairs. Maybe Via could step up 😛 This is such a frigging let down.

            • khands
            • 8 years ago

            I think we’ll actually see ARM desktop CPUs before VIA becomes interesting again :/

            • chuckula
            • 8 years ago

            Really.. you want to play the time warp game? If the Pentium IV had come out in 1987 it would have completely destroyed any computer that could fit under a desk in performance too.. but that doesn’t mean it was any good.

            • khands
            • 8 years ago

            Bulldozer was originally supposed to launch then, AMD’s inability to get anywhere close to their launch date has been the bane of their existence for quite a long time now.

          • Krogoth
          • 8 years ago

          Did you even read the article?

          Bulldozer has an excellent memory controller, its memory bandwidth and latency are comparable to Gulftowns despite being dual-channel DDR3.

          Cache speed is a mixed bag, but Bulldozer pulls ahead once you go beyond 2MiB.

            • TravelMug
            • 8 years ago

            Yes, I did. That bandwidth improvement means nothing here. Their goal was to improve the integer performance apparently and the increased bandwidth won’t help you there. Lower latencies would, but those are worse than their previous products not to mention the current Intel line. Here’s a nice table with results:

            [url<]http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/6[/url<]

        • Kaleid
        • 8 years ago

        Since it’s not really faster than PII then yes, it’s a big failure.

        • sschaem
        • 8 years ago

        Allot went wrong.

        a) The voltage had to be pushed way above AMD expectation for the FX processors
        b) The cache is running at half its expected rate
        c) They didn’t give developers , specially microsoft info on their processor scheduling requirement in time.
        d) It was late almost 6 month from GF delays

        What happen to the overall picture when AMD can stop overvolting its FX processors?
        When it can fix its cache issue? and finaly can get windows sheduler to play well with bulldozer?

        I think what we saw so far is not the best bulldozer as to offer. Because allot went wrong 🙂

        Disclaimer, none of the tweak will chnage the fact that Sandy ^ Ivy bridge will for ever continue to be the fastest 2 and 4 core design.

    • yogibbear
    • 8 years ago

    Considers yanking out my Q9450…. LOL.

    Decides to twiddle thumbs till the next TOCK.

      • ronch
      • 8 years ago

      Hey, aren’t you the world’s smartest bear??

    • yogibbear
    • 8 years ago

    Looks at BFBC2 benchmark…. sniggers….

      • Arclight
      • 8 years ago

      I know , right?

      1100T 76fps > FX 8150 75 fps>A8 3850 69 fps = FX 8120 69 fps

      Srsly A8 3850 on par with 8120 and 1100T beating the new 32 nm, 2 billion transistor flagship? What a flop….

    • sweatshopking
    • 8 years ago

    what about this? “The FX-8120 costs $205 and it’s unlocked — all FX processors are — so it can be easily modified to match or exceed the operating specifications of the FX-8150. Therefore we feel the cheaper FX-8120 gives the Core i5-2500K a serious run for its money and it’s a worthy alternative. Meanwhile the FX-6100 is also great value at $165, as it undercuts both the Phenom II X6s but it wasn’t always faster.”

      • mboza
      • 8 years ago

      Only on heavily threaded benchmarks, mainly the ones where the 8150 loses to the X6s. And comparing an overclocked FX-8120 to a stock i5-2500K (or even a i5-2500 at the same $205 price) is getting a little apples to oranges.

        • NeelyCam
        • 8 years ago

        Yep. In fact, an alternate way to look at this would be to overclock [i<]2500K[/i<] instead to the same load power consumption point as 8120 or 8150, and see how the performance is. In a way this would be apples/apples, as it implies equal energy cost and cooling noise/expense.

    • ALiLPinkMonster
    • 8 years ago

    :'(

    • chuckula
    • 8 years ago

    Randy Allen Brought forward into 2011 says it all! Blow away Sandy Bridge in Every Dimension!

    [url<]http://www.youtube.com/watch?v=pSw2GTLHtQo[/url<] Performance Per Watt is growing! Or is that watts per core? Oh well.. something is growing!

      • RtFusion
      • 8 years ago

      Hehe, also found this one too:

      [url<]http://www.youtube.com/watch?v=G_n3wvsfq4Y[/url<]

    • chuckula
    • 8 years ago

    Look you Intel Fanbois! This is a great victory for AMD! Intel totally cheats and AMD is at a major disadvantage! Look at all the unfair and inhuman limitations that AMD’s engineers had to work with when they Bulldozed Intel:

    1. 100% Larger transistor budget [edited once I found out how big BD *really* is]
    2. Larger die size
    3. Higher Clock Speeds
    4. Not having to dedicate any resources to GPU
    5. Larger power envelope
    6. Five years to come up with a new design instead of four years.
    7. Ability to study multiple generations of competitor’s designs before having to come up with anything new.

    Intel couldn’t design a pocket calculator with these cripplingly horrible chains that were bound to the AMD engineers by Intel. I mean AMD should get a Nobel — interruption, whispering — Uh… I’ve just been informed that some of you stupid Intel fanbois think that the points I raised above aren’t horrible limitations that prevent the pure Geniuses at AMD from designing a chip. Well, just look at the benchmarks: OBVIOUSLY AMD is at an unfair disadvantage!

      • ermo
      • 8 years ago

      Time will tell whether AMD can refine the BD concept with new steppings and make it run closer to its full potential. For starters, GloFo and AMD will need to work well together for this to have a chance of happening.

      Whether this full potential is enough is an entirely different kettle of fish.

    • chuckula
    • 8 years ago

    AMD Legal had to edit the marketing info for Bulldozer a bit:

    This is the greatest day in the history of AMD!* I mean…. look at the scores!** The Eight Core chips are INSANELY cheap compared to Intel’s!***
    The FX-8150 completely blows the 6 core chips away in every dimension!† The power consumption per core is amazing!†† Gaming benchmarks show Bulldozer’s INSANE power compared to those overpriced Intel Chips!‡ Magical OS patches FTW!‡‡ AMD has out innovated Intel AGAIN!‡‡‡

    * Except for the K6-III, Athlon, and Athlon 64… we’re on the fence about Phenom.

    ** Looking at the scores may cause nausea and eye bleeding in AMD fanboys, please consult with your doctor before looking at the score.
    *** Comparisons being made to 8 Core EX series processors ($3,897.00 on Newegg). No other Intel chip could possibly compete… right??

    † In at least some benchmarks, as long as you don’t overclock them to the same speed as Bulldozer. Oh, and we mean 6 core Phenoms, the 6 core Intel chips broke our meters when we ran the benchmarks.

    †† Just take 125 / 8 = 15.625 watts per core. Totally better than 95 / 4 = 23.75 watts per core. Any other power measurement is a lie told to you by Intel marketing.

    ‡ By which we mean a low to mid-range GPU at high resolution will give roughyly the same frame rate in games and only the $1000 CPU can
    possibly ever be used to play games, the 2500K cannot be used with any form of GPU at all, ever, period.

    ‡‡ Magical OS patches include source code not yet accepted into Linux to fix cache aliasing due to poor design choices in making the modules. To quote Ingo Molnar: “It could also go away easily if this bug is fixed in a future CPU.” (http://us.generation-nt.com/answer/patch-x86-amd-correct-f15h-ic-aliasing-issue-help-204200361.html)

    ‡‡‡ The term “innovation” shall be construed to mean: 1. spin off the assembly to another company; 2. Sell huge chips at cut rate prices;
    3. manipulate the numbers to look like you are making a huge profit while the fab eats the loss.

      • ermo
      • 8 years ago

      Interesting thread on the Linux scheduling issues. The comments do seem par for the course, though.

      Quirks are nothing new, and subsequent patchset posts confirm that AMD will keep refining the patch until it’s politically acceptable, which really just means that the kernel developers have gotten used to the idea and told AMD how it needs to look before they’ll accept it, I guess.

      • NeelyCam
      • 8 years ago

      [quote<]3. manipulate the numbers to look like you are making a huge profit while the fab eats the loss.[/quote<] LOL, I remember that one!

    • bogbox
    • 8 years ago

    Now we have the reason why Dirk Meyer was sacked.

    AMD design was wrong. I mean creating the modules .This is not a 8 core processor ,is 4 module processor.AMD said :One integer core is 50% less then a Phenom core .But also much smaller than core.

    Honestly I was expecting this,but the single core performance was weak. 2000 millions is waste, GF is really having big problems.

    This is like Fermi.Big massive inefficiency .

      • Silus
      • 8 years ago

      At least the GTX 480 actually beat the competition, despite being big and power hungry…If you want to compare this with a GPU launch, compare it with AMD’s first GPU, R600, which was big, hot and could barely catch NVIDIA?s second best, the 8800 GTS 640 at the time.

    • I.S.T.
    • 8 years ago

    Sad day for AMD, but for those who’ve been reading up on the Bulldozer arch before today, not entirely surprising.

    I do have a question about the benchmarks. Namely the Metro 2033 1680×1050 Medium Detail benches. The i7 2600K is nearly twice as slow in low FPS as the 2500K. SMT’s problems rearing their ugly head again?

    • elmopuddy
    • 8 years ago

    I’m still somewhat confused why they get away with calling this an 8 core CPU, its just a slighty better version of HT ( some more hardware) and not true full 8 cores, even in the review Scott showed some concern over thread scheduling.. a concern that wouldn’t exist with “real” cores.

    I didn’t expect much, but I am still dissappointed with this product, was hoping for it to be a little more competitive at least.

      • shank15217
      • 8 years ago

      Windows treats HT cores differently than real cores, the point is with AMDs current configuration, two threads that have nothing to do with each other maybe scheduled on the same module, and two threads that do share data maybe scheduled on two separate modules. This would be minimized with Intel’s HT aware scheduler.

      • just brew it!
      • 8 years ago

      I think calling it “just a slightly better version of HT” is a bit harsh, though I see where you’re coming from. For integer workloads it is essentially an 8-core CPU; for floating point it is more like 4 cores + HT.

        • bcronce
        • 8 years ago

        “For integer/x87 workloads it is essentially an 8-core CPU; for AVX it is more like 4 cores + HT.”

