Intel’s Core i7-3960X processor


Truly high-end desktop PCs have been in a curious position for nearly a year now. Intel’s X58 platform and the corresponding Core i7-900-series processors have reigned supreme in their segment for three successive years, with little real competition from AMD. Yet in the 11 months since the debut of Intel’s Sandy Bridge processors, this high-end platform hasn’t been unequivocally faster than the more affordable mid-range options. The source of the trouble is pretty simple: newer technology. The Core i7-900 series is based on the older Nehalem CPU microarchitecture, and the newer Sandy Bridge architecture delivers substantially more performance in every clock cycle. Even the thousand-dollar Core i7-990X, with six cores and three memory channels, can’t distance itself too much from its quad-core Sandy Bridge cousin, the Core i7-2600K.

Thus, the X58’s raison d’etre has largely been reduced to its utility as a vehicle for multi-GPU solutions. Since quad-core Sandy Bridge chips can only support dual eight-lane PCIe connections, the X58’s more ample PCIe connectivity has been prized in certain circles. We can’t say we’ve been part of those circles, though, since we’ve never been able to measure any real performance differences between SLI and CrossFire on an X58 and a Sandy Bridge rig. Heck, we even quit recommending the X58 as the basis for the beastly Double-Stuff Workstation in our famous system guides, subbing in a Core i7-2600K instead.

In short, Intel’s high-end platform has been in desperate need of a refresh. Fortunately, the time for that update has finally come, and Intel is looking to remove all doubt about who’s the top dog, unleashing a massive new chip—known as Sandy Bridge-E, or Sandy Bridge Extreme—that doubles up on nearly every resource included in the quad-core versions of Sandy Bridge. If that prospect isn’t enough to get you salivating, you probably don’t associate a hunger-type response with computer performance, which is entirely normal. Still, some watering of the mouth in this case would help me build drama for the technical specs I’m about to catalog, so oblige me, if you would.

The chip

Sandy Bridge-E is one formidable hunk of silicon. Natively, it includes a total of eight CPU cores capable of tracking 16 threads via Hyper-Threading, 20MB of L3 cache, quad channels of DDR3 memory, and 40 lanes of next-gen PCI Express I/O connectivity. If you’re like me, you’re probably thinking those specifications sound like they belong to a multi-socket system, but Sandy-E crams all of that stuff into a single socket. While that fact sinks in, we’ll consider an annotated picture of the chip’s die, highlighting what’s where.


Sandy Bridge-E die layout. Source: Intel.

There’s much to consider, and we should start with the chip’s eight… er, six CPU cores. (I didn’t tell you wrong; Sandy Bridge-E natively has eight cores, but Intel has elected to disable a couple of those cores in the first products based on this silicon. We’ll discuss that shortly.) Each of those cores is based on the Sandy Bridge microarchitecture, which is essentially a total overhaul of Intel’s fundamental CPU building block. We’ve detailed the more notable changes in the Sandy Bridge architecture elsewhere, but the highlights are worth calling out, including the improved branch predictor, a cache for decoded micro-ops, and the ability to execute two 128-bits loads from memory per clock cycle. All of these changes contribute to higher per-clock instruction throughput. Also, graphics and media workloads should benefit from the inclusion of the AVX extensions to the traditional x86 and SSE instruction sets. AVX doubles the width of floating-point vectors to 256 bits, and Sandy Bridge can execute both a 256-bit add and a 256-bit multiply in one tick of the clock, if the stars (and the data) align correctly.

In order to enable such considerable throughput, Sandy Bridge chips incorporate a ring-style interconnect between the cores. Intel has said each ring stop can transfer up to 96 GB/s at a clock frequency of 3GHz. If that holds true for Sandy-E, with eight stops, the total internal capacity of the ring should be a staggering 768 GB/s.

Not only is the on-chip communication capacity tremendous, but Sandy Bridge-E talks to the rest of the system at unprecedented rates, as well, starting with the quad channels of DDR3 memory provided by the on-chip memory controller. The Core i7-900 series has “only” three memory channels and has never officially supported memory speeds above 1066 MT/s. With Sandy-E, DDR3 transfer rates up to 1600 MT/s are officially supported (though only with a single DIMM per channel). That works out to 51.2 GB/s of memory bandwidth across all four channels—again, a staggering rate, more than double the throughput of lower-end Sandy Bridge desktop chips. What’s more, the first chips and motherboards expose the correct multipliers for a range of higher memory speeds, including 1866, 2133, and 2400 MT/s.

Most Sandy-E mobos will be outfitted with eight DIMM slots, so the memory capacities are considerable, too. The manual for Asus’ board cites a max capacity of 64GB, while MSI’s claims 128GB is possible. These outsized numbers point to Sandy Bridge-E’s heritage as a single-socket variant of Intel’s upcoming 2P server product, Sandy Bridge-EP. Just as the X58 platform was a sawed-in-half version of the Nehalem-EP platform, so Sandy Bridge-E is one side of a future Xeon configuration. Those Xeons aren’t slated to be available for a while yet, though, and Intel had to make a few concessions in order to bring Sandy Bridge-E to the desktop on schedule.

One of those concessions involves the 40 lanes of PCI Express connectivity built into the processor. The incorporation of PCIe directly onto the CPU is a topology change; the X58 chipset played host to the PCIe lanes in the prior-gen platform. Moving this I/O onto the processor die ought to reduce latency and has the potential, at last, to allow superior multi-GPU performance in Intel’s high-end platform. However, you’ll notice that Intel claims Sandy-Bridge-E is capable of supporting “PCI Express 2.0 graphics,” in spite of the fact that this chip was expected to support the next-generation PCI Express 3.0 standard. The reality is that Sandy Bridge-E and its supporting motherboards should technically meet the requirements of the PCIe 3.0 specification. However, the rest of the world doesn’t yet have enough PCIe 3.0-capable devices ready to roll. Without access to GPUs and other chips with PCIe 3.0 connectivity to test against, Intel isn’t ready to make robust claims about this platform’s PCIe 3.0 support right now.

We should see PCIe 3.0-capable graphics processors hitting the market late this year or early next, based on the latest scuttlebutt. Once that happens, we think the likelihood is that most or all PCIe 3.0-capable GPUs will be able to achieve PCIe 3.0 transfer rates when plugged into a Sandy Bridge-E system. Of course, it’s also possible there could be an interoperability snag that would prevent that from happening, which is the reason for Intel’s caution in making claims about this product’s feature set.

If it works out well, the transition to PCIe 3.0 transfer rates should be a very positive outcome, at least on paper, because PCIe 3.0 has essentially twice the bandwidth of second-gen PCIe. The third-gen standard supports higher transaction rates, 8 GT/s versus 5 GT/s, and employs more efficient encoding to take it to double the bandwidth. PCIe 2.0 uses 8-bit/10-bit encoding, with 20% overhead, and thus transfers data at 4 Gbps. PCIe 3.0, meanwhile, uses a 128-bit/130-bit encoding scheme, with much lower overhead, resulting in a transfer rate that rounds out to 8 Gbps. Like much of the rest of Sandy Bridge-E, the math on this PCIe 3.0 connectivity adds up fast. Each third-gen PCIe lane can transmit 1 GB/s of data in each direction, so a PCIe x16 link can transfer a total of 32 GB/s, and the 40 lanes of Sandy-E have a total capacity potential of 80 GB/s.

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
Gulftown Core i7-970, 990X 6 12 12 MB 32 1168 248
Sandy Bridge Core i5, i7 4 8 8 MB 32 995 216
Sandy Bridge-E Core-i7-39xx 8 16 20 MB 32 2270 435
Deneb Phenom II 4 4 6 MB 45 758 258
Thuban Phenom II X6 6 6 6 MB 45 904 346
Orochi/Zambezi FX 8 8 8MB 32 1200 315

As you might expect, cramming all of this processing power, cache, and I/O into a single chip isn’t without consequences. Sandy Bridge-E is fairly enormous, with 2.27 billion transistors in a die area that makes even AMD’s Bulldozer look dainty. Since it will be sold exclusively as a high-end desktop, workstation, and server processor, though, the sheer size probably won’t cut into Intel’s bottom line too much.

The products: the Core i7-3000 series


Source: Intel.

Intel has plans for a trio of CPUs in the Core i7-3000 series, two of which are hitting the market today. The one we have for review is the top dawg, the Core i7-3960X, with the beefiest specs in the table above and a hefty price tag of $990 to match. Like other Extreme products from Intel, the 3960X has unlocked multipliers to ease overclocking.

As we’ve noted, the 3960X has portions of the die disabled, namely two of the CPU cores and 5MB of L3 cache. Releasing a top-of-the-line desktop chip with portions disabled runs contrary to our expectations, but we think Intel had some cogent reasons for making this choice. That is, in order to reach the same 3.9GHz peak Turbo Boost frequency as the top Sandy Bridge quad-core product while staying within the 130W power limit, a couple of cores had to be sacrificed. For a desktop-oriented product where performance in lightly threaded applications remains very important, we can see the logic of this tradeoff. After all, it wouldn’t do for the 3960X to be measurably slower than the Core i7-2700K in a number of applications.

The flip side is that it’s difficult to imagine the typical person who’s willing to shell out nearly a grand for a high-end processor being happy in the knowledge that substantial portions of his new chip have been disabled. We also can’t help but notice that the 3960X is nine bucks short of Intel’s usual $999 price tag for an Extreme Edition product. A true eight-core part would sure slot in neatly at $999 at some point down the line, don’t you think?

Since the table above doesn’t show it, we should relay the exact Turbo Boost behavior you can expect from the 3960X. The top frequency of 3.9GHz can be reached when only one or two cores are active. With three to four cores active, the chip drops down to 3.7GHz, and with five to six cores active, the clock dips to 3.6GHz. Like most Sandy Bridge-based CPUs with Turbo Boost, we’ve not seen the 3960X spend any time at its 3.3GHz base frequency. When none of the cores are busy, the processor settles at 1.2GHz via SpeedStep or the C1E halt function.