        There, fixed that just a bit.

          • just brew it!
          • 8 years ago

          I don’t think that’s correct.

          I’m pretty sure SSE goes through the shared FPU as well.

          I also haven’t seen anything to imply x87 gets executed in the integer ALU. x87 is largely irrelevant these days anyhow; AFAIK 64-bit compilers generate SSE2 instructions instead of x87 instructions for floating point math.

            • Damage
            • 8 years ago

            x87, SSE, AVX, and even MMX/XOP (integer SIMD) are all handled by the FPU.

            But then the FPU can track two threads, so it’s more like a Hyper-Threaded quad-core arrangement than a four-thread quad-core. FWIW.

            • bcronce
            • 8 years ago

            SSE is only 128bit. The FPU is 256bit. The only way for a single core to monopolize the FPU unit would be to run a 256bit AVX instruction

            The “Flex FPU” can do any combination of these at the same time

            4 32bit floats, 2 64bit floats, 128bit SIMD

            Both cores can do any of those at the same time without stepping on each other’s toes. But a 256bit AVX instruction will require one core to completely take over and force the other to stall if the other also needed the FPU.

    • ermo
    • 8 years ago

    *wipes egg off face*

    *looks down drain*

    “Hello? Anybody hooome? Dirk? Barcelona? … Anyone?”

    The only bright spot so far: 7-zip can utilize all 8 threads with optimized code paths and give us a glimpse of what BD’s ‘forward looking’ ‘throughput-oriented’ design can do when utilized fully. Likewise, x264 also seems like it can be optimized to take better advantage of BDs architecture. Of course, this just reflects that the cost and risk associated with enabling new code paths for a small open source tool is low.

    But Games? How many major titles will put out a patch just for BD? Answer: Not a lot. Not economically feasible. And in any case, only the already multithreaded stuff would likely benefit.

    And commercial desktop/workstation software tools? Why bother, intel owns this space anyway. I do wonder if these suites only enable SSE3+ features on intel hardware, based on the (at the time correct) assumption that AMD hardware didn’t support it?

    So yeah, the results seen in the TR test suite are likely going to be representative for the next 6-12 months, and even then, it’s hard to be optimistic, much less excited for BDs performance prospects on the desktop going forward, even if it may carve a niche out for itself in the server market.

    This is a D- for GloFo and AMD. The only reason it doesn’t get an F is the (well hidden, apparently) potential for optimizations that the open source tools hint at.

    They say that you learn the most when you fail. Well, AMD + GloFo has a lot of learning to do, it seems. Here’s hoping that both will have got their act together by the time PileDriver gets released.

    That 955BE I just bought two months ago for what amounts to peanuts and plopped into my old 790FX DDR3-1600 system is looking like an awfully good value right now…

    • Althernai
    • 8 years ago

    [quote<]At the same time, Bulldozer resurrects a concept that's fallen out of favor in PC processors in recent years: it's a "speed demon," optimized for higher clock frequencies rather than maximum instruction throughput in each clock cycle. The Pentium 4 "Netburst" microarchitecture—particularly in its troubled "Prescott" incarnation—gave frequency-optimized designs a reputation for high power draw and iffy performance.[/quote<] It's fallen out of favor for very good reason: high power draw and iffy performance is exactly what you get here. Why did they ever think this was a good idea? Even Intel gave up on CPUs that require a high frequency to be effective and Intel's manufacturing prowess is second to none.

      • Farting Bob
      • 8 years ago

      This was the point in the review that i know it was going to be as disapointing as i had feared. They based their whole architecture on one of the worst CPU design choices Intel have ever made, rather than the excellent designs they have done ever since.

    • dale77
    • 8 years ago

    It is great that AMD finally got this chip into the market. Shame it doesn’t improve markedly on the Phenom X6, but nevermind, hopefully they’ll be able to build on this new architecture.

    • ronch
    • 8 years ago

    It’s really interesting that AMD was touting performance/watt/dollar a year ago when they revealed major details about the BD architecture. Sadly, it now looks like they missed all three by a mile. Performance is nothing revolutionary considering BD only has a few bright spots, and touting 8 cores + more cache + higher clocks for your money vs. only 4 from Intel doesn’t help when users see that performance is nowhere near where they expected it to be. Power is nothing amazing compared to K10 either, especially considering it’s built on a smaller node with lots of power management features specially built in. And that large die just isn’t going to bode well for AMD’s profitability, especially with rumors that GF can’t get 32nm running very well. As Intel is set to move on to 22nm, AMD is stuck with a low-yield 32nm node trying to make large chips. Oh boy.

    I have no idea either why AMD marketing decided to call this ‘FX’. It just kinda ruins the brand and the prospect of using the FX name in future, truly impressive products should AMD manage to pull them off. It also somewhat deludes/fools buyers into thinking they’re getting something as ground-breaking as K8-based FX chips once were. If you ask me, this should have been called Barcelona II, except this time, AMD is not one, but two, architectural generations behind Intel before this architecture came out. AMD marketers sure are a lousy bunch, and I couldn’t care one bit for them except perhaps their decision to unlock Phenom II and FX, if it was really them who convinced management to do so in the first place.

    On a more positive note, I’m glad AMD can now focus on tweaking BD. Being the first officially released revision I expect BD to have a lot of quirks that AMD simply didn’t have the time to fix in order to push a 2011 launch. I would think AMD should focus some effort in getting the branch predictors to work more efficiently, in light of the long pipeline. Also, in time they could lower TDPs by fine-tuning the silicon, Turbo Core, and as GF’s 32nm node matures. Thirdly, I hope they could tweak BD to achieve higher clocks to meet their original aspirations. Lastly, Windows 8 should provide better thread allocation more well-suited to BD’s architecture, although asking me to get Win8 just so I can get the most out of my FX system isn’t really something I’d like, especially given the early indications that Win8 will be a radical shift for Microsoft.

    Overall, reading this article gave me mixed feelings about where AMD stands now. Bulldozer is out and we all finally get to learn more about the architecture we’ve all been excited about for the past year. But it’s also disheartening to see how BD has Phenomenally failed to deliver on its goals, particularly in the desktop space. lts chances of succeeding in the server space may be brighter though. I don’t know. Let’s see what AMD does from here.

      • ermo
      • 8 years ago

      Right now, it can’t be much fun either at AMD headquarters, or at GloFo. They must have been bracing for impact for months now, as evidenced by them having already put out slides of the improvements targeted for the next stepping.

      Looking on the bright side, there’s at least room for improvement. *Ahem*.

      [quote<]But it's also disheartening to see how BD has Phenomenally failed to deliver on its goals, particularly in the desktop space.[/quote<] Ooooh. I see what you did there.

      • Voldenuit
      • 8 years ago

      [quote<]I have no idea either why AMD marketing decided to call this 'FX'. [/quote<] The 'FX" stands for "fuxx0red".

      • squeeb
      • 8 years ago

      Yea, what a waste of the “FX” name. The original FX chips were pretty bad ass.

      • madseven7
      • 8 years ago

      Like what you wrote, but I strongly disagree that they didn’t have time to push out a 2011 launch. They have been working on this for years. There is not excuse for this crap. Heads must roll! Starting with the board. If I were the investors I would pull the plug and let the just get bought out or die. Wasted money for what, more payouts? underperforming chips?
      The board has to go! What have investors said?

        • NeelyCam
        • 8 years ago

        [quote<]What have investors said?[/quote<] "Me likey this unrivalled FX thingy. Can I please buy some moar stock?" (-> AMD stock up 1.5% today) How is it that investors don't know about this yet..? On Yahoo! Finance's AMD page, the only news headline related to BD is: "Unlock Your Record Setting AMD FX Series Processor Today (Marketwire)" Ah, never mind. It'd down 1.5% after hours.

          • khands
          • 8 years ago

          That’s still a fair amount less than I was expecting, but I suppose that’s good on AMD’s part then.

          • sschaem
          • 8 years ago

          AMD reached its 52week low October 4th when bulldozer benchmarks of final shipping version got disclosed under the table.

          Could AMD go lower, yes, I wouldn’t get it passed AMD to destroy another billion of its worth.
          But I think the Zambezi B2 steping reality is already baked in the stock price for a while.

        • sschaem
        • 8 years ago

        The board should be dissolved but not for this reason. And investors are pulling the plug on AMD, but I think its more over Windows8 and other higher level problems.

      • JumpingJack
      • 8 years ago

      [quote<]I have no idea either why AMD marketing decided to call this 'FX'. It just kinda ruins the brand and the prospect of using the FX name in future, truly impressive products should AMD manage to pull them off.[/quote<] This is he underlying tragedy here. AMD withdrew the FX brand shortly after Conroe specifically because that brand was meant to represent the cream of the crop and best in class. They were holding off using it, and switched to "Black Edition", until they could get to a point where it was meaningful. Reanimating that brand with Bulldozer does to the FX brand what the Pentium 4 did the to overall Pentium brand (which is now relegated to the low end just a notch above Celeron). Sad really.

      • Theolendras
      • 8 years ago

      Well they will need better turbo management with real sensor to push the thermal more granularily. This does look a lot like Barcelona launch, a launch with a transitionnary product, not fully mature. But pressure from Intel is very hard at this time, and they got almost no choice to go forward. Still the basis might be good enough to build upon. The concept is different, and I for one give them credit for that, and might be interesting in a few server workloads, especially integer heavy ones. Still it is an much bigger breakup architecture than Barcelona, from which they need to get to their goals, bring them up in clockspeed is one of them.

      Still for a consumer point of view it feels unrewarding for most. That’s the pain of low ressources must go troutgh, Intel is spending almost as much on R&D than AMD ever will. So this is a transition, with some bumps.

      Piledriver will really need to deliver, but it is not impossible. By the time of it’s release, it might be pretty close to Windows 8 launch, giving it a 5% boost from the scheduler on many loads. The 32nm process will hopefully mature, with maybe a 10% boost trough clockspeed. AMD is anticipating a 10-15% boost from IPC as well. FMAC and AVX seems to perform fine and will probably a little more revelant in some realworld benchmark by then. I would expect them to enhance further still the turbocore feature in Piledriver to get another 5% from single thead.

      So if this is all realised, the optimistic figure would put bulldozer roughly in 30 % higher performance from the same thermal enveloppe in single-thread and 30% in multi as well. Sure Intel will be at Ivy bridge by then giving them the 7% they anticipate on IPC trough little tweaks to SB and maybe 15 % from the process.Speculatively putting them a little closer. But then Intel has a lot on it’s hand, lower thermal budget right now, getting hyperthreading further down the line, or Quicksync for that matter.

      What is disenhearting is the sharing concept does not seems to save transistors so much… Almost feels like there is disabled logic, to get to the market in time, or diminish a thermal/leak problem.

      If Piledriver is another disapointement, They will really need the GPGPU to get steam, otherwise they’re pretty much cooked. Maybe then it could put itself up for sale then. Hey Samsung, you’ve got deep pockets and lots of logic capabilty and interest, would you mind.

        • ronch
        • 8 years ago

        I’d hate it if AMD gets acquired by Samsung, and I’m not sure the FTC or some other regulatory body would allow it. I’d rather have IBM snap them up.

          • destroy.all.monsters
          • 8 years ago

          Without the x86 license – which does not transfer – AMD’s value is entirely on the graphics side. Why would Samsung buy them when they can license a core from Imagination Tech? IBM has no reason to buy them at all and doesn’t even have an x86 license any more.

          Please stop bringing up unworkable scenarios. There is no white horse in the background. AMD will stand or fall on its own. The only possible buyer would be ARM and they’re not looking to buy, just partner. Oracle might want them just to get special deals from intel however Oracle has no interest in the consumer space.

          Again, the x86 license does not transfer.

            • FuturePastNow
            • 8 years ago

            AMD owns an awful lot of patents in its own right.

            • destroy.all.monsters
            • 8 years ago

            And at a fire sale price they’d buy them for the patents, possibly spin off the graphics division, and kill the x86 side (which they’d have no choice in).

            The real issue is that ronch – and others on the site – seem to think that someone will swoop in and save AMD like Superman rescuing Lois Lane and somehow magically keep the consumer and x86 sides when that’s pretty much impossible (or more correctly – no business has doing so in its own best interest).

            AMD has done damn well at being less than a tenth the size of intel and putting out competitive product despite having management, board, and resource issues. I find the whole “if it’s not as top notch as intel then it’s nothing” sentiment short-sighted and frankly idiotic. AMD and intel are not equals.

            tl,dr: I agree that harvesting AMD for patents and killing or selling the rest particularly at a fire sale price would be a good plan for a handset developer or in some limited smaller circumstances. There is no white knight.

            • ronch
            • 8 years ago

            Chill out, man. We’re just speculating. I’m fully aware that it’s not that simple, legal-wise and business-wise for both AMD and IBM. When Barcelona came out and AMD looked like they’re about to sink, there was also a lot of talk on the streets about IBM snapping up AMD.

    • UberGerbil
    • 8 years ago

    One correction: on the AVX page, you have[quote<] At this point in time, finding applications or benchmarks that make use of AES isn't easy.[/quote<]While that's probably true, in context it looks like you probably meant AVX rather than AES there.

    • geekl33tgamer
    • 8 years ago

    Wow, underwhelmed much??? I’ve got a X6 1100T already, and was expecting the FX-8150 to obliterate it. But it didnt happen – AMD’s not really closed up the performance gap on Intel to any great degree in common applications (games, encoding etc), and when Ivy Bridge comes out, the gap gets even bigger… 🙁

    • entropy13
    • 8 years ago

    [url<]http://i47.tinypic.com/2vnhzra.gif[/url<]

      • shank15217
      • 8 years ago

      hahahahahaha

    • achaycock
    • 8 years ago

    OK, I can see I’m not alone in being hugely disappointed with AMD’s new FX processor, especially when you consider that it was their top end SKU being tested here. I am curious though about the possibility of overclocking the northbridge/L3 cache ala Phenom. My 1090T runs at 3.8GHz on air but more importantly the NB is at 3GHz and this yielded massive performance gains even without the cores being overclocked. In fact I prioritised NB over core overclocking because of this (hence cores not at 4GHz+). I’m not being a Bulldozer apologist, its performance or lack thereof is frankly inexcusable no matter how forward looking the design, but I would love to know if NB clocks matter.

      • entropy13
      • 8 years ago

      A review can’t get past 12×200 on the NB.

        • achaycock
        • 8 years ago

        I’m concerned there. A 200MHz increase wouldn’t really yield any decent improvements even if an overclocked NB did help. Shucks, that’s a real disappointment. I hope this makes for a decent server part otherwise this architecture is a dangerous failure. Phenom II has actually been a pretty decent processor, but lacks the halo effect of Intels’ Core architecture. AMD just cannot afford a failure right now.

    • beck2448
    • 8 years ago

    tragic

    • Ashbringer
    • 8 years ago

    All these years I’ve been patiently waiting to see Bulldozer in action, and what I see is very sad.

    Most of the time it’s slower then a X6 or X4 from AMD. Its most likely going to cost more then those chips, and still doesn’t come close to Intel’s flagship processors.

    Unless there’s something like 3DNow that isn’t being utilized yet in applications, I see no reason to go crazy for one. I got a Phenom X4, and I see no reason to look at upgrading.

      • ronch
      • 8 years ago

      Same here. I was hoping (not expecting) BD will give me something to look forward to when I ditch my Phenom II sometime in the future, but I guess my Phenom II will suffice until IB, Haswell, or AMD gets BD running more smoothly. As it is, I hope today’s BD rev. is just reminiscent of how the K5 came out. Earlier versions didn’t meet expectations and were labeled SSA/5. Later revisions set the chip in order, but unfortunately it was too late. AMD had better move fast.

    • k00k
    • 8 years ago

    Man, those synth floating point numbers are tragic. Integer numbers are good though, but not good enough to bring AMD in competition with Intel’s current high-end lineup. Was AMD wrong in creating shared FPUs in each module? This feels like Barcelona/K10 all over again. We’ll likely see Zambezi parts discounted in a few months’ time.

    I’ll probably get a Thuban before the year ends to tide me over and wait for this microarchitecture’s refinement, hopefully, within a year or two. That’s a big ‘hopefully’, AMD. If not, I may just get my first Intel-based desktop in a decade.

    • Geistbar
    • 8 years ago

    Maybe I’m extrapolating wrong, but it appears to me that Bulldozer had it’s most (only?) impressive moments with work loads that, to my knowledge, should be mostly integer based, with floating point tasks being very dismal showings. My initial reaction was to think that not duplicating the FP units was a mistake, but I don’t think many of the FP tasks used scale past 4 threads anyway. So, am I just looking at it wrong, or might the floating point unit have some major flaws?

    As to the overall performance, it was pretty disappointing. I hope AMD can turn things around. Maybe AMD will figure things out and it’ll have more in common with the P4 than expected. Willamette to Northwood was a pretty significant performance improvement, if I recall (it happened just before I started following this stuff). Though, I think even that much of an improvement might not be sufficient for Bulldozer. We’ll see.

    • sschaem
    • 8 years ago

    AMD did chnage its front page to a bulldozer announcement… alright

    Marketing message

    “Perform mega-tasking and get pure core performance with new “Bulldozer” architecture”

    Was this translated from Chinese ? WTF ?

    I’m not making this up.

    “Enjoy stable, smooth performance with impressive energy efficiency thanks to a 32nm die”

    AMD is promoting that their cpu are stable ? (why is this even mentioned) and efficient ? (AMD must think we are all stupid?)

    And the new CEO is endorsing all this…

    AMD marketing is so lame, I’m going to bed.

      • destroy.all.monsters
      • 8 years ago

      It is pretty terrible. Trying to make a selling point of having more cores is a tough one. The second argument comes across as “you can trust us even though it’s not an intel processor”.

      Try not to have a conniption fit man. The older we get the higher our blood pressure is anyway.

        • Farting Bob
        • 8 years ago

        [quote<]Trying to make a selling point of having more cores is a tough one.[/quote<] No it isnt, its a vey easy one, along with clockspeed. The vast majority see clockspeed and core count as the only things about a processor. When they are shopping in best buy and one PC has "4 intel cores" on its label and the next has "8 cores" on it, alot of people will go with the 8 cores. When it comes to anything that people are not an expert on, find at most 2 numbers that you can highlight as being better (higher) than your rival and you will sell products.

          • OneArmedScissor
          • 8 years ago

          The problem is that people don’t go desktop shopping at Best Buy in the year 2011.

            • Waco
            • 8 years ago

            Of course they do – those people just aren’t the ones on this site (or any site, for that matter).

            • OneArmedScissor
            • 8 years ago

            I don’t even remember seeing desktops last time I was there. They have tons and tons of laptops, covering the main display area, and then a few monitors up against the back wall, but just the monitors. Places like Staples are also set up this way.

            Normal people buy laptops and it’s been that way for years.

      • just brew it!
      • 8 years ago

      Marketing has always been a (very) weak point for them.

        • sweatshopking
        • 8 years ago

        possibly because it’s hard to market a slower, hotter, more expensive chip?

          • ermo
          • 8 years ago

          +1’ed for delicious snarkiness. 😀

          • just brew it!
          • 8 years ago

          Intel did a pretty good job of marketing the Pentium 4…

            • kamikaziechameleon
            • 8 years ago

            AMEN!!!

          • shaq_mobile
          • 8 years ago

          If there’s anything we’ve learned from Apple’s marketing it’s that people need to be told what they need. Marketing isn’t reminding people that your device is at the bottom of Maslow’s Heirarchy, it’s telling them it is. But I do love the response. 🙂

            • kamikaziechameleon
            • 8 years ago

            ^^^this

          • bwcbiz
          • 8 years ago

          Well, up until the FX, they’ve rarely been more expensive than Intel. This is probably the worst mis-step by AMD since they fell behind Intel tech in the first place.

          • kamikaziechameleon
          • 8 years ago

          ^^^This

        • madseven7
        • 8 years ago

        Hire proper marketing team is all I have to say.