Of course, the 3960X is practically superfluous given the specs and price of the next model down the line, the $555 Core i7-3930K. The 3930K gives up 100MHz in base and peak clock speeds and has 3MB less L3 cache than the 3960X, but as a K-series part, it also is unlocked and ready to overclock. Pretty much no one should pay the premium for the 3960X when the 3930K will sell for $555, practically a bargain in the rarefied air of Extreme-class processors. We had hopes of reviewing the 3930K processor today, but Intel wasn’t providing advance samples of this product to reviewers. We’ll have to snag one later.

If you really like the idea of the Sandy Bridge-E platform but aren’t ready to pony up the big bucks as the price of entry, Intel may have a more appealing option for you in the first quarter of 2012: the Core i7-3820. With four cores and a 3.9GHz Turbo peak, the 3820 looks poised to follow in the footsteps of past enthusiast favorites like the Core i7-920, which offered reasonably priced access to the X58 platform a few years back. We don’t yet know the exact pricing of the 3820, but we’d expect it to slot between the options above and below it in the table, hopefully somewhere south of the $400 mark. Intriguingly, Intel plans to produce a native quad-core variant of Sandy Bridge-E, so the 3820 may not just be a larger chip with portions disabled.

The final product we’ll call out from the table above is the Core i7-2700K, a non-Extreme Sandy Bridge part that recently took its place at the top of Intel’s mid-range lineup. The 2700K is only 100MHz faster than the previous champ in that segment, the 2600K, so it’s nothing to write home about. Still, we were intrigued about how it would stack up against the 3960X, so we’ve included performance results for it in the following pages.

A new chipset: the X79 Express


Source: Intel.

Although Sandy Bridge-E integrates a boatload of high-bandwidth connectivity, it still needs a support I/O chip to handle various lower-speed I/O types. That chip is known as the X79 Express. As you may have noticed while peering at the block diagram above, the X79 looks very much like the P67 support chip used in the mid-range Sandy Bridge platform.

Most of what you’d want in a core-logic chipset is included, with only a couple of exceptions. For one, the USB ports don’t support the new SuperSpeed or 3.0 standard; motherboard makers will have to add that feature via a third-party USB controller chip. Also, only two of the six SATA ports can achieve 6Gbps transfer rates; the rest top out at 3Gbps. That may not seem like a horrible limitation given that only SSDs can really make full use of 6Gbps interfaces; putting a mechanical drive on a 6Gbps interface won’t get you much. Still, the X79’s collection of six SATA ports may seem kind of skimpy given the other specifications of this platform. In fact, Intel built another SAS/SATA disk controller into the X79, along with an additional PCIe x4 interconnect to the CPU, in order to provide more fast disk I/O capability. Unfortunately, those two features had to be disabled in this version of the X79 for “silicon health reasons.” We expect a future revision of this chip will restore that functionality, probably when it’s deployed in dual-socket servers.

Another casualty of the X79’s apparent rush to market is the Smart Response scheme that lets a small, fast solid-state disk act as a cache for a larger mechanical hard drive. This feature debuted in the Z68 chipset for Sandy Bridge processors, but the X79 ships without it. Intel intends to add Smart Response caching to the X79 via a future driver update, but its timeframe for doing so is quite a ways off: by mid-2012.

A new and rather large socket

Of course, with all the changes to platform plumbing, Sandy Bridge-E will drop into a new socket, pictured above. With a land grid array-style layout and a cheeky 2011 total pins, the new socket type is known as LGA2011. Its size may not be evident from the image above, but have a look at these pictures.


From left to right: LGA1155, LGA2011, and LGA1366

Sandy Bridge-E is a big chip, and it comes in a big package—substantially larger than the LGA1366 package used for the Core i7-900 series and a heckuva lot larger than its Sandy Bridge sister. Then again, it looks like it needs to be that large in order to accommodate its outsized pin count. I remember former AMD CTO Fred Weber once saying that pin count is a pretty good indicator of a chip’s I/O capacity. If so, well, this one wins.

This socket uses a funky dual-lever retention mechanism that take a little getting used to. You have to move two levers in a specific sequence to open or close the CPU retention mechanism. Seems like a bit of a pain to me, but Geoff, who is reviewing X79 motherboards for us, says he likes it.

Yep, Intel is slapping its brand name on an Asetek liquid cooler, much like AMD is doing for its FX processors. In this case, the Intel-branded cooler will be sold separately from the CPU for between $85 and $100. Intel says the cooler is compatible with both Sandy Bridge/Ivy Bridge systems and Sandy Bridge-E. Curiously, it also says the cooler will work with Ivy Bridge-E, the presumptive successor to today’s new chip. Since Intel mentioned it, we pressed on this front, inquiring whether Ivy Bridge-E will be compatible with LGA2011 motherboards and when it might hit the market. Unfortunately, Intel wasn’t ready to talk about it, leaving us with nothing more than a sense that, hey, the coolers should be compatible.

The cooler itself is quite nice, with a four-pin PWM fan rated for 21 dBA at 800 RPM and 35 dBA at 2200 RPM, which is pretty quiet as these things go. In our experience, it’s an acoustically unobtrusive and effective way to cool one of these CPUs while adding a little bit of thermal headroom for overclocking. Of course, there will be a fairly broad selection of third-party air and liquid coolers for this new platform, as well. If you’ve already blown your budget on the CPU itself, you may be pleased to learn Intel will also be selling a basic air cooler for under 20 bucks.

Also pictured above is Asus’ P9X79 Deluxe motherboard, which we originally intended to use for our performance testing. Unfortunately, issues with this board’s Turbo Boost behavior caused our performance results for the Core i7-3960X to be somewhat inflated. We’re hoping Asus can fix this problem via an update to the mobo’s firmware, but given limited time, we chose to switch over to Intel’s DX79SI Extreme motherboard instead and re-test. I believe Geoff will be addressing Asus’ peculiar choices regarding Turbo Boost policies in his X79 mobo roundup.

Finally, the RAM we used is a four-module kit from Corsair that’s expressly tailored for Sandy Bridge Extreme. These 4GB DIMMs are rated for 1866 MT/s operation at 1.5V.

Now, let’s move on to our test results…

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 AMD
A8-3800 2.4GHz

AMD
A8-3850 2.9 GHz

Atom
D525 1.8 GHz
AMD
E-350 1.6GHz
Motherboard Gigabyte
A75M-UD2H
Jetway
NC94FL-525-LF
MSI
E350IA-E45
North bridge A75 NM10 Hudson
M1
South bridge
Memory size 8GB
(4 DIMMs)
4GB (2 DIMMs) 4GB (2 DIMMs)
Memory type Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CM2X2048-
8500C5D

DDR2 SDRAM

Corsair

CMD8GX3M
4A1333C7

DDR3 SDRAM

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

drivers

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

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

 

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

Core
i7-3960X
Motherboard Gigabyte
X58A-UD5
Asus
P8P67 Deluxe
Intel
DX79SI
North bridge X58 P67 X79
South bridge ICH10R
Memory size 12GB
(6 DIMMs)
8GB
(4 DIMMs)
16GB
(4 DIMMs)
Memory type Corsair

CMP12GX3M
6A1600C8

DDR3 SDRAM

Corsair

CMD8GX3M
4A1600C8

DDR3 SDRAM

Corsair

CMZ16GX3M41866C9R

DDR3 SDRAM

Memory speed 1333 MHz 1333 MHz 1600
MHz
Memory timings 8-8-8-20 2T 8-8-8-20 2T 9-9-9-24 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

INF
update 9.2.3.1022

RSTe 3.0.0.9012

Audio Integrated

ICH10R/ALC889 with

Realtek 6.0.1.6235 drivers

Integrated

P67/ALC889 with

Microsoft drivers

Integrated

X79/ALC892 with

Realtek 6.0.1.6449 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:

  • The Core i7-2700K results that appear in the following pages are “simulated;” they come from a Core i7-2600K whose Turbo Boost policy has been modified to match the 100MHz-faster speeds of the 2700K. Due to limitations in the motherboard’s firmware, we weren’t able to modify the 2600K’s baseline clock speed from 3.4GHz to 3.5GHz in order to match the 2700K. However, we’ve not seen a Sandy Bridge chip spend any significant time resident at its baseline clock when both Turbo and SpeedStep were enabled, as was the case here. We pinged Intel PR about whether the baseline clock multiplier mattered to performance, and PR in turn pinged an Intel engineer, who didn’t have a definitive answer to the question, either. So… your mileage may vary. We’ve elected not to include the 2700K in our power consumption testing, since we didn’t have a sample of the actual product.
  • 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.

Sandy Extreme’s memory hierarchy achieves more throughput at every step of the way than any other desktop processor. This chip’s massive L3 cache gives it a commanding throughput lead at the 16MB block size, where no other desktop CPU comes close. Unfortunately, we suspect few desktop workloads have a working data set large enough to fit into this L3 cache and not into the caches of more pedestrian chips. As a result, Sandy Bridge-E’s large last-level cache may be a victim of the law of diminishing returns.

On paper, Sandy Bridge-E has a tremendous amount of memory bandwidth compared to anything else out there. In reality, Sandy Bridge-E has a tremendous amount of memory bandwidth compared to anything else out there, too. What more can you say about this one? With twice the memory channels of its sister, Sandy Bridge, the Extreme part delivers a little more than two times the throughput in our directed test.

I’ve included this result for completeness, but the outcome is probably a little bit deceiving. I believe the 3960X’s 15MB L3 cache is masking some of the access latency at this 16MB block size. Next time out we’ll have to sample from a different spot or use another tool to measure latencies to main memory.

For what it’s worth, our tool reports Sandy Extreme’s L1 cache latency at four cycles, L2 latency at 12 cycles, and L3 latency at 30 cycles.

Battlefield: Bad Company 2

So begins our suite of gaming tests, which present us with a bit of a dilemma when it comes to a product like the Core i7-3960X. As you can see, the fastest solutions are pretty well bunched up together at the top of the chart. That’s true probably for several reasons. First, most games don’t take advantage of more than three or four threads, max, so any triple- or quad-core processor that can execute those threads quickly will excel here. Second, most of today’s games aren’t particularly CPU-intensive, since they’re cross-developed for game consoles like the Xbox 360 and PlayStation 3, whose microprocessors have more in common with the engine control computer in a Ford Contour than they do with a Sandy Bridge Extreme. Third, to the extent that it matters when the minimum frame rates are higher than most display refresh rates, it’s possible our test is GPU-limited, even at this relatively low resolution. That will happen on a modern PC, whether it’s with a mid-range graphics card like this one at a middling resolution or with a high-end graphics card on a four-megapixel display.