      • ronch
      • 8 years ago

      Funny, I couldn’t see FX on the front page of AMD’s website unless I dig down their products listing. Seems not all their regional websites were updated. AMD must not be too proud today…

    • Mr Bill
    • 8 years ago

    Several comments that Bulldozer is better for workstation / server configurations. I wonder how well these benchmarks would run inside a virtual machine?

      • UberGerbil
      • 8 years ago

      AMD does enable all the virtualization acceleration technologies (v.IO, nested page tables, etc) in all of its chips (and BD continues this AFAIK); this is in stark contrast to Intel, which loves to leave some of them out in the interests of product segmentation (want an unlocked clock but don’t want to pay for hyperthreading? Sorry, n vt-d for you in the 2500K).

      And BD’s relative strengths appear to be integer, as you say; moreover, it does remarkably well on the TrueCrypt bench that exercises the AES instructions, All of which points to it being a better (relative) competitor in servers. Virtualized web servers couldn’t care less about fp performance.

      On the other hand, if you’re looking for virtualization benchmarks to find it a competitive advantage on the desktop, well, that’s a pretty small niche. Granted, virtualization is getting more mainstream, but even ubiquitous “transparent” virtualization (sandboxing within the OS) is beyond the commercial lifespan of this iteration of BD.

        • just brew it!
        • 8 years ago

        Yup, that’s still a big selling point for some people. AMD leaves the virtualization stuff enabled. Everything from mid-range on up still has support for ECC RAM as well (though mobo makers sometimes choose not to implement it); contrast to Intel, where AFAIK the mainstream CPUs and chipsets don’t support it at all.

        • Deanjo
        • 8 years ago

        Virtualization (and IOMMU support) is a big factor for me. I will however hold off until all the linux patches are in the mainline kernel and by then we should see new steppings hopefully with a bit better performance. Until then I have no reason to replace my current 1090T rig.

        • mnecaise
        • 8 years ago

        It’s a small niche; but, some of us do care about virtualization on the desktop… On my development boxes I run VMs for testing and alternate OS work. Keeps it simple (one God box vs multiple workstations) and if I screw up I just revert to a prior snapshot. If the performance of the AMD chip is better in virtualization then I would benefit.

        I applaud AMD’s choice to enable all the technology in all their processors. In order to avoid the vagueness created by Intel’s product segmentation, my employer ended up buying me a Xeon based workstation. It was an expensive way to go.

        Having said that, I don’t think we can expect you to run the benchmarks inside virtual machines. There are to many factors that affect the performance… and each implementation has it’s own issues. I see significant differences between VMware Workstation and Virtual Box, for example (both of which I use regularly). Trying to run the benchmarks would become a study of the virtualization software as much as the underlying hardware.

      • BaronMatrix
      • 8 years ago

      With the several reviews I’ve seen the biggest takeaway is there is no way to tell which chips will beat which. There are cases where the 15 2400 is ahead of the 980X and the i7 970 is faster than the 2600K. There has long been talk of “non-optimizations” for this new chip and I see that a n issue. AMD has admitted that Win7 doesn’t really understand the module process and there’s probably thrashing going on as data is shuffled in and out of the L2 between modules.

      It’s clear from the AVX tests that AMD is not looking to support the older ISAs anymore but to move totally to XOP\AVX with FMA. The INT numbers were in lien though there seems to be some latency issues for the FP.

      Since Linux has a totally different scheduler mechanism I’d be interested to see how it handles the various mixes of instructions in the apps tested. I’d especially like to see OpenGL under Linux.

      All in all, it’s definitely not reaching its potential with the current OS and compiled software. I expect as we move closer to PileDriver and Win8 we will see more optimizations of Zambezi’s memory model. Since we are beyond the NDA we should start to see Linux LinPack numbers and SPEC testing. These will show the real potential through optimizations.

        • JumpingJack
        • 8 years ago

        Well Baron

        I must admit, I was impressed that your extensive simulations yielded a 20% number for IPC — as that is almost exactly what we observed, it is just that it was 20% in the wrong direction.

        [quote<]It's clear from the AVX tests that AMD is not looking to support the older ISAs anymore but to move totally to XOP\AVX with FMA. The INT numbers were in lien though there seems to be some latency issues for the FP.[/quote<] Clear, really?? Looking at the data, AMD's implementation of AVX looks fairly ineffectual, and FMA did not seem to really help much either. [quote<]Since Linux has a totally different scheduler mechanism I'd be interested to see how it handles the various mixes of instructions in the apps tested. I'd especially like to see OpenGL under Linux.[/quote<] News flash! The OS does not schedule instructions, this is done in the hardware schedulers as they pick which instructions execute based on several criteria. The 'scheduler' you read about in the OS is how it dispatches threads to various cores (you know, thread, the instruction stream which is comprised of sequential series of instructions strung together which are then executed out of order and in parallel when ever possible). Also, I thought OpenGL (under Linux or Windows) is a graphics API, why would it be interesting on Bulldozer? You think it waves magic pixie dust? [quote<]All in all, it's definitely not reaching its potential with the current OS and compiled software. [/quote<] You got this from your extensive simulations I suppose. [quote<]I expect as we move closer to PileDriver and Win8 we will see more optimizations of Zambezi's memory model. [/quote<] What memory model would that be? Are you talking about the cache system? Looking at the data, it may be beyond repair, it would take a complete redesign. XXOO Jack

    • smilingcrow
    • 8 years ago

    The TDP signposted the poor power efficiency months ago:
    Intel 95W (CPU+GPU) versus AMD 125W (CPU).
    So even if performance matched the i7-2600 it was still going to be trounced in the performance per watt stakes.

    FX-8150 v i7-2600K – idle / load

    76W / 209W v 64W / 144W (TR – Cinebench)
    85W / 229W v 78W / 155W (Anandtech – x264 HD)

    Viewed as the absolute difference between the two platforms: idle / load
    12W / 65W (TR)
    7W / 74W (Anandtech)

    Viewed as the relative difference between the two platforms as in AMD consumes x% more than Intel: idle / load
    18.8% / 45.1% (TR)
    9% / 47.7% (Anandtech)
    36.9% / 51% (Xbitlabs)
    42.9% / 57.6% (Hothardware)

    As above but using estimated values for using an IGP for which I deducted 25W from all values: idle / load
    30.8% / 54.6% (TR)
    13.2% / 56.9% (Anandtech)

    Edit: Added Xbitlabs and Hothardware but only for 1 table.
    Note: Ivy Bridge is going to make the power efficiency difference much larger I presume.

      • sschaem
      • 8 years ago

      AMD message to us all regarding FX power efficiency:

      “Enjoy stable, smooth performance with impressive energy efficiency thanks to a 32nm die”

      Realistically, who here can say that they where impressed with the FX energy efficiency ?

        • just brew it!
        • 8 years ago

        Impressed? No. About the best thing you can say about the energy efficiency is that it didn’t totally suck. Which would be forgivable if they would at least price the damn thing competitively.

          • smilingcrow
          • 8 years ago

          Considering it’s the latest AMD platform versus a 9 month old Intel platform which is due to be replaced in 6 months time or less and they are both fabricated at the same node I’d say the power efficiency really sucks.
          If you average the differences across the 4 reviews I quoted above the AMD platform consumes 27% more at idle and 50% more at load. And that’s for the whole platform with Xbitlabs showing that the CPU itself is consuming 86% more. They also show that the FX-8150 with 1 thread loaded consumes the same as the i7-2600K with all threads loaded.

          • sschaem
          • 8 years ago

          If you use cinebench this is not the case:

          FX-8150 power usage: 14.3
          i7-2600K power usage : 8.5

          Bulldozer require 70% more energy to accomplish the exact same task!!! ***70%***

          Build a render farm and your energy cost will almost DOUBLE. How can you state that this doesn’t totally sucks?!

      • gc9
      • 8 years ago

      XBit labs reported that the 990FX has 5-6 watts higher TDP than the 990X or 970.
      [quote=”xbit-labs”<] [url=http://www.xbitlabs.com/articles/cpu/display/amd-fx-8150_5.html#sect0<]Desktop Platform Refresh[/url<] [code<] AMD AMD AMD 990FX 990X 970 SB950 SB950 SB950 ... TDP 25.6 W 20.0 W 19.6 W [/code<] [/quote<] Everyone seems to be testing only the 990FX chipset. However the tests typically do not use the extra I/O capability of the 990FX. I wonder if the 990X or 970 are more efficient, or just have lower max.

    • cygnus1
    • 8 years ago

    Well, I think AMD is foobar. There’s no way they’re going to be able to make enough money, or doop any more investors, in order to catch up to Intel’s performance and process lead. The Bulldozer tech might do well in the server space, but I do not see this doing well scaled down to ultrabooks or tablets. Piledriver needs to get here ASAP. Hopefully one of those clocked down below 3GHz with the integrated Radeon can compete against Intel at least in the budget laptop space where there is lots of volume.

      • sweatshopking
      • 8 years ago

      fubar*

        • NeelyCam
        • 8 years ago

        No C for u.

    • shank15217
    • 8 years ago

    Its clear now why Dirk was let go (mediocre mainstream parts, unclear mobile platform), this type of performance doesn’t bode well for anything except certain types of server workloads. This architecture has a lot of potential but it didn’t meet AMDs own performance target, that’s very clear. Windows 8 and software optimization will improve performance by 5-8% over the course of weeks but Llano seems to have better IPC than bulldozer.. why didn’t AMD slap 8 llano die together, remove the GPU and improve the UN-core a bit.. it would have fared a lot better than this chip.

    • just brew it!
    • 8 years ago

    I’ve been saying for a while now that Bulldozer sounded like an architecture that was aimed more at the data center than the desktop. This more or less confirms it.

    Disappointing performance overall. Probably not bad enough to take AMD out of the game entirely, provided they can get decent yields and adjust their prices to be more in line with the performance; but certainly mediocre enough that they’re going to continue to struggle.

    The take-home message for me personally (given that I currently own several AM2+ and AM3 systems) would appear to be that I should just drop some Phenom II X6 CPUs into my existing boxes rather than doing an AM3+ refresh, and sit tight for a year or two. Anyone have any idea how quickly the Phenom II parts will start disappearing from the distribution channel?

      • shank15217
      • 8 years ago

      It doesn’t look that good on server workloads either.. 10% lower IPC and 4 extra cores doesn’t give me much confidence. Shrinking Mangy-Cours to 32nm would have given AMD more breathing room than this monster.

        • just brew it!
        • 8 years ago

        [quote<]Shrinking Mangy-Cours to 32nm would have given AMD more breathing room than this monster.[/quote<] I wonder if that might still be an option...

    • kristi_johnny
    • 8 years ago

    had High Hopes for Bulldozer, 2500k is a better deal, not to mention the arrival of Ivy Bridge in ~6 months time. Pity.

    PS: if AMD sells it for like 180-200$, than it would be an option.
    Unfortunatelly, Intel runs alone in this race .

      • sschaem
      • 8 years ago

      And the 2500K comes with a free GPU and a HW trans-coding engine…

      I think AMD didn’t price the FX line correctly. The fx-8150 as it stand should have a $195 MSRP.

      AMD botched the Bulldozer marketing and message.

        • just brew it!
        • 8 years ago

        Yup. If the yields are good enough that they can make it competitive on price/performance (by lowering the price) and still make money on it, they need to do so pronto. Otherwise, there’s really no good reason for people to buy this chip.

        • madseven7
        • 8 years ago

        No they botched the CPU period. This piece of crap should be cheaper than $195. Should be more like $180. It had at times worse performance than the I5 2400.

    • Lazier_Said
    • 8 years ago

    Poor power efficiency on the server side.
    Very poor single/dual threaded performance on the desktop.
    Very large die prevents value pricing on either.

    None of which is any surprise because AMD no longer has the development resources to stand up to an executing Intel even when they are executing correctly themselves. Let alone 6 months late and 500 mhz short.

    Forget about SB, it’s far out of AMD’s league. As a desktop product BD doesn’t make a convincing case against 18 month old Thuban. Or even against three year old Yorkfield.

    I didn’t expect it to be good but I sure didn’t expect it to be this bad.

    • sschaem
    • 8 years ago

    BTW, is it just me or the AIDA result look like bulldozer is not working correctly ?!

    And bulldozer L1 cache is 3 time slower then SandyBridge ?!?
    L2 cache nearly half the speed, and L3 cache nearly 3 time slower…. No wonder this chip is slower.
    and the ddr3 mem controller doesn’t even get 50% of theoretical ?

    Who ever work on the ddr3 controler at AMD need to be fired. Its been a disaster since the K8.
    How can BD not even get half of theoretical in a stream test? not even half of what intel gets with their implementation.

    I think the power consuption is also a bug, AMD had to raise the voltage (over-volt) their design and causing those huge power draw

    Its not impossible for AMD to lower voltage… but the L1, L2, L3, memory controller issue seem serious enough to hinder the design going forward.

      • Arclight
      • 8 years ago

      Indeed something feels wrong. Nonetheless there is no word if they intend to adress those issues and the borked product was already made commercially available. I wouldn’t hold my breathe, so if any friend of mine asks for advice when choosing between FX, Phenom II or Core i5/i7 my answer will be clear…..

        • ermo
        • 8 years ago

        Actually, they’ve already begun talking about the next stepping, which implies that they are at least trying to address the issues with the current stepping.

          • Anonymous Coward
          • 8 years ago

          This is [i<]just like[/i<] the original Phenom.

          • Arclight
          • 8 years ago

          Talking about new stepping and future releases is a given for a big company, but have they given a specific date for the release of a revision that will actually fix this arhitecture? Is it 2013? Man, *profanity* them.

            • OneArmedScissor
            • 8 years ago

            That depends on what you mean.

            For the next [b<]stepping[/b<], there's the FX-8170, which is 3.9 GHz base and 4.5 GHz turbo, and the rest are pushed down to 95w. There's not a set date, but it's not supposed to be very far away. Trinity, the Fusion version is [i<]allegedly[/i<] pushed ahead enough to beat Ivy Bridge, so that may just be a few months, and it uses the [b<]second generation[/b<] design, Piledriver, and also has an integrated PCIe controller. Assuming they also dropped the L3 cache, it should have all the advantages and none of the disadvantages. Just cross your fingers that they have a more dynamic system of trading off between the CPU and GPU. Llano was a real mess.

          • madseven7
          • 8 years ago

          Screw that, start from scratch. The architecture is utterly garbage.

            • Anonymous Coward
            • 8 years ago

            AMD never seems to launch a product that runs quite optimally, but no doubt that they will have versions of this which are much more competitive (if they go up against the current competitors anyway).

      • dragosmp
      • 8 years ago

      …add to this the CPU can hardly be clocked over 4.6GHz on air (<180W) and the CPU-NB can’t get over 2.4GHz.

      This architecture was probably targeted to start at 4GHz and go up, but it would up at 2.4-3 GHz.

      Design shouldn’t rely so much on manufacturing to push performance. AMD made a grave execution error putting so much weight on the shoulders of GloFo since now they don’t even control manufacturing they way they used to.

      • WillBach
      • 8 years ago

      Looking at the measured bandwidth to L1, L2, L3, and memory, the FX-8150 actually isn’t far behind the i5-2500K, especially for very small (8KB and under) and very large (1MB and over) blocks.

      If you’re looking at the i7-990X, that’s not Sandy Bridge, and it has 50% more cores than the 8150 has modules, and 50% more memory channels, too. That’s why is costs a thousand dollars.

      I agree that Bulldozer is somewhat disappointing, but I think your criticism is misdirected.

      • rhysl
      • 8 years ago

      My Corei7 920 @4.2Ghz can beat these cpus ?.. Sigh .. Come on AMD you can do better?..

      Or are you permanently crippled in this competitive market??

        • NeelyCam
        • 8 years ago

        Maybe they should sue Intel for another $1.5bil…? Thit Read guy needs his golden parachute, and they can’t rely on BD sales to make enough dough for that.

    • kmieciu
    • 8 years ago

    Well all that windows-compiled-for-lowest-common-denominator benchmarks tells me nothing. Even Phoronix ubuntu-compiled-for-lowest-common-denominator are not satysfying. Is there somebody who does gentoo-ricer-insane-cflags-optimized benchmarks?

      • just brew it!
      • 8 years ago

      That’s only going to be a big win if gcc already implements Bulldozer-specific optimizations…

    • Jigar
    • 8 years ago

    I personally feel Bulldozer belonged to 22nm process.

      • smilingcrow
      • 8 years ago

      I feel it belonged on a drawing board which is seemingly how it was designed unfortunately.

      • FuturePastNow
      • 8 years ago

      Considering the benefits 32nm had for the K10.5 cores in Llano, they should have just shrunk the X6 and either boosted its clock speed or added two more cores…

      …and the result would have run rings around the FX-8150. No wonder they fired Dirk.

        • Deanjo
        • 8 years ago

        I was thinking the same thing. I would love to see the benches on a x6 on a smaller process and have them tweak the it a bit with support for the newer instruction sets and maybe a bit larger cache.

          • khands
          • 8 years ago

          Hell, they may have been able to launch an x8 on the smaller process 🙁

            • OneArmedScissor
            • 8 years ago

            They already had X12s for servers. And what would you do with it? :p

            There’s the real problem with this iteration of Bulldozer.

            • Deanjo
            • 8 years ago

            There are plenty of ways for some of use to utilize all cores. Not everyone is a gamer.

            • OneArmedScissor
            • 8 years ago

            For a PC, really? I don’t play PC games at all. I probably adopted multi-core CPUs a lot quicker than most people here because I needed them to work. I remember being very annoyed that I could push even a Core 2 Quad into locking up on me, and there was a time when I figured I just needed even more CPU cores.

            And then came all the quad-cores with higher clock speeds, integrated memory controllers, significantly faster RAM, and several times the cache. The bottlenecks and compromises have been eliminated, which were the real problem.

            Now even dual-cores have these things, along with hyper threading. I haven’t seen a PC slow me down in what seems like forever. Very cheap CPUs outdo very expensive CPUs of not too long ago because they aren’t held back in the same ways, despite their lack of cores.

            For server and workstation CPUs, the only places you’re going to find tasks that can truly scale to more than a few threads, clock speed compromises and memory bandwidth limitations are still a greater issue than the amount of cores. Intel and AMD are both piling onto the memory system at a much faster rate than they’re adding cores, and that’s what is eating into power limitations.

            But even those things have scalability limitations. With Sandy Bridge EX, Intel is actually reducing the amount of cores vs. Westmere EX, and increasing clock speeds. What does that tell you?

            I think you’re going to have a lot of trouble pointing out something people do with PCs that wouldn’t be better served by higher clock speeds and lower memory latency. If more cores is beneficial to you, then why aren’t you using such a CPU already? Because it would only be worse, due to the inability to scale to that many cores, and the huge deficit in clock speed.

            The only time constrained thing I could think of for PCs is video encoding, and there’s QuickSync for that now, which will only get better.

            Specialized approaches are immeasurably more beneficial than piling on cores.

    • Meadows
    • 8 years ago

    Disappointing, but not entirely unexpected.

    • Arclight
    • 8 years ago

    The fact that this chips have 2 billion transistors and yet perform the same as the old generation (in some tests the 1100 T even bests the new generation) tells me that something is rotten in the company named AMD. Heads need to fall, this is an embarasment.

      • NeelyCam
      • 8 years ago

      I hope the [i<]right[/i<] heads will fall... scapegoats are too easy to find.

        • ermo
        • 8 years ago

        You have to wonder if Dirk was a scapegoat or if the board was indeed right in firing him.

    • TravelMug
    • 8 years ago

    Epic FAIL!

    It’s not even up for a debate. Fail after fail in the benchmarks and when there would be a glimpse of hope it’s a totally irrelevant win. Then come the power consumption results as icing on the cake.