With that said, we had intended to include some additional testing, using newer games and a high-end graphics card and going inside the second with the top CPUs, but the aforementioned issues we had with the Asus mobo’s Turbo policy forced us to scuttle those plans. We’ll get back to it eventually, but don’t get your hopes too high. Our recent look at Battlefield 3 performance with a lowly Core i5-750 produced prevailing frame times near the 16-millisecond mark with the faster GPUs. That’s a solid 60Hz or 60 FPS, which doesn’t leave loads of room for tangible improvements with a quicker CPU.

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.

That first result is obviously GPU-limited, and removing that limitation with the “no render” setting allows the 3960X to assert it superiority over the prior-generation champ, the Core i7-990X.

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.

The 3960X comes out on top here in both cases, with a slight edge over the 990X. In the “no render” test, the six-core processors are markedly faster than the Sandy Bridge quad-core chips, a testament to how fully multithreaded this game happens to be.

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.

Even at the lowest resolution, this game doesn’t run any faster on the 3960X than it does on the 2700K, for many of the same reasons we discussed in the Bad Company 2 results. We’ve included the higher resolutions so you can see what happens when a GPU bottleneck takes over, but I wouldn’t recommend dwelling too much on those results if you’re trying to compare CPUs.

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.

Phew. This is our last set of gaming results, so we can move on to applications where the CPU is the primary performance bottleneck. Obviously, that’s not the case here.

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.

In this nicely threaded test of CPU performance, the Core i7-3960X assumes its rightful spot at the top of the pecking order.

Productivity

SunSpider JavaScript performance

This test is lightly threaded, so the 2700K manages to eke out a slight victory. Still, the 3960X’s Sandy Bridge cores give it a measurable edge over the Core i7-900 series CPUs.

7-Zip file compression and decompression

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 supports Intel’s custom-tailored AES-NI instructions, so the encoding of the AES algorithm, in particular, should be very fast on the CPUs that support those instructions.

A couple generations of Intel processors have supported AES acceleration, and AMD’s FX processors have recently joined the club. The new club president, though, is the Core i7-3960X.

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.

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. For the sake of brevity, we’ve included a single overall score for those real-world tests, along with an overall score for picCOLOR’s suite of synthetic tests of different image processing functions.

The 3960X dominates our image processing tests, jumping out ahead of the Core i7-990X and the 2700K.

If you haven’t noticed, we’re getting to the portion of our benchmark suite that’s nicely parallelized, and the 3960X is liable to be fastest in every test. I’m going to be limiting my commentary, as a result, to places where there’s something interesting to discuss.

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.

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.

Neither of our image processing applications is perfectly multithreaded. Pass one of x264 only uses a few threads, and Windows Live Movie maker doesn’t appear to benefit from the presence of more than four cores. Still, the 3960X finishes first across the board.

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.

POV-Ray rendering

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

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.

With the exception of POV-Ray’s benchmark scene, all of our 3D rendering tests scale well with the addition of more cores and threads. No surprise, then, that the 3960X reaches new performance heights in each.

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.

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.

Euler3D uses double-precision floating-point datatypes, which are particularly bandwidth-intensive. As we saw in our synthetic memory benchmarks earlier, the Core i7-3960X is unmatched in terms of memory bandwidth among today’s desktop CPUs. That helps explain its large margin of victory here.

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.

Pretty impressive results. Our Core i7-3960X system draws no more power at idle than a P67-based Sandy Bridge quad-core rig. At peak, the 3960X system uses less power than any of the test systems based on its predecessors in the Core i7-900 series.

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.

Our Sandy Bridge-E test system uses less energy to render the Cinebench test scene than anything else we’ve tested. The keys to its efficiency are relatively modest power draw at peak and, most crucially, the ability to complete the task in the shortest amount of time.

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 AMD’s Bulldozer, as well as Intel’s Sandy Bridge family: AIDA64 from FinalWire. AIDA64 includes several small, synthetic tests that can be accelerated with AVX’s new instructions.

In order to measure the impact of those new instructions on performance, I tested several of the CPUs 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.

We have another round of AVX tests from the latest version of SiSoft’s Sandra. Although these quick benchmarks are labeled “Multimedia” in Sandra, in truth they’re simply fractal computations like the AIDA64 Julia and Mandel tests.

Sandy Bridge-E’s prowess in these early AVX-enabled benchmarks is impressive. We expect this platform’s substantial memory bandwidth will allow it to put AVX to good use as more and more real applications add support for these instructions.

Overclocking

The 3960X turned out to be a surprisingly sweet overclocker, happily agreeing to run all six cores at 4.8GHz at only 1.425V. We tried to push it further, to 4.9GHz, but no amount of voltage up to 1.55V was sufficient to keep it completely stable at that speed.

That’s probably just as well, because we found that even the nifty Intel-branded liquid cooler we were using wasn’t quite up to the task of keeping the processor cool at 4.8GHz and 1.425V. After running a Prime95 torture test for three or four minutes, the CPU would eventually begin to throttle back to lower clock speeds, presumably because it was getting too hot. (I’d tell you that with more confidence, but we used the Asus P9X79 Deluxe for our overclocking tests, and Asus’ monitoring tools for Windows weren’t yet compatible with that board.) More than likely, the throttling would go away with beefier cooling. Still, we were able to get a decent set of overclocked benchmark results before throttling ruined the party.

One impact of having used the Asus motherboard for our overclocking attempts is somewhat higher power draw in the test above. Still, most of the increased power consumption comes from the CPU itself, a consequence of a large number of transistors operating at a much higher clock speed and voltage.

Conclusions

Obviously, the Core i7-3960X is the new top dog in the desktop CPU performance sweeps, as a look at our value plot will show.

Although it’s still not a great value, the 3960X provides a solid incremental improvement over the Core i7-990X in our overall performance index. Peel back a layer, and you’ll find that the 3960X also clears up any questions about whether the Sandy Bridge-based Core i7-2700K might be faster in lightly threaded applications. At its best, the 2700K can only hope to match the 3960X’s performance in such cases—and it’s no contest at all when lots of threads are in play. If there’s a memory bandwidth limitation in the mix, the 3960X has the potential to far surpass the prior-gen Core i7-990X, as we saw in the Euler3D computational fluid dynamics simulation.

Step beyond raw performance, and the positives continue to add up. The 3960X’s power consumption is relatively tame for a chip of this class, especially at idle, where our test system drew no more power at the wall socket than a quad-core Sandy Bridge-based system. The 3960X is the new leader in our task energy test, a measure of power-efficient performance, as well. This chip also comes with quite a bit of built-in headroom for overclocking, if our experience is anything like typical.

With that said, we can’t help but notice the Core i7-3930K would surely occupy a more enviable position in our overall value plot, with performance only a smidgen below the 3960X’s for hundreds of dollars less. Having seen the overclocking potential of Sandy Bridge-E silicon, our initial impression is now stronger than ever: no one should buy a Core i7-3960X when the blissfully unlocked 3930K exists.

We’re duly impressed by the overall potential of the Sandy Bridge-E platform, which is unlike anything else that’s come before; it’s as if Intel crammed a dual-socket workstation into a single socket. For the PC enthusiast, this is truly lust-worthy computer hardware. However, there are still some negatives to the platform that we can’t overlook, mostly related to the fact that this product introduction feels strangely rushed in some ways. The state of the motherboards is unusually messy for a notable Intel platform launch, with major-brand boards and fairly critical firmware fixes still trickling into our labs at the eleventh hour. The X79 chipset’s relatively modest SATA capabilities and initial lack of SSD caching only add to that impression. Although it’s not really Intel’s fault, the PCIe 3.0 situation also feels unfortunate. Finally, as we’ve noted, we wonder whether the typical buyer of a thousand-dollar CPU and a $300-plus motherboard will welcome the fact that his CPU has two of its cores and 25% of its last-level cache lobotomized. There’s a certain psychology involved in buying the best, and we’re not sure it’s compatible with Intel’s decision to nix those bits of the chip.

That psychology is particularly important because, honestly, having and owning the best has to be one of the major motivators for folks buying or building a Sandy Bridge Extreme system. If you’re just a hard-core gamer who wants glassy-smooth gameplay or a power user who wants a system that feels as snappy as possible, you can get sublime satisfaction out of one of the quad-core variants of Sandy Bridge, like the Core i7-2700K or even our pick for best value, the Core i5-2500K. The best reasons for looking past those options to Sandy Bridge-E are either a yearning for bragging rights or a specific usage model that requires additional parallel processing power—say, video editing or heavy image processing work. Well, and let’s not rule out the Great Justifier, which is some combination of a CPU-intensive usage model and raw desire.

However it works in your head, if your mix of needs, wants, and means leads you to seek the best PC money can buy, there’s no question where to look. Sandy Bridge Extreme is it. Just try not to think about those two cores permanently sitting idle.

Comments closed
    • Mr Bill
    • 8 years ago

    Would this be the used transistors or does that number include the unused cores?

    “As you might expect, cramming all of this processing power, cache, and I/O into a single chip isn’t without consequences. Sandy Bridge-E is fairly enormous, with 2.27 billion transistors in a die area that makes even AMD’s Bulldozer look dainty.”

    • Mr Bill
    • 8 years ago

    Question: Is it always the same two cores that are disabled on the die, or can it be any two cores? In other words could it be that these are dies that for one reason or another (other than thermal envelope) could not be validated for 8 cores to be run at lower frequencies in servers?

    • Mr Bill
    • 8 years ago

    Ya really have to wonder why AMD has not jumped on the memory bandwidth wagon. First it was AMD with two channels and Intel with three channels, now Intel has upped the ante to 4 channels? Hello, AMD? I’m a huge fan, but if this continues, my next build will be an Intel.

    • internetsandman
    • 8 years ago

    Whats the odds that motherboard makers release a UEFI tool to attempt enabling those two extra cores? They did it forever with the old Athlons and Phenoms, would it be possible here as well?

    • gmskking
    • 8 years ago

    For most people, I really do not see this processor being worth $1000 when you can get a 2600k for $300 for a little less performance. I doubt Intel will sell too many of these at that price. But what do I know. I still like my Q6600 and have been given no reason to upgrade yet.

    • Abdulahad
    • 8 years ago

    They are both BLUE….But what’s the difference…..???
    VIAGRA: You buy one…. one hour of erection
    COREi7-3960X: You buy one…… one year of impotency

    IMPOTENCE INSIDE…:-)

    • Abdulahad
    • 8 years ago

    They must be crazy if they think I’m going to buy such expensive chips, allowing my wallet to get deep fried and my waist holding the pang all year through….

    It seems that the BLUE Team’s high-end prices make you see BLACK, knock you out straight and by the time you regain consciousness, your system is fossilized….

    • HisDivineOrder
    • 8 years ago

    I’ll agree with the consensus. Damn shame about those two cores they left lopped off at the factory and the fact that the efficiency is so eh compared to the rest of the SB line. Plus, not being able to offer Quicksync without the iGPU was a problem when they had to release a new chipset (and license some third party software) to enable usage of it combined with a discrete GPU back at SB launch and its a problem now when its lopped off along with the iGPU for SB-E.

    The power it’s sucking down at the high overclock is nothing short of ehhhh, but I see the market they’re aiming at and appreciate the fact they’re not targeting most people. I also get that they’re going for the crowd who will buy the $1k+ 3960, a $350+ motherboard by Asus or eVGA, and then slap in four 580’s with 3GB VRAM and laugh like a maniac when the few games that really scale with true quad-SLI work out. Sure, they’ll be sweating like they’re in a sauna, but that’s who’ll really appreciate this thing.

    Everyone else… well, they’ll probably be ready for an Ivy Bridge upgrade that runs cooler with GPU’s that are newly die shrunk and incredibly optimized for power after being designed entirely during the post-power-sipping generation began. These folks will be able to leave the A/C off during the summer while the SB-E Monolith crowd will be installing A/C units for their computer room to compensate for the heat rolling off it in waves.

    Then again, they’ll be the only ones with the money to do so anyway. It’s great that IB will offer 99% of the performance for a fraction of the cost, though. Kinda funny that AMD couldn’t match the performance of regular SB (or its own precursor line) and still managed to beat the power/heat INefficiency of SB-E all at the same time with Bulldozer.

    That takes skill.

      • chuckula
      • 8 years ago

      [quote<]Kinda funny that AMD couldn't match the performance of regular SB (or its own precursor line) and still managed to beat the power/heat INefficiency of SB-E all at the same time with Bulldozer.[/quote<] Reader FAIL. Go back and read the TR review where the 3960X has the lowest task energy of any CPU TR has ever tested to complete Cinebench. People confuse peak power consumption (which BTW is *lower* for the 3960X compared to the Bulldozer) with energy efficiency all the time. Even though TR is just about the only site on the Internet that knows that energy is measured in joules and not watts, people still apparently don't want to read what TR writes....

    • Bensam123
    • 8 years ago

    And the price whoring has started. Indeed, I’m sure two disabled cores is solely to keep it in the 130w envelope. If that were the sole reason and Intels intentions were good they could be enabled in the bios if the user wishes.

    Adding to this, who wants to bet the whole SATA 3.0 and USB 3.0 BS is to push the horrid Thunderbolt? The direct competitor (USB 3) getting more of the shaft.

    All of this sounds planned. All these ‘oopsies’ and teething issues are waaaay too coincidental and well planned out. I bet the original delay for Intel was that they wanted to see what BD had in store and then make some last minute adjustments to their product. The milking equipment is already running.

    Poor AMD, I miss you already and you haven’t even finished walking out the door. πŸ™

    PS the gaming graphs make me more depressed.

    • glacius555
    • 8 years ago

    Soo..

    It is LGA2011 because it was planned for 2011? πŸ˜›

    • j1o2h3n4
    • 8 years ago

    The VT-d bug (hardware accelerated virtualization) still exists, so its permanently disabled altogether (source: arstechnica). Meaning we are about to own / pay for something that’s not fully functional and we don’t care. Most don’t even know it existed.

    • Hallucinosis
    • 8 years ago

    Try VLC Media player for encoding to H.264. On my i7 980x it uses all 6 cores (+6 HT threads) at nearly 100% while I encode 2560×1440@FPS AVI to H.264. Does it better than any other conversion program I’ve attempted to use–most seem limited to 1080p.

    [url<]http://youtu.be/5z0La5pZMMs?hd=1[/url<]

      • indeego
      • 8 years ago

      How long did that take?

      • yokem55
      • 8 years ago

      VLC simply uses x264 internally for its h264 encoder.

    • TurtlePerson2
    • 8 years ago

    Three things impressed me with this chip: the memory access times, the overclocking headroom, and the efficiency (power consumption-performance). All three of these are things that are extremely difficult to do from a performance perspective.

    Cache design and memory systems have been more or less figured out for a long time. Modern microprocessors pretty much just add capacity and improve pre-fetching logic in order to reduce the cache miss rate. Clearly Intel has done both very well to earn those scores in the memory system tests.

    4.8 GHz is a very impressive clock speed even for an overclock. A decade ago it seemed that 4 GHz was the barrier that few would ever hit. This kind of overclock just goes to show how it’s TDP and cooling that’s holding back IC design and not the transistors anymore.

    The fact that Intel has managed to create a faster version of its previous chips while slightly improving the power efficiency is very impressive. You generally pay a large efficiency price to improve performance, but somehow they have done so without paying any price.

    That all having been said, the price on this thing is ridiculous. I could build two decent gaming rigs for the price of one of these CPUs.

      • NeelyCam
      • 8 years ago

      [quote<]This kind of overclock just goes to show how it's TDP and cooling that's holding back IC design and not the transistors anymore..... ...You generally pay a large efficiency price to improve performance, but somehow they have done so without paying any price.[/quote<] Transistors are still the limiter. Faster and more efficient transistors -> more performance within a given TDP envelope. The price Intel is paying is the development cost of a new process node. Better process is the key reason for improved power efficiency.

    • boogaloo
    • 8 years ago

    As hilarious as having the atom processor added to the benches is, it really makes the difference between the modern processors difficult to see.

    • kamikaziechameleon
    • 8 years ago

    the fact that they still don’t have a consumer 8 core processor available simply so they can maintain the 130W thermal envelope is a bit goofy. I’m actually rather annoyed by intels choice here. It feels more like they engineered it then said, hey wait we will release those 2 cores with the fab revision in 12 months otherwise there is no way those products can compete.

      • Farting Bob
      • 8 years ago

      If they fitted 8 cores of SBE onto the chip it would consume more than 130w under load and thats a worrying amount and would bring bad publicity. At the same time, they wont want their flapship extreme CPU to be beaten by a Quad core version at 1/4 of the price in single/lightly threaded tasks as that wont sell many chips. So they have to keep clock speeds at the same or higher that the quad parts are at but trying to put in double the cores. This is an acceptable balance. If you NEED 8 true cores then you should wait for better things to come or spend serious money on server parts.

        • kamikaziechameleon
        • 8 years ago

        I read that part of the review. I don’t like how their is wasted space on there. They can make reasons/aka excuses end of the day it just makes the 1,000 dollar part irrelevant.

          • OneArmedScissor
          • 8 years ago

          Ever bought an Intel CPU that wasn’t like $300? Nearly all of them have “wasted space” below that price.

            • kamikaziechameleon
            • 8 years ago

            yeah my Q6600 was great!

            • HisDivineOrder
            • 8 years ago

            Except this is a $1k+ CPU. When you’re paying more for a CPU than the highest end iPad or a low-end Macbook Air, then you just want it to be a fully armed and operational battle station. You don’t want it to be drifting there in the middle of space, unarmed, hoping against hope its parts aren’t showing.

            Besides, we know Intel is just giving themselves breathing room to release another version of this CPU in the next six months “but now with 2 extra cores!”

            Perhaps around the time Ivy Bridge arrives and shows us what the 3xxx series is really about.

            • OneArmedScissor
            • 8 years ago

            [quote<]Except this is a $1k+ CPU.[/quote<] No, it is $1,000 even, which gets you a CPU worth $1,000 even. Xeons go up to $5,000. Sorry, but that's what these chips are. [quote<]Besides, we know Intel is just giving themselves breathing room to release another version of this CPU in the next six months "but now with 2 extra cores!"[/quote<] Don't count on it. That would require that their high end Xeons be substantially devalued, and then it would still likely cost $1,500 and not $1,000. Here is the frame of reference for what "we know:" The current cheapest 8-core Westmere EXs start at $1,400 and if you only pay that much, you get less cache, memory bandwidth, and much lower clock speeds than a "full" SBE. If you want the "full" Westmere EX, you're going to have to cough up $4,800.

            • flip-mode
            • 8 years ago

            Do the Xeons use socket 2011? I don’t set Intel’s pricing policies, so my opinion means zilch, but I don’t think that the pricing of the desktop Extreme Edition has much bearing on the pricing of the multi-socket server chips. Don’t those server chips have a bunch more memory support too? Either way, Intel may make the EEs from Xeon silicone and so may share most of the features, but even if the silicon was “fully functional” on the EE there would be platform constraints. They serve totally different customers, so the pricing of one should have little bearing on the pricing of the other – in my completely irrelevant opinion.

            • Krogoth
            • 8 years ago

            Next-generation Xeons are going to use LGA2011.

            It is Intel’s next-generation workstation/server platform which is set to replace LGA1366.

            This is where quad-channel DDR3 support and integrated PCIe controller will come into play. πŸ˜‰

            • kamikaziechameleon
            • 8 years ago

            There has never really shown to be a linear relationship between prices for consumer single socket processors VS. the multi socket server market. If I’m not mistaken you could get more cores for less from the server market for a while.