    • madseven7
    • 8 years ago

    This must be a joke!! It can’t even keep up with the x6 1100T. Wow..especially having faster clocks!!! And they priced this between a 2500k and 2600k. Are they sick!! Consumes more power as well!! I never thought I’d say this but AMD should just go bankrupt! This is a huge disappointment. I was an AMD fan until now. They are not competitive and never will be until the get rid of the board! They squander their investors money when they could use it to be competitive..good-bye AMD

      • kristi_johnny
      • 8 years ago

      Hope AMD don’t go bankrupt, it be desastrous to leave Intel alone on the market

        • JustAnEngineer
        • 8 years ago

        What, you don’t want to go back to the days when Intel consumer desktop CPUs cost over $1000 and the server parts were over $3K?

          • sweatshopking
          • 8 years ago

          I really doubt that amd is the cause of the cpu drop. they’re given FAR too much credit for the minuscule impact they have on intels prices.

            • wierdo
            • 8 years ago

            Looking back at the time when the K7 came out, I would disagree.

            Off the top of my head, Intel’s plans at the time were to bump clockspeeds from around 450mhz to something like 600mhz. And their prices for their 450mhz CPU were in the several hundreds of dollars.

            Then K7 came out, and couple years later, speeds hit 1Ghz, and I bought a 650mhz CPU for around $50 or $60.

            So yeah, I think competition helps based on such experiences. It’s just a shame the current AMD product doesn’t look as exciting as the K7 when it came out, I hope there’s some kinda design kink for the market’s sake, cause I’m not really impressed at the moment.

            • sweatshopking
            • 8 years ago

            “Then K7 came out, and [b<] couple years later, [/b<] speeds hit 1Ghz, and I bought a 650mhz CPU for around $50 or $60." ? so... in a number of years.. you purchased an old cpu for cheap? color me surprised. AMD's market share is low. corporations are going to buy intel regardless of the price. Unless intel was suddenly FAR FAR FAR more expensive. they're quite a bit more now, and they OWN the server and desktop market for business. It would take a long long long time for amd to catch up in market/mind share if they were even twice as fast for the same price.

            • wierdo
            • 8 years ago

            No, I didn’t purchase an old CPU, I purchase a new one, but the prices were a steal after the K7 came out.

            Not everyone buys top of the line CPU, the price/performance is not very good up there obviously. For example I remember the 1Ghz K7 was a ridiculous $1000 because at the time Intel couldn’t compete with that speed, it was just a mojo competition (Jerry Sanders had a funny Gigabucks dollar bill printed with his head on it going around corporate HQ to celebrate and all that PR stuff). Funny thing is that CPU went down to a couple hundred after Intel came out with their competing product eventually.

            Point is the market went from mainstream CPUs selling between $200-800 to a market of $50-$500 in just two years when I was looking to upgrade, and it wasn’t very difficult to notice it at the time at all.

            The part about the corporations I’ll agree with, my observation is about the other half the market, consumer space. Still, sometimes pressure from consumer space can translate into market acceptance in the business space. AMD had a few chances in the past, but they don’t capitalize on it very well unfortunately, not even considering the past Intel/Dell exclusivity arrangements and whatnot.

            • sweatshopking
            • 8 years ago

            no. but my point is that you bought a cpu that was a few years old tech. i can buy an arrandale chip for nothing. saying “when we passed the 1ghz mark, the 650 mhz chips got cheap” is redundant. it’s old tech. i buy cheap too, but to suppose that amd is somehow keeping intel honest is silly. I like AMD. i wish they were executing better, but they’re not.

            • wierdo
            • 8 years ago

            You misunderstand me, when the K7 came out, the market moved so fast that the same processor family spanned speeds from 500mhz all the way up to ~1.2Ghz in that period, it was a ridiculous shift in the market, clockspeeds quadrupled in the span of couple years when it should’ve been just a 200mhz bump (discounting the P4 getting rushed to take over where P3 left things).

            And this has nothing to do with “corporate honesty” you’re alluding to, rather it’s basic Econ 101 from freshman college, supply/demand basics. In a monopoly/duopoly situation (like cellphone service market for example) the business will charge “what the consumer can bear” rather than what can make the most successful sales results vs competition, so for example:

            If you’re the only farm in town you plot a chart of price vs sales, and then pick the sweet spot where you can make the most money with the least amount of production. It’s smart business. Now if you have more than one farm selling milk, then for the same quantity/quality of product the consumer will go for the best deal, and the companies try to undercut eachother to make more sales, this drives prices down faster and the companies make up the low pricing with higher production quantities.

            Anyway, I think/hope that market will regulate itself due to competition from another surprising source, and that’s the move toward computing devices as appliances, like the Arm market that’s eating up the tablet market and starting to get into Netbooks as well, hopefully they’ll move up some more and threaten the x86 market enough to dictate price points positively for the consumers.

            • sweatshopking
            • 8 years ago

            I understand the economics of the situation. I just think you’re overestimating AMD’s impact. now ARM, there is a real threat. THEY will be the ones to keep intel competitive.

            • destroy.all.monsters
            • 8 years ago

            Yeah that will help us so much on the desktop.

            /sarcasm.

            • sweatshopking
            • 8 years ago

            Arm is coming to the desktop. Not today, but reasonably soon. they’re clearly not competitive on the high end, but when it can do 90% of you required processing, and costs 20$, intel will have a hard time selling more powerful cpu’s by the bajillions. is it going to affect people on the high end? no. but when you can buy a 100$ i3, or a 20$ arm to facebook, then why would you spend the extra 80$? You’re right, you wouldn’t. Arm is coming, it’s just a matter of time.

            • destroy.all.monsters
            • 8 years ago

            It is immaterial to games and to what people whose jobs depend on computing horsepower – graphics designers, CAD/CAM, architecture etc. do _on the desktop_. It’s relevancy, if it ever gets there in those markets, is at least 5 years away.

            Don’t forget that ARM was wooing AMD not so long ago. I venture to guess that attaining access to those markets was a part of that.

            • kamikaziechameleon
            • 8 years ago

            AMD does offer some compelling low end solutions but anything they try pushing over 180 dollars has not market relevance ATM.

          • madseven7
          • 8 years ago

          That won’t ever happen now.

          • Bauxite
          • 8 years ago

          This, keep cheering for intel and wishing amd to die but a fully operational [s<]death star[/s<] monopoly is a terrible thing to behold. (aka open your wallet and bend over)

        • madseven7
        • 8 years ago

        The way it looks like right now…Intel is alone..I had such high hope but this really sucks the bag…I’ve built numerous systems all AMD due to value and performance. With that said I can no longer do that. Disappointing is putting it mildly. This is no different than a P4.

          • NeelyCam
          • 8 years ago

          [quote<]I've built numerous systems all AMD due to value and performance. With that said I can no longer do that.[/quote<] Step into the light. Let the warmth of goodness embrace you.

        • madseven7
        • 8 years ago

        What AMD just released, INTEL is the only one in the market. To call a new chip with worse performance than their last chip the future, heads should roll, starting from the top.

      • just brew it!
      • 8 years ago

      While I agree Bulldozer is a pretty major disappointment, you’re over-simplifying. If it was that easy to go head-to-head with Intel in the x86 segment, AMD wouldn’t be the only one still doing it. Several other companies have tried and failed over the years; the fact that AMD is essentially the last man standing amongst Intel’s x86 competition says that they’ve done a lot of things [u<]right[/u<].

        • madseven7
        • 8 years ago

        While I agree with you that they’ve done a lot of things right, they have done more wrong. The board should go! They have let go some of their best leaders and hired some of the worst. Hector RuizThe board didn’t see what he did to Motorola? They have no direction. They have squandered tons of money (payouts to the one mentioned above) and not once have I seen them promote their chips to the extent of Intel.
        Disappointed in not the word. They should be ashamed. If I worked at AMD I would be. How could they call this a Processor. True that Servers make more money than the desktop, but to totally screw your followers/userbase and take a step backwards is totally unacceptable!
        Disappointed I’m not, Pissed off, ashamed and screwed I am!

          • just brew it!
          • 8 years ago

          [quote<]Disappointed in not the word. They should be ashamed. If I worked at AMD I would be. How could they call this a Processor. True that Servers make more money than the desktop, but to totally screw your followers/userbase and take a step backwards is totally unacceptable! Disappointed I'm not, Pissed off, ashamed and screwed I am![/quote<] Wow... just wow. 😮 Get a grip, it's just a CPU chip. Life goes on. I've been using AMD CPUs pretty much exclusively in my own builds since the K6 days. I like the bang-for-the-buck AMD usually provides, and I like to support the underdog. As I've noted elsewhere, I plan to buy a couple of X6 chips for my AM2+/AM3 systems, then sit tight and see what happens over the next year or two. Maybe they'll get their house in order; I hope they do. We need the competition. "Followers"? You make it sound like AMD users are like some sort of fanatical religious cult. (I suppose some are...) They haven't "screwed" anyone, other than maybe their own shareholders. Nobody's forcing you to buy these chips at the launch price. It's not like they've been on the bleeding edge performance-wise lately, and quite a few intelligent people have been predicting that Bulldozer probably wasn't going to put them back in that position either. So the current situation should not be a complete surprise to anyone. What *did* surprise me a little was the uncompetitive pricing, given the performance. And provided their yields don't suck too badly, they can easily remedy this by dropping the price. If the price comes down to where it is commensurate with the chip's performance, I'll consider buying one.

      • shank15217
      • 8 years ago

      AMD will deliver, its a given..

        • madseven7
        • 8 years ago

        Still waiting 10 years later…..soon I’ll be dead by the time AMD becomes competitive.

          • destroy.all.monsters
          • 8 years ago

          Are you sixty two – or is it that you keep raising your blood pressure to insanely high levels due to things you read on the internet?

      • Bensam123
      • 8 years ago

      Sensationalism.

    • sschaem
    • 8 years ago

    If AMD sell the FX-8150 for $189 they have a competitive product. The X6 1100T is $175 right now on newegg

    At $245, I cant find a single workload where it makes sense over a i5-2500k, or even an almost 100$ cheaper 1100T

    I hope AMD break their silence and explain their logic in pricing this chip.