        • kamikaziechameleon
        • 8 years ago

        I’m a bit of a goof but using power gating and turbo core they could have easily done the 8 cores that would have crushed the multi threaded apps and then turned off cores to turbo boost up 1 or 2 cores for single and double threaded apps or am I mistaken. There is no reason as far as I can see to legitimately turn off the cores beyond leaving headroom for future

    • flip-mode
    • 8 years ago

    Disabling the two cores is lame. If I paid $1,000 for a CPU I’d not want it to be a hobbled CPU.

    Other than that I suppose SB-E is impressive in several ways while at the same time making SB feel even more right than it already did.

    435mm^2 sure is big.

      • MathMan
      • 8 years ago

      You realize that all your DRAM is hobbled in the same way, right? And the NAND flash on your smart phone. etc.

        • flip-mode
        • 8 years ago

        No, I didn’t realize, but that does not make this CPU hobbling less lame. Again, this is “The King” of CPUs, while my DRAM and the NAND flash in my phone aren’t really anything special that are marketed under the Intel brand name and the “Extreme” moniker or come with a $1000 price tag. Intel’s top of the line Extreme enthusiast CPU should not have anything disabled – that’s my feeling on the matter. And really, if that’s the only thing to complain about on this processor then Intel’s in a pretty good place, ya know.

          • OneArmedScissor
          • 8 years ago

          Westmere EX is a 10 core chip, and they never sold that with more than 6 cores under $1,000. This is, in some ways, the chip that will be used as its successor, so why would things be different at the same manufacturing node?

            • flip-mode
            • 8 years ago

            I don’t pay much attention to what goes on at the $1,000 CPU price level, so I didn’t know that there have been previous Extreme Edition CPUs that also had disabled parts. I don’t really care as I’ll never be able to afford one, but it seems lame when I can’t afford it so I’m sure it would seem lame if I could afford it. Lame or not, it’s still the fastest CPU. I suppose it’s a lot less lame than AMD’s 8 core CPU that can’t even perform as well as Intel’s 4 core CPU.

            • OneArmedScissor
            • 8 years ago

            They haven’t sold one with disabled cores before, but they pretty much always have disabled parts/featured, or aren’t the fastest parts.

            The one exception I can think of is when they actually sold the dual-socket board as the enthusiast platform, like with the QX9775 and Skulltrail, but those CPUs were $1,500, just like their Xeon counterpart.

            Point being, it always costs what its Xeon counterpart does. They’re the same chips with the same boards, so if they charged less, they’d be letting someone pay less for the same thing.

            This go around, they just opted to use the ginormous version of the chip as the standard, kind of mixing together a Bloomfield/Gulftown and Nehalem/Westmere EX replacement.

            It’s really not that different from how a Core 2 Quad Extreme was two Core 2 Duo Extreme chips stuck together, making a much larger chip, which cost quite a bit more. If you bought the Core 2 Duo Extreme for $999, it’s not as if you were being denied something.

          • flip-mode
          • 8 years ago

          Darn Nvidia haters are thumbing me down again. Stinking legions of Nvidia haters at this site – shame on them!

      • NeelyCam
      • 8 years ago

      [quote<]Disabling the two cores is lame. If I paid $1,000 for a CPU I'd not want it to be a hobbled CPU. [/quote<] Not this again... You're paying $1,000 for a six-core CPU. That's it. If you want an 8-core chip, pay for an 8-core chip when they become available, and it's surely going to be more than $1,000. [quote<]435mm^2 sure is big.[/quote<] Yes it is...

      • kamikaziechameleon
      • 8 years ago

      “Disabling the two cores is lame. If I paid $1,000 for a CPU I’d not want it to be a hobbled CPU.

      Other than that I suppose SB-E is impressive in several ways while at the same time making SB feel even more right than it already did.”

      My thoughts exactly this launch has only made the i7 2700 and i5 2500 look even better.

        • kamikaziechameleon
        • 8 years ago

        as explored in other posts it is apparent that with power gating and turbo core there is no legit reason to turn off the other 2 cores beyond to basically leave consumer processor performance gains head room down the road.

    • OneArmedScissor
    • 8 years ago

    Request:

    The list of CPUs has become overrun with “duplicates” which could just be left out. I’m guessing they’re there because you already had the results from before, but since so many of them have two of the same chip with about the same clock speed right next to each other, it’s looking kind of jumbled.

    It’s fine to have about that many, as the full chart still fits in my dinky laptop browser window, but it would be nice if there were, say, the lowest and highest end of a CPU if it’s going to be in there twice, rather than the two highest.

      • Farting Bob
      • 8 years ago

      I like Anandtech’s bench section. You can search for any card they have ever tested on the same test setup and compare, but in the main reviews usually only relevent ones are shown. So you can compare your old CPU or GPU to the newest ones in the benchmark section but in the main review it will just show the current generation.

        • NeelyCam
        • 8 years ago

        You know, a year ago saying something nice about Anandtech was considered punishable by death.

        How times have changed… I guess the AMD Fanboi Thumbdown Brigade disbanded after the BD fiasco, or moved to the safety of SemiAccurate forums.

        If you feel particularly nostalgic, go to SA forums and try to post a link to Anandtech – preferably to a page that says something positive about Intel – and see what happens.

          • Fighterpilot
          • 8 years ago

          Wait a sec…you,NeelyCam…,Intel shill and self admitted SA forum troll are whining because the “AMD Trolls” [i<]aren't[/i<] here?

            • NeelyCam
            • 8 years ago

            I wasn’t whining about anything. I was just pointing out something that I found interesting.

            BTW, I’m not an SA forum troll, and there really aren’t many others, either. SA Forums are pretty closely monitored, and banhammer swings on a regular basis. But even then, minor anti-intel trolling seems to be accepted and even encouraged (Rich Wargo is one of the key anti-Int-El trolls).

    • Dr_b_
    • 8 years ago

    Better wait for Ivy Bridge. This product is too expensive, and there are too many things not available in the chipset, like USB 3.0 and the missing SAS ports. With a crippled CPU, the chipset issues, and the expense it’s not worth the upgrade for the performance gain unless you are doing something heavily threaded, and maybe already do not have a decent CPU now. It is geared towards the workstation market based on these specs and limitations, but they are also selling it to the extreme/enthusiast crowd who should also not buy it, because right around the corner is Ivy Bridge or a respin of this chipset and new steppings of this same CPU variant. It is not a smart play to be an early adopter here.

    • maxxcool
    • 8 years ago

    soooo charlie was right? pci-e 3.0 *is* somehow borked on SB….

      • NeelyCam
      • 8 years ago

      His update says PCIe3.0 is working and enabled, but not certified because Intel couldn’t find validation boards.

      • chuckula
      • 8 years ago

      No, charlie was not right. Intel incorporated a *standards compliant* implementation of PCIe 3.0. The trick is, having a standards compliant implementation is useless if there are incompatibilities with actual working devices that other manufactures make that are *not* necessarily 100% compliant with the standard. Intel has not had an opportunity to test with real hardware since the real hardware isn’t even available yet. Because of that, they took the ethical route and only said that SB-E is compliant with PCIe 2.0. I’d say there’s a 90%+ chance that SB-E will work fine with next generation video cards using PCIe 3.0, but you can’t tell for sure until the cards are in place and running.

        • sschaem
        • 8 years ago

        Most likely, PCI3.0 device manufacturer will use this platform as part of their validation.

        • NeelyCam
        • 8 years ago

        How do you know Intel’s implementation is standards compliant if it’s not certified?

          • Farting Bob
          • 8 years ago

          Im guessing at this point there is no certification but Intel followed the required spec for PCIe3 fully, and thus if everyone else follows the same spec fully there is no issues. If others deviate from the spec in their final implementations and that because the de facto standard then Intel’s might not be fully PCIE3 compatible. But it will work at PCIE2 if all else fails, that is a guarantee.

    • Krogoth
    • 8 years ago

    Sandy Bridge E = Workstation and Server-class hardware rebranded to epenis junkies with more money than sense. It is a poor value for desktop users.

    The only reason Intel disabled the two cores is because the chip couldn’t the same clockspeed as 2600K and 2700K with 8 cores enable without becoming a blast furnace. Intel couldn’t have their “XE” line-up be slower at desktop related tasks (clockspeed is king here). πŸ˜‰

    If you want to get octal-core version, you have to get Xeon brand. I suspect that an 3.0Ghz 8-core SB-E would actually pull ahead from i7-3960X in tasks where number of threads are king and set new records in power efficiency.

    • supercomp
    • 8 years ago

    This review got me more exited about the Ivy Bridge-E, and how far away can it be if the SB-E cooler will work with Ivy Bridge-E? πŸ˜‰

      • Ushio01
      • 8 years ago

      A year at best.

    • Anomymous Gerbil
    • 8 years ago

    Why on earth does anyone care if two cores are disabled?

    There could be 47 disabled cores in there for all it matters – you know what you’re getting for your money (i.e. 6 cores and the performance described in here), and you make a decision on that basis.

      • chuckula
      • 8 years ago

      I tend to agree that there is a lot of manufactured outrage over the fact that 2 cores are disabled. If somebody really wants an 8 core CPU then it will be possible to get a workstation class board and slot in a xeon when they are released.

      –> Ah, and I see by the downrating of Anonymous Gerbil’s post and my post that the AMD squad is in full damage control mode on this one…

        • clone
        • 8 years ago

        their is no AMD squad, only those who wish their was one for the sake of competition and those who want to rattle the fanboy sabre atop an empty hill all by themselves, AMD is in the process of reinventing itself in a race to compete in markets Intel hasn’t taken yet.

        the CPU war is over, it ended after the introduction of Intel’s Core series cpu’s several years ago…. it might resume in 7 years but it’s doubtful that CPU’s will matter much in 7 years let alone AMD would still be trying to compete on architecture when they will never & never did compete with Intel’s manufacturing capability.

      • Krogoth
      • 8 years ago

      This product doesn’t any sense for normal desktop users. It is just as fast as the regular i7-2600k in non-server and workstation tasks for 2x to 3x platform cost.

        • Anomymous Gerbil
        • 8 years ago

        Your reply is not relevant to my post.

        • sschaem
        • 8 years ago

        ?? the I7-3820 is CHEAPER then the i7-2600k. And an lga2011 MB itself doesn’t make so your over system cost will be 3 time higher.
        And the 3820 is ~5% faster then the i7-2600K

          • Krogoth
          • 8 years ago

          Again, there’s no “official” pricing and availability for the i7-3820. I also don’t see any official benchmarks on it either.

      • NeelyCam
      • 8 years ago

      This. +1.

        • NeelyCam
        • 8 years ago

        ThumbDownNeelyCam() script is alive and well.

          • Meadows
          • 8 years ago

          That’s no script, sunshine.

      • flip-mode
      • 8 years ago

      Why on earth does anyone have to defend wanting a fully functional piece of silicon?

      A broad range of opinions and views have typically been welcomed at this site. It’s pretty weird to hear people criticized for wanting an option to buy the fully functional piece of silicon. It’s not a fanboy issue – if AMD had launched Zambezi in a partially disabled state, well, that would have resulted in the surprising situation of something that was actually even more disappointing than Zambezi already was (which is frankly difficult for me to imagine).

      The competitive landscape is pretty grim when Intel can release a partially disabled chip and it’s still easily the fastest CPU in the world.

        • Anomymous Gerbil
        • 8 years ago

        I think you missed my point. My simple point is that the fact that *this* chip has two disabled cores is totally irrelevant to anyone’s purchasing decision. One day the 8-core version will come out, and then people can make a decision on *that* chip’s performance, price, power consumption etc. It’s got nothing to do with fanboys, Zambezi, or anything else.

          • esterhasz
          • 8 years ago

          To the purchasing decision, yes. I still deplore that there isn’t a lower clocked 8 core version available at launch. I am shopping for a datamining machine for the office *now*.

            • Anomymous Gerbil
            • 8 years ago

            Patience, Grasshopper πŸ™‚

          • flip-mode
          • 8 years ago

          Your point is completely valid and I did not miss it, but just as your point is valid, there is also validity in the disappointment that others are feeling in the fact that the chip is partially disabled and in the fact that the new Extreme Edition offers no more cores than the old Extreme Edition.

      • Bensam123
      • 8 years ago

      If it was just ‘just’ two cores this one time on this one unit baring the looming shadow of something bigger happening in the future. Some people think about a lot more then just the implications of one thing at one point in time.

    • derFunkenstein
    • 8 years ago

    Well, here’s the first review that makes me want to replace my 2.5-year-old system…with a regular LGA1155 Sandy Bridge setup. :p

    It’s a ridiculous CPU, with power to spare – mostly because many of the apps can’t use all the cores, but in the ones that do, it blows the doors off of everything else out there.

      • FuturePastNow
      • 8 years ago

      Kind of makes me want to start prowling eBay for cheap, used LGA1366 processors being replaced by fools who buy this monstrosity.

        • Bauxite
        • 8 years ago

        You’re behind the curve, the nuts started dumping last week to have cash for this week.

        My 920 from 2009 just got replaced by a gulftown steal from craigslist πŸ™‚ $100 swap.

    • ronch
    • 8 years ago

    I wonder whether AMD will ever be able to get back on their feet and offer something which consistently tops the charts again.

    • rechicero
    • 8 years ago

    Just one quick suggestion. Would it be possible to add diagonals in the scatter plot diagrams? That would be the best way to asses the performance/$ ratios.

    And it should be really easy to do.

    Another way could be a bar diagram similar to what you sometimes do with SSD and HDD

    Thanks!

      • jonbanh
      • 8 years ago

      Ideally, you would consider what your budget for a CPU is and navigate to the correct column to pick the highest performer for what you’re gonna spend. Diagonals would not be the most useful, but you can take a screenshot and go line crazy to find out for yourself.

    • srg86
    • 8 years ago

    The amount of cache on these things is amazing. Just to put it into perspective, this chip has 20MB, with 15MB enabled in this desktop part.

    15MB is just 1MB shy of the entire physical memory address space of the 80286, MC68000 or 386SX.

    With this amount (and even better with the 20MB version) you could pretty comfortably run an entire DOS/Windows For Workgroups 3.11 system. Running things like Office, CorelDRAW! and maybe a few others entirely from cache, without needing to install any DRAM what-so-ever, just runnning in SRAM!

      • derFunkenstein
      • 8 years ago

      Except that pretty much nothing will POST successfully without RAM.

        • srg86
        • 8 years ago

        True, but my point more more theoretical, not practical.

      • OneArmedScissor
      • 8 years ago

      Now consider a Westmere EX server. 240MB of L3! I think my Windows fits in that.

    • Forge
    • 8 years ago

    Damage: Thank you for burning the midnight oil re-re-rerunning these benches. It’s a glorious thing.

    Intel: The king is dead, long live the king. If you’ve got a spare LGA2011 setup, I can offer it a *very* appreciative home.

      • Duck
      • 8 years ago

      Way to suck up there

    • ModernPrimitive
    • 8 years ago

    If not for the lacking features on the x79 I would almost consider a 3930 for a build over a 2600k. Just my opinion, but for a $1000 a pop (3960x) I’m gonna have to ask for those 2 disabled cores and cache to make it seem worth it. As always though, for those with the cash to play with, more power to ya and Il’l enjoy reading about your exploits.

    • odizzido
    • 8 years ago

    The minimum FPS in metro2033 is pretty terrible with this CPU compared to much cheaper options.

    Nice idle power draw though.

      • OneArmedScissor
      • 8 years ago

      [quote<]Nice idle power draw though.[/quote<] Yeah, that's impressive with all the memory channels and cache. I fully expected this to be X58 2.0, but they finally stepped up their game in the platform shortcomings. Good luck with that, AMD...

    • Mourmain
    • 8 years ago

    Wait, I thought I read the entire article, and yet I must have missed the reason why those two cores are disabled. Where is that explained again?

      • kroker
      • 8 years ago

      Anand says SB-E is a die harvested Xeon / Sandy Bridge EP, with 2 cores and 25% of the cache disabled. So basically, you pay premium price for a Xeon reject. The reason is probably related to yields of such a large die, and lack of competition.

        • smilingcrow
        • 8 years ago

        The reason is because it isn’t financially viable for them to design a native 6 core SB-E just for the desktop market. So it’s a cut-down Xeon BUT with higher clock speeds which makes sense for the intended market.

          • kroker
          • 8 years ago

          Yes but the original question was why is it cut-down, since this is a premium desktop part. So, it has 6 cores in order to allow higher clocks in the 130W TDP?

            • smilingcrow
            • 8 years ago

            Yes.

            • OneArmedScissor
            • 8 years ago

            That’s not a very good explanation when there’s such extensive power gating and turbo boosting. It’s about price and the fact that desktops do not dictate what the high end is.

            • NeelyCam
            • 8 years ago

            Are you sure you understand what TDP is, and how that translates to continuous computing performance (as opposed to short-term, turbo-enhanced performance)?

            • OneArmedScissor
            • 8 years ago

            Are you sure you understand what turbo boost is? They could just leave the other cores there, with probably the exact same clock speeds, but with the 8 core mode running at a lower clock and voltage.

            They’re already doing one better, leaving the full 20MB cache, as well:

            [url<]https://techreport.com/discussions.x/21557[/url<]

            • NeelyCam
            • 8 years ago

            Yes, I understand what turbo boost is. Do you understand what TDP means?

            A 6-core chip designed for 130W TDP can turbo longer and higher than a 6-core chip designed for 95W TDP. Turbo is great for short boosts and low thread counts, but when all cores are running full blast, higher TDP enables higher performance.

            • OneArmedScissor
            • 8 years ago

            Epic reading comprehension fail.

            • NeelyCam
            • 8 years ago

            Agreed… my bad.

            I blame caffeine deficiency.

            • kamikaziechameleon
            • 8 years ago

            I was reading this and about 4 posts ago I’m wondering what is NeelyCam getting at. πŸ˜›

            This is clearly a power play by intel because of lack of competition and to leave headroom for performance gains on the next fab revision.

            There are a dozen reasons why 2 cores are disabled and non of them show respect for their consumer.

        • OneArmedScissor
        • 8 years ago

        [quote<] you pay premium price for a Xeon reject.[/quote<] No, you pay the same price as the actual Xeon equivalent. You think the 8 core, full speed Xeon is going to cost $1,000? Maybe multiply that number a few times and you'll be in the ball park.

        • Shrikered
        • 8 years ago

        Yes. This is a huge die (21 mm on a side) and undoubtedly has low early yields. By locking out 25% of the area, yields go way up — far more than proportionally,since you use fuses to lock out whichever cores/cache turn out to be bad. E.g. 60% yield would go up to 80% yield.

        In many cases 7 or even 8 of the cores may work fine, if you know how to get at them.

        • Mr Bill
        • 8 years ago

        I wonder why AMD did not do the same thing with its 8-module 6200 series. Then they might have had a horse in the race.

      • Damage
      • 8 years ago

      Top of page two:

      [quote<]As we've noted, the 3960X has portions of the die disabled, namely two of the CPU cores and 5MB of L3 cache. Releasing a top-of-the-line desktop chip with portions disabled runs contrary to our expectations, but we think Intel had some cogent reasons for making this choice. That is, in order to reach the same 3.9GHz peak Turbo Boost frequency as the top Sandy Bridge quad-core product while staying within the 130W power limit, a couple of cores had to be sacrificed. For a desktop-oriented product where performance in lightly threaded applications remains very important, we can see the logic of this tradeoff. After all, it wouldn't do for the 3960X to be measurably slower than the Core i7-2700K in a number of applications.[/quote<]

        • TREE
        • 8 years ago

        I can see why you think Intel made this choice, however I reckon they rather would have taken a cue from AMD and allowed for the user to manually activate the extra cores and cache within the motherboard BIOS. It would make sense to allow this as the Sandy-E chip is more targeted towards high end users, whom could probably benefit from the disabled units.