      • destroy.all.monsters
      • 8 years ago

      I expect that price to drop very quickly after the initial shipment. The problem is that it looks like yields are still low and that means it costs more to put out.

      I would expect some new Windows 7 drivers in 4-6 weeks if the board isn’t as braindead as hp’s.

        • sschaem
        • 8 years ago

        The scheduler tweak will help.. (BTW AMD knew this was needed 3 years ago… surprising this wasn’t part of windows7)
        Also this doesn’t seem to bring that much ~5%.

        The issue I see is with AIDA L1, L2, L3, mem, the result are pathetic, specially vs a sandybridge processor.

        Bulldozer should have this refined, yet the number are almost 3 time lower then expected… something feel borked.

    • sschaem
    • 8 years ago

    The biggest disappointment is the task energy. The rest we pretty much knew already.

    BTW, TR. It seem Radeon VS geforce does matter.
    Other review show for example that F1 2011 to be faster on a 8150 vs a 2600K on a Radeon 6950 VS your choice of a 460 GTX.

      • NeelyCam
      • 8 years ago

      I don’t know what’s going on… 32nm SOI should’ve helped.

      Netburst deja vu..

        • sschaem
        • 8 years ago

        I think AMD had to bump up the voltage (over volt) and that proved disastrous to power efficiency.

        The question is, can AMD run bulldozer at high clock rate on GF 32nm process with lower voltage ?

        Maybe TR can do some undervolting test to see how this would affect power consumption with BD ? (if their ‘sample’ undervolt)

        The concern is that from what we can guess, since AMD doesn’t like to talk to its customers, they have been trying for a long time now to lower voltage but cant… so if they haven’t been able to do so in the past 12 month, a quick fix might not happen.

        Best case, AMD get a break thought soon and can lower voltage by ~15% in their next cpu rev.
        This would dramatically reduce power consumption under load.
        They also get the scheduler patch (2 to 10% speedup) and fix their cache issue (unless AMD denies it, Its a fact 🙂 and so increase performance by another 5-15%.

      • willmore
      • 8 years ago

      I was thinking the choice for an nVidia card for an AMD processor was strange. If anyone is going to optimize their drivers for the new AMD chips, it would be AMD.

    • omegajvn1
    • 8 years ago

    Clearly AMD is in a hole right now…and I’m extremely sad, being an AMD fan here. I believe their only saving grace now is to integrate their VASTLY superior graphics architecture into all of their CPU’s just as Intel has. Until that happens, AMD’s silicon just can’t compete with intel with a standalone CPU. There was a reason they purchased ATI all those years ago. They need to exploit this ace in the hole…

    • GokuSS2
    • 8 years ago

    I would like to see what the Folding PPD with -bigadv checked is on this chip?

    • sschaem
    • 8 years ago

    Big fiesta at AMD, they updated their web front page finally after weeks! to Dragon age II..

    Its like bulldozer doesn’t even exist… sigh

    • Archer
    • 8 years ago

    Who would want this in a server? It runs too hot and uses too much juice. If it was designed as a server part first, then they failed even harder in that respect than on the desktop.

      • DancinJack
      • 8 years ago

      It won’t be clocked at 3.6+ in a server.

        • tfp
        • 8 years ago

        Why not IBM can do things like that with it’s POWER chips without issue

          • TravelMug
          • 8 years ago

          Different markets…

            • tfp
            • 8 years ago

            How’s that?

            “With the six-core Power7 chip running at 3.7 GHz, the Power 710 Express has 16 GB of memory and those two disks at a cost of $8,120.”

            [url<]http://www.theregister.co.uk/2010/08/17/ibm_power7_low_high/[/url<] This looks to be in the same price range as x86-64 servers and this is over a year old. The only reason AMD won't clock at 3.6+ in a server is if they can't get the chip stable enough. Intel was running huge clocks with P4s in servers back in the day AMD hasn't hit some magical power/heat wall with these chips.

            • TravelMug
            • 8 years ago

            What do you mean “How’s that?” Single vendor hardware/software/services market vs, commodity x86 market. Nobody will buy the product if another one will do the same work with half the power consumption. Heck, nobody is doing it already today, hence the market share of Opterons in the server markets.

            • tfp
            • 8 years ago

            Ah ok you’d like to pretend that x86 is the only U2 server in town. I see your point then.

            • TravelMug
            • 8 years ago

            Maybe this will be hard to comprehend for you but for AMD yes, x86 is the only U2 server in town. And when there are competing products offering better price/performance or performance/watt then it’s totally irrelevant what IBM does with POWER in it’s own market.

            BTW, the market is not defined by 1U, 2U, 4U etc. size of the box. It’s defined by what is inside that box and what is it licensed for. The non-x86 IBM and Oracle products bear no relevance here. Especially if in case of overlap of these lines with x86 (for example SAP) the current Opteron and from the looks of it the upcoming Bulldozer based range is trounced by Xeon.

      • NeelyCam
      • 8 years ago

      [quote<]Who would want this in a server?[/quote<] [url<]http://www.ornl.gov/[/url<]

    • ronch
    • 8 years ago

    Big excitement followed by ‘Oh well.’

    I was hoping, not expecting, AMD to get back on top, but creating something as complex as a microprocessor is one of the most difficult engineering challenges. So, I still give my kudos to Mike Butler and his team for finally finishing Bulldozer.

      • Krogoth
      • 8 years ago

      Bulldozer was never meant to be a desktop chip. It was quite evident from the day that first details about Bulldozer’s underlying design that it was going to be a workstation and server chip.

      TR’s benchmark suite makes it painfully apparent.

        • shank15217
        • 8 years ago

        If this was a server chip then AMD should have released the server chip first. It fundamentally changes the viewpoint if AMD releases the sever refresh and shows the performance improvement of this platform. Instead they released the desktop variant that shows the weaknesses of the architecture, how is that ever a good thing.

          • Krogoth
          • 8 years ago

          There are server versions already coming out very soon.

          This is purely a marketing move to refresh the existing Phenom lines, while at the same time help pay off the massive R&D to develop the architecture.

          • Chrispy_
          • 8 years ago

          The server chip [b<]was[/b<] released first: Interlagos was shipped almost two months ago.

          • ronch
          • 8 years ago

          Agreed. Ironically I believe the server parts shipped out first but the desktop parts got reviewed first. Crazy AMD marketers. Off with their heads!

        • axeman
        • 8 years ago

        Yeah, I bet with a server workload like some sort of database benchmark, the fact that there is only one floating point per every two integer isn’t going to matter. In stuff where you’d want a “real” 8 core chip like content creation, though, it’s disappointing to see a Phenom X6 best it.

    • Jigar
    • 8 years ago

    Q6600 to 2600K here i come…

      • Krogoth
      • 8 years ago

      I would personally wait for Ivy Bridge. The Q6600 is no slouch and 2600K is only significantly faster at workstation related workloads. I suppose that you also want lower power consumption. The 2600K consumes a fraction of the Q6600’s power when fully loaded.

        • Jigar
        • 8 years ago

        hmm, you have a point. Ivy is suppose to launch in march right ?

          • NeelyCam
          • 8 years ago

          Thereabouts.

          • derFunkenstein
          • 8 years ago

          Do you really want to hold on another 6 months? I mean, IMO, the Q6600 is long in the tooth unless you’re overclocking it. AMD can actually beat it, for one, and can do so at less than $100, for two. The difference between a 2600K and a Q6600 is already ridiculous.

            • Krogoth
            • 8 years ago

            For general computing and gaming. The Q6600 can still effortlessly handle anything you can throw at as well as the existing Phenom X4s. Games are typically limited by the GPU not the CPU with a few exceptions here and there.

            2600K is pointless from a performance standpoint unless you are doing workstation related tasks.

            The only reason you would want upgrade to a 2600K is if you want lower power consumption.

            It perfectly illustrates why CPU performance in the desktop arena doesn’t matter that much anymore. There is simply no killer apps that brings dual-core chips, let alone quad-core chips to their knees.

            It is part of the reason why Bulldozer’s current strategy on parallelism falls short for desktop users. It is also why Gulftown and upcoming Sandy Bridge-Es are poor values for desktop users. They only shine under heavily-threaded workloads which almost never happen under a gaming/desktop system.

            • flip-mode
            • 8 years ago

            However, any Sandy Bridge is going to have significantly better single threaded performance than a Q6600. That may make the upgrade worthwhile. Otherwise, I generally agree that Joe and Jane Six don’t need more than a Q6600 or even an E6600, but Jigar may be more than a Joe / Jane Six.

            • swampfox
            • 8 years ago

            My E7300 frequently runs both cores in the 80%+ area while playing Fallout 3 at 1920×1200. Probably wouldn’t be an issue if I had a quad-core though. You can definitely max out dual core CPUs these days.

            • derFunkenstein
            • 8 years ago

            That’ll still be true of IB when it launches, too. I don’t get what you’re saying; that the Q6600 is good enough and he should just sit on it? Or that he should upgrade on your schedule?

            • OneArmedScissor
            • 8 years ago

            To be fair, it’s not $100 when you need a new motherboard and RAM, and I don’t think I’d buy any AMD board right now. Socket FM2 is too close, which actually has a legitimate future.

            You could argue that socket 1155 has an “upgrade path,” but it’s about as inconsequential as AM3’s.

            It’s either buy Sandy Bridge or wait for something else.

          • MrDigi
          • 8 years ago

          Some delay is expected. Intel still needs IB to replace desktop SB for DX11 and faster graphics. CPU performance could be held back based on no competition and keep it under SB-E offerings. The bigger opportunity is IB mobile chips, for power consumption and graphics improvements.

        • shank15217
        • 8 years ago

        Last I checked the Q6600 is very much a slouch, its like calling 8800GTX a great card, yea it was a good deal… 4 years ago.. The difference between Q6600 and SB2600k is around 2-3x the performance, what else did you want?