        For Intel to assume that it’s targeted customer, an enthusiast or someone who needs this kind of computational power, doesn’t know how to configure a BIOS, is rather short sighted and unlikely. I have a hunch that it could be a yield issue… It has taken a while for this chips release afterall.

          • Krogoth
          • 8 years ago

          Intel did it because, “Desktop users” don’t need 8-cores, (IMO they only need quad-cores in a few case, while they can get by with dual-cores). They need the MEGAHURTZ, since clock speed is still king in the desktop world. An full 8-core i7 can’t hit the same clockspeed as the normal Sandy Bridges without becoming a blast furnace. You also got the aforementioned yielding issues from making such a huge piece of silicon.

          They are selling their second-rated “Sandy-Bridge-E” yields to early adopter, epenis types who don’t know about better.

            • shank15217
            • 8 years ago

            With full 8 cores enabled the processor wouldn’t be able to hit the clock speeds required to compete with even it’s own i7 2700K in many desktop tasks.

            • kamikaziechameleon
            • 8 years ago

            look above at the Discussion between OneArmedScissor and NeelyCam. Using power gating with the turbo boost they could have made it work just fine.

        • kamikaziechameleon
        • 8 years ago

        using all of intels other existing tech I’m sure they could have rigged power gating and turbo boost in such a way as to keep it under TDP and atain the same clock speed and efficiency easily for single or dual threaded applications.

    • Pez
    • 8 years ago

    People slated the BD for having such high power draw overclocked yet this is praised for being much higher?

      • Arclight
      • 8 years ago

      If FX 8150 OCed consumed more than this chip and yet it performed the same or slightly better, hell even slightly worse nobody would have said anything, methinks. But look at the tests man, the 8150 is nowhere near to this beast…….

        • Pez
        • 8 years ago

        True and wasn’t insinuating it was – just seemed strange.

        • travbrad
        • 8 years ago

        In addition to the performance, this is also aimed at the ultra-high-end enthusiast, as evidenced by the $1000 price tag (and $300 mobos). That kind of consumer is a lot more likely to have a 1000W+ PSU than a more “mainstream” consumer who the 8150 is priced for.

        Obviously for typical gamers/users a 2500K/2600K have a vastly higher perf/price, but some people just want the best of the best, or do heavy video encoding/rendering/etc.

      • smilingcrow
      • 8 years ago

      The Anandtech review shows a very different story when over-clocked:

      Intel 4.6GHz 320W
      AMD 4.8GHz 406W

      It might that TR had to push the voltage too high to obtain the extra 200MHz at 4.8GHz which is why the power figure look so high. Those last few hertz can often really hurt the power consumption.

        • spuppy
        • 8 years ago

        In my experience with this beast, it seems to draw more power, the hotter it gets. On liquid, I was able to reach 4.6 GHz, and it was drawing 385 watts at load, at about 77-79C

        On air, I could reach 4.5 GHz stable, but temperatures reached 82-85C. At that level, it is drawing 411 watts

          • OneArmedScissor
          • 8 years ago

          More heat means more friction and reduced efficiency.

          Thanks for sharing what you found. People need to keep that in mind as yet another reason why you can’t compare multiple power figures apples to apples.

            • spuppy
            • 8 years ago

            I am playing around with Hyperthreading disabled, and using the more beefy AMD FX watercooler right now. So far I’m up to 4.9 GHz.. You can follow my progress at @hardcoreware if ya like

            • Waco
            • 8 years ago

            It makes a bigger difference than I ever expected. Going with a full water loop on my rig dropped power consumption by more than 100 watts…and that’s including the extra power used for my pump and fans. Granted, mine is an extreme case (because of my components), but cooler components use less power. See the K computer for reference. πŸ˜›

      • maroon1
      • 8 years ago

      Sandy-bridge-E has much better performance per watt than Bulldozer

      • Krogoth
      • 8 years ago

      It has the performance to back the heavy power consumption.

    • r00t61
    • 8 years ago

    Pretty underwhelming, especially given the significant delay in bringing to market.

    Neutered cores, neutered chipset, and Intel can’t even be bothered to include a cooler for the $1,000 you’re shelling out. Yeah, you get 10-15% improvement over previous Gulftown, but that’s from the base SB architecture change. No special SB-E sauce. Those two missing cores are really annoying. Like Intel giving you a big middle finger for spending a grand. “Here’s your bonus. Two of your cores are disabled!”

    For like 99.9% of users, even enthusiast users, no point in getting SB-E over a vanilla 2500k. Save that $700 or buy a bunch of video cards.

    Intel can only get away with a release like this because normal SB kicks so much ass and AMD can’t seem to launch a competitive mid-to-high-end chip to save their lives.

    Bring on Ivy Bridge.

      • Krogoth
      • 8 years ago

      Intel is trying to milk the early adopter, epenis types to be beta-testers for their next generation server/workstation platforms.

      It was the same case back when Bloomfield and Gulftown came out.

        • Bauxite
        • 8 years ago

        Bloomfield was a better buy if not sometimes cheaper than 1156 builds bought later (thanks micro center deals and late 09 ram spike) and came out earlier instead of vice versa. Or are you trying to say nehalem wasn’t a real improvement over core2?

          • Krogoth
          • 8 years ago

          For a while, Bloomfield was the premium option. The chips in it commanded a higher premium, the motherboards were more pricy. The only two chips that are “affordable” were the i7-920 and i7-930. At stock, the chips weren’t much faster than their Core 2 predecessor. The chips just scaled better when you OC’ed them and you needed to OC’ed to them in order to get a comfortable lead.

      • sschaem
      • 8 years ago

      If I have a choice. i7-3820 or i7-2600k, I would go with the cheaper and faster I7-3820.
      (So far Intel seem to have priced the 4core SB-E under $300)

      Ok, you loose the SB GPU, but you gain quad channel DDR3 1600, faster base and turbo clock and larger L3 cache, and bonus, possibly PCI 3.0. A trade I will make anydays, specially when the SB-E model is 10% cheaper.

        • Krogoth
        • 8 years ago

        i7-3820 is currently MIA and with no “official” pricing yet.

        I suspect it will be price a little higher than i7-2700 and demand may push it higher. FYI, most X79 boards cost more than their P65/X65 counterparts. IF you want to deck it out, you need to get four-DIMMs. Although, memory costs isn’t that bad these days.

      • NeelyCam
      • 8 years ago

      [quote<]"Here's your bonus. Two of your cores are disabled!"[/quote<] They are not your cores unless you paid for them. You paid for 6 cores - you got six cores. I don't understand how some people have such hard time comprehending this.

        • flip-mode
        • 8 years ago

        Where’s the 8 core option?

          • NeelyCam
          • 8 years ago

          Coming later (Xeon)

            • flip-mode
            • 8 years ago

            So there will be no fully enabled Extreme Edition?

            • Krogoth
            • 8 years ago

            It will happen eventually when Intel get better yields on their 8-core SB-Es. I suspect it will be the next re-fresh.

      • shank15217
      • 8 years ago

      Ivy Bridge will bring like a 5-8% increase in performance per clock. A lot of the low hanging fruits have already been picked.

      • esterhasz
      • 8 years ago

      One thing that’s great though: the ability to get to 32GB on the cheap.

    • marvelous
    • 8 years ago

    meh

    • chuckula
    • 8 years ago

    It’s too bad they didn’t get a 3930k to play with too. I thinkg that chip will overclock better than the already awesome 3960x due to the smaller l3 cache. At under $600 it’s actually reasonably priced for the performance you can get out of it.

      • mczak
      • 8 years ago

      I’d be VERY surprised if the difference in cache size would make any noticeable difference for OC. It should however likely reach the same clocks when OC, hence you could say it overclocks better because it starts out 100Mhz lower.
      (FWIW I think the 3960X actually has all l3 cache activated it can – looking at the xeon lineup I think it is not possible that the l3 cache tied to deactivated cores can be used – so 2.5MB per core is the maximum on this chip. It is however possible to deactivate more cache which apparently the 3930k does with “only” 2MB per core.)

        • derFunkenstein
        • 8 years ago

        I thought the same thing relating to the cache at first, but then if it’s really “shared” is any of the cache really “tied” to the cores? Maybe it is, and I just don’t get it…

          • mczak
          • 8 years ago

          The L3 cache is shared, but still each core “owns” a chunk of it (for that matter the latency isn’t quite fixed neither). The cores (and cache slices) are connected via a ring bus. There was also some talk that the L3 cache runs at the frequency of the respective core and hence if the core is downclocked it would be slower, I think in light of that it makes sense the l3 cache cannot work without the core tied to it.

    • chuckula
    • 8 years ago
    • chuckula
    • 8 years ago
    • RtFusion
    • 8 years ago

    I am reading TR’s and HardOCP’s reviews and it really just cements the 2500-K as THE CPU to get for a new build.

    I mean, the new CPU from Intel is fast but not really worth for a majority of us. Unless you want to break some OC or benchmark records and/or have a lot of cash lying around the house.

      • Airmantharp
      • 8 years ago

      Overclocking considered, I agree there is very little reason to consider anything but the 2500k. It’s the least expensive CPU that can be pushed to ~4.8GHz with ease. So unless you need as much raw CPU power as you can get, it makes very little sense to spend more, especially for gaming.

      • AssBall
      • 8 years ago

      Fair enough, but this shit is bananas. B A N A N A S. 6 cores or two at 4.8ghz and the memory bandwidth is nutsack.

      I guess they call it an enthusiast proc. for a reason.

    • Jigar
    • 8 years ago

    Impressive beast, but some how i am still not convinced to replace my Q6600 which is Oced at 3.7GHZ. I think i’ll see substantial performance gain once Ivy bridge comes in, till than Q6600 will cruise without any issue.

      • flip-mode
      • 8 years ago

      Er, I think you’ve had the option for substantial performance gain for some time now. A 2600K would substantially outperform and out-clock you Q6600 and so would some processors from the generation before the 2600K. I can only imagine what’s really holding you up is that you don’t really need more than what you have.

        • Jigar
        • 8 years ago

        Errr.. No, i already have made tons of systems based on 2600K and 2500K and no when C2Q is Oced at around 3.7GHZ you don’t get to see substantial performance gain… Yes 2600K @ 5 GHZ would blast anything in benchmarks, but for daily fooling around, no it is not a justified upgrade against 3.7GHZ C2Q.

          • NeelyCam
          • 8 years ago

          I mostly agree with you, but why do you think Ivy Bridge would give you substantially better performance than SB or SB-E?

            • travbrad
            • 8 years ago

            Indeed. From the rumors I’ve heard it sounds like IB is more focused on power efficiency than pure performance. It’ll certainly be faster than SB, but I don’t think the difference will be massive (apart from the GPU).

          • shank15217
          • 8 years ago

          Yea whatever, 2600K at nearly 30% the ipc boost over Q6600 and a 40% power reduction all you’re getting out of an overclocked processor 2 process generations older is the right to say you have an over-clocked processor. Also the Q6600 isn’t that fast at 3.7 GHz. I am sure even a i3 2100 can beat it in several benchmarks.

            • Krogoth
            • 8 years ago

            You are giving the i3 2100 a bit too much credit. It is fast, but it only has two real cores with HT. I doubt it catch-up to the OC’ed Q6600 in multi-threaded applications.

            The big difference is that the i3-2100 doesn’t need a huge HSF with a loud fan to achieve that level of performance.

            • sweatshopking
            • 8 years ago

            I can’t get my q6600 past 2.8ghz. seems to crash if i go higher. still boots, but everytime you do anything it laggggggggggggggsssssss. it still does what i need, but it’s not 3.7ghz.

            • canoli
            • 8 years ago

            imo it’s the chipset that’s buggered this launch. It really has to mature before I lay out my hard-earned dough; hopefully it’ll be IB.

            2 measly SATA III ports? Still no native USB 3.0? And Intel really shouldn’t be commended for being unable to certify PCIe 3.0. I don’t say slam them for it but let’s not pretend they’re being noble.

            The conclusion of the article hits home for me. This launch feels rushed, not-ready-for-prime-time. The disabled cores – I understand the 130w limit, that’s fine. On the other hand those dead cores illustrate the problem perfectly with this offering. Maybe Intel figured we were losing patience with all the hype about SB E! So they shoved it to the altar married to a pretty but under-whelming chipset.

            Does it matter two cores are disabled, as in a “we got screwed” sense? Of course not. It’s perfectly reasonable they’re non-functioning [i<]if[/i<] the market had been crying out for a new high-end processor. It isn't. X79 has a few steps on X58, no question. Not enough for me but I admit I lust after native SATA III. But how many steps does it have on Z68? Any? You lose the SSD caching, you don't gain any additional SATA III ports. Like others have said, for gaming there are already very similar-performing processors for the same (or less) money. For video-editing, 3D modeling and heavy image-processing I'll stick with my Gulftown.

            • flip-mode
            • 8 years ago

            Your post is great – speaks to the true concerns of the enthusiast. Too much attention is being paid to the disabled cores. Way too much really. The platform really should be better. And it will be, I’m sure, in time. It really does feel prematurely launched. If Intel would have held off longer it may have been able to launch both a better platform and the silicon with all cores enabled.

            • gmskking
            • 8 years ago

            Could be your power supply, might not be enough wattage. Most likely the motherboard though. I can get mine to 3.6GHz with a 750 chipset. Something not right with your setup.

          • smilingcrow
          • 8 years ago

          If you don’t notice a significant difference it sounds as if you aren’t running very intensive applications so no point in upgrading really.

      • canmnanone
      • 8 years ago

      i agree man. i too still have the q6600 oc to 3.6gig that runs all my apps and games very well. especially when games re-use the same old game engine from the past. but bf3 really does need more processing power from a high system. but thats the only one game that makes upgrading worthwhile. so for now will stick to my old system.

      • derFunkenstein
      • 8 years ago

      Your CPU is similar in performance to a Phenom II X4 980, and it’s getting left way behind. I think you could easily upgrade today and see a nice improvement.

        • Krogoth
        • 8 years ago

        I doubt it, unless you are running workstation applications or just want to reduce power consumption.

        i7-2600K at stock barely outruns the heavily OC’ed Q6600, however it doesn’t need a huge heatsink with loud fans to keep it cool. πŸ˜‰

      • OneArmedScissor
      • 8 years ago

      [quote<]I think i'll see substantial performance gain once Ivy bridge comes in[/quote<] Think again, unless you want to cut power use a little bit, but mostly just because you've overclocked yours. You're setting yourself up for disappointment. There was never any suggestion Ivy Bridge would have more cache or cores or something of that nature. Intel has said over and over that it's extremely similar to Sandy Bridge and something in the neighborhood of 5% faster per clock, which is going to be best case scenario, and they're unlikely to raise the clock speed any higher than another 100 MHz or so for the desktop parts because they wouldn't dare sell one that's potentially as fast or faster than Sandy Bridge E.

      • d0g_p00p
      • 8 years ago

      Your CPU seems to run at different OC’d speeds everytime you defend your old CPU. I’m just sayin’

        • Jigar
        • 8 years ago

        Very correct, i usually keep my CPU hanging at around 3.2GHZ, and my OC changes as per the game or application requirement, the max i have done is 3.7GHZ stable.

          • d0g_p00p
          • 8 years ago

          I just find that hard to believe since like i said you seem to change how fast your CPU is overclocked every time you say that your ancient CPU is fine and current CPU’s offer no benefit. So you tell me that your CPU is stable at 3.7GHz and you change the OC settings based on your needs? That makes absolutely no sense at all. If it runs stable at 3.7GHz why not keep it at that speed? I call shenanigans you even indicate in this very thread that you run your CPU at 3.7Ghz you make no word that it’s not what you run it at 24/7 and only fall back in this response when I ask you about it. Also I would like to know what you used to clock you Q6600 at 3.7Ghz since most of the elite overclockers cannot bust 3.3 on air and on hit in the 3.5Ghz range on water.

          I’m not saying that you are lying but 3.7Ghz on a Q6600 requires FCC and even that does not run 24/7. If you indeed have this magic Q6600 you should trade or sell it on xtremesystems because someone will pay top dollar for this chip and you can get the system you have been wanting.

            • Kaleid
            • 8 years ago

            It makes sense. I can o/c my i3 530 to 4.2Ghz but I have no need for that speed when I’m surfing the web and writing, 3Ghz will do fine. If I start to game I simply load an o/c profile from bios to enjoy better speed.

            • d0g_p00p
            • 8 years ago

            I can see something like that if you could load profiles from Windows. Maybe it’s me but I would not want to reboot my computer to change the speed of my CPU. I also don’t care what speed my CPU is overclocked at as long as it’s quiet. Too each his own I guess.

            • kamikaziechameleon
            • 8 years ago

            ^^^+1 this,

            Had a Q6600 and everything he says makes no sense to me.

            @Kaleid, I would point out your cpu isn’t always running at the OC run cpu z and watch how it scales the OC setting is your ceiling for 100 percent utilization. I OC my 1090T x6 to 3.9 ghz on air and its not noisy(from fan noise) unless I’m playing a demanding game like BF3 and the CPU utilization across all cores is floating between 80 and 100 percent.

            • Waco
            • 8 years ago

            Agreed on this. The power draw to get a Q6600 that high is IMMENSE. I literally had my DFI board burst into flames trying to get just 3.6 GHz stable…and that was with a full watercooling loop to cool the veritable inferno that is a Kentsfield chip at anything over 3.4 GHz or so.

            So yeah, shenanigans. It has been called.

      • jounin
      • 8 years ago

      the performance gain is there… i just upgraded from a q6600 at 3.2ghz and i gain over 20fps avg in f1 2011… with a 2500k at 4.2ghz

        • Jigar
        • 8 years ago

        So you got 20FPS on 1 GHZ and i should get 10FPS on 500MHZ.. I think i’ll pass.

          • jounin
          • 8 years ago

          im sure it wont work like that since the clock per clock peformance is way better on the 2500k than the q6600 i can underclock it to 3.2 and see what i get

          • derFunkenstein
          • 8 years ago

          I don’t think your math is working.

      • Duck
      • 8 years ago

      Hate all these I “have a Q6600” posts.

      I get by with a 2.1GHz Athlon X2.

      • kamikaziechameleon
      • 8 years ago

      Dude Upgraded to a crap 1090T x6 last year and it was such an amazing improvement in preformance. Just sayin if you game you could do with a new processor. For regular websurfing etc you are still golden.

    • Vasilyfav
    • 8 years ago

    An absolute beast of a CPU. I’m torn between building a new 3930K system and waiting for Ivy Bridge.

    This truly seems to be a very future proof CPU, especially for gaming and 30 fps 1080p x264 streaming.

    Also, I admire your choice of the Canon t3i for taking the pictures of the CPUs for socket size comparison πŸ™‚

      • DeadOfKnight
      • 8 years ago

      I would definitely wait for a quad-core Ivy Bridge.

        • swampfox
        • 8 years ago

        This.

        Ivy Bridge should improve performance and/or decrease power over quad-core SB, and cost half as much (my guess) as the 3930K.

        • PainIs4ThaWeak1
        • 8 years ago

        If we ALL waited for the “next-gen”, then we’d either A. Still be using legacy hardware, B. Hardware manufacturers wouldn’t exist as they do today, or C. Both.

        Why not start telling people to “upgrade if its worth it to YOU” ?

          • DeadOfKnight
          • 8 years ago

          Then don’t wait for Ivy, I’m just saying that Ivy will definitely offer more bang for your buck than this thing. He’s obviously a gamer, so really the 2500K would be a better investment if he wants to buy now. If he’s truly torn between this and Ivy though, Ivy is the way to go.

      • Arclight
      • 8 years ago

      If you got the money don’t play the waiting game, this CPU is certainly future proof. It reminds of the first generation i7s, which although more power hungry can keep up quite easily with Sandy Bridge with a little overclocking.

        • Airmantharp
        • 8 years ago

        …and with a little overclocking, Sandy Bridge will beat the pasts off of it’s ancestors.

          • Arclight
          • 8 years ago

          At gaming, maybe, and i doubt it will be that drastic.

Pin It on Pinterest

Share This