          • Krogoth
          • 8 years ago

          It is twice as fast in the best case in workstation and server related task which can fully harness the power of the 2600K. Otherwise, the 2600K isn’t really that much faster in most other tasks. The gap comes mostly from its clockspeed advantage (2.4Ghz versus 3.4Ghz). It is less pronounced when most Q6600s out there are OC’ed to 3.0-3.2Ghz with little or no effort.

          The main benefit of getting Sandy Bridge is its far superior power consumption = you don’t need a large, loud HSF to keep it cool.

      • canmnanone
      • 8 years ago

      hey what was the highest youve oc’ed your q6600? mine was 3.6 stable 24/7 at 1.750 volts.

        • Jigar
        • 8 years ago

        3.7 GHZ stable at 1.45 V

        • sschaem
        • 8 years ago

        I’m going to hold on that q6600 (2.4ghz@1.152v turbo 3.2ghz@1.24v) and my amazingly cool running overclocked/undervolted 1090t for a while longer.

        I will wait for LGA2011 with a 22nm IV-E PCIe3 for my q6600 replacement. But I will pickup an fx-8150 as soon as the price drop below $200

    • Krogoth
    • 8 years ago

    Barcelona called, it wants its disappointing performance at launch back.

    Anyway, Bulldozer seems to be suited as a server chip. I am curious to see how well it fares in the server arena.

    I didn’t expect Bulldozer to be a winner in the desktop market, but it doesn’t really matter that much anymore. The demand for faster CPUs isn’t that big anymore. The new game is power consumption and miniaturization. Intel has the ball in that court.

      • Anonymous Coward
      • 8 years ago

      This is a [b<]hell of a lot[/b<] like the original Phenom launch. Why does AMD set expectations high when they know the product is essentially equivalent to what they already have?

        • destroy.all.monsters
        • 8 years ago

        Public relations and reality have a tenuous connection at best. That said, all of these processors including the 1100T are still way more than what the average consumer will ever really need. Even for gaming – unless you’re going for absolute max everything – they’re fine.

        You can’t expect anyone’s PR department to go – “yeah, barely better than last generation and our competitor kicks ass on our parts in a lot of areas”. That’s suicide. I get that you’re disappointed – I am too.

        We can hope that drivers make it better – but ultimately this is what AMD has until Piledriver. You can either feed the beast that is intel – and continue to hurt AMD’s chances – or buy AMD, or wait.

          • madseven7
          • 8 years ago

          Feed the beast. Why would I have to wait for new drivers? Don’t you think all this should have been done already prior to the release of this chip? That’s just an excuse to say that this chip sucks period.

        • Krogoth
        • 8 years ago

        FYI, Barcelona = Phenom I 😉

      • dragosmp
      • 8 years ago

      Looks like Intel has the ball in all courts

        • Krogoth
        • 8 years ago

        Save for one, GPUs. 😉

    • SecretMaster
    • 8 years ago

    I just started reading the article but I’m surprised; given Tech-Report’s pop-culture savvy when it comes to writing articles, why didn’t Scott start the article with Tik-Tok 😛

      • DancinJack
      • 8 years ago

      They did use The Big Bang Theory for the video stuff at least.

    • VILLAIN_xx
    • 8 years ago

    I just held out long enough to see what kind of price/performance ratio would be after TR’s review.

    So, whose going with Blue Team after this review?

      • Sencapri
      • 8 years ago

      I’m going to wait for Ivy bridge only because I can 😀

        • VILLAIN_xx
        • 8 years ago

        Yeah, might as well. Whats after Ivy Bridge though? Haha, by then ill be saying “Oh lets wait for piledriver”…

          • UberGerbil
          • 8 years ago

          Haswell, and several people have mentioned waiting for it already.

            • OneArmedScissor
            • 8 years ago

            It’s quite possibly the only CPU that’s had a significant emphasis on improving the way PCs can be used since Core 2. May as well wait. All we’re getting before then is Sandy Bridge 1.5 and Bulldozer 1.5, regardless of what they may paint them up to be.

            • NeelyCam
            • 8 years ago

            +1. Ivy Bridge probably brings some evolutionary improvements (lower power, improved graphics). Haswell, being a new architecture, could bring in something revolutionary. The 10-day standby battery life on an ultrabook sounds pretty good to me.

          • bcronce
          • 8 years ago

          Or he’s someone like me. I have an i7-920. While the CPU is plenty powerful, the motherboard is crap, and I want/need USB3/SATA3/etc.

          I do not want to purchase a motherboard for my 3 year old chip, so that means I need to purchase an up-to-date motherboard for an up-to-date CPU. But why would I want to purchase a current mobo when some new PCIe3/etc ones will be debuting next year. I’ll just wait for the Ivy Bridge.

          It feels like a such a waste because my i7-920 is plenty fast, but I really need the newer interfaces, and the fact that it’s a Dell mobo from a $800 PC from 3 years ago…… yeah…. crap…

            • elmopuddy
            • 8 years ago

            I’m in same boat, my i7-875K is plenty fast, but I want real SATA 3, so 350$+ raid card, or new CPU and MB.

      • destroy.all.monsters
      • 8 years ago

      Not I. We can’t afford a 1 vendor market in the cpu space.

        • Farting Bob
        • 8 years ago

        So you are going to go with a more expensive, power hungry, slower CPU? I dont think the profit they will get out of you (if any, those chips are HUGE) will make 1 bit of difference, but it will certainly suck for you having a very inferior chip.

          • khands
          • 8 years ago

          Well, if there are more people like him they combined might be able to keep AMD afloat for another month or two…

          • OneArmedScissor
          • 8 years ago

          You could also just buy nothing. If you bought a new CPU a few years ago, none of these are really going to do much for you, if anything.

            • Bauxite
            • 8 years ago

            Spend on new GPU in the next 6 months or so, keep overclocked i7 9×0 from 2009 awhile longer: I won’t be alone.

          • destroy.all.monsters
          • 8 years ago

          The difference will ultimately be negligible for me and for most. I’m not driving an immense screen, not doing uber sli/crossfire.

          A better question is how will you feel when there’s no other competitor and intel is charging thousands per cpu.

      • l33t-g4m3r
      • 8 years ago

      Guess my next upgrade will be intel, although it’ll be a while before I need to replace my 1090T.

      • ronch
      • 8 years ago

      Intel must be having a good day today.

        • NeelyCam
        • 8 years ago

        One good thing that could come out of this is that, with Intel “winning” the desktop race, they’ll start focusing on super low power chips to compete with ARM. IvyBridge and Haswell laptop chips, cell phone chips…

      • Anonymous Coward
      • 8 years ago

      This didn’t really make AMD’s situation worse.

      Except that apparently they could have saved themselves the trouble of paying a lot of engineers for three years. Hell, they could have just about beat this with a mildly modified A64 (that is, 32nm and sporting relatively small tweaks).

        • ermo
        • 8 years ago

        Except that they already tried this with Llano. Look what that got them.

          • Anonymous Coward
          • 8 years ago

          What are you talking about? Llano is fine except for production volumes.

            • ermo
            • 8 years ago

            According to the rumor mill, something doesn’t gel between the 32nm process and the way Llano was shrinked/tweaked. The fastest quad core 32nm part is at 2.9GHz. The fastest comparable 45nm part is at 3.7 GHz.

            So based on what I’ve seen, I remain unconvinced that the approach you advocate (basically a X6 Llano with cache instead of GPU) would’ve been a better bet than BD, which at least has future evelopment potential, whereas the 32nm Llano is pretty much the end of the line in terms of development potential for that particular microarch.

            • Anonymous Coward
            • 8 years ago

            I doubt AMD invested too much effort in “porting” K10 to 32nm. They apparently plan on having it on the market for less than a year before the successor arrives! Probably there are also general problems with 32nm for AMD right now, hence the supply shortages and delays.

            Also recall how 90nm K8 (at 3.2ghz) beat 65nm K8 at clockspeed for a very very long time. They didn’t need to reinvent the processor to get a 3.7ghz quad on 45nm.

      • ronch
      • 8 years ago

      It must be a sad day at AMD today. I’m not gonna kick them when they’re already down. Instead, I’m gonna cheer them on and probably still proudly go get an FX next year when it matures and more widely available in quantity.

        • madseven7
        • 8 years ago

        What, you should kick them. Maybe that’s what it’ll take to get the company going. Get rid of the moron’s on the board and get some true leadership and engineers. Get them back to designing good chips and not overpriced crap.

          • destroy.all.monsters
          • 8 years ago

          The market is kicking them just fine. You can see from other places (hp being one, yahoo) that insane boards are allowed to rule for far too long and that CEOs are picked for specious reasons.

          You have a plan? Let’s hear it!

            • madseven7
            • 8 years ago

            Bring in ICANN

            • destroy.all.monsters
            • 8 years ago

            Yeah, that’ll sure help.

      • srg86
      • 8 years ago

      Well, once real life stuff allows, this certainly isn’t going to change my plans for my next rig to be Sandy or Ivy Bridge. Especially as where reviews have code compilation, BD is pretty poor even compared to Phenom II.

      • can-a-tuna
      • 8 years ago

      Not me. Everything starting with “i” is crap.

    • shaq_mobile
    • 8 years ago

    ferst

    wow thats almost as bad as the toms hardware review. im pretty bummed.

    sorry, let me clarify. the results are disappointing. the quality of the article is excellent as usual. (thanks ssk)

      • sweatshopking
      • 8 years ago

      you’re referring to the scores, not the quality of the review, correct?

        • shaq_mobile
        • 8 years ago

        i dont read toms articles anymore, i just glanced at the numbers. their site format has improved but not enough for me to spend time there.

      • clocks
      • 8 years ago

      As an Intel shareholder, I am pleased with these results. As a computer junkie, I am saddened to see AMD flounder, as I believe the industry needs more than a Intel monopoly to keep innovation alive.

        • shaq_mobile
        • 8 years ago

        i always appreciate the little guy. makes me sad to see this cpu sell for so much and deliver so little on the gaming front.

      • shank15217
      • 8 years ago

      Dont troll toms hardware, their review was accurate just like this one.

        • NeelyCam
        • 8 years ago

        He meant BD scores were almost as bad as in Tom’s Hardware review.

Pin It on Pinterest

Share This