AMD’s Phenom II X4 965 Black Edition processor


Since the introduction of its Phenom II processors at the beginning of this year, AMD has been, uh, chipping away at Intel’s performance lead. (Yeah, I wrote that.)

Well, that’s sort of true, at least. The Phenom II is a fine processor and very much a match for Intel’s Core 2 chips, but it’s not really up to the challenge of going toe to toe with the Core i7. Instead, AMD has concentrated on matching up against the Core 2 processors in the middle of Intel’s lineup. Thanks to aggressive pricing and some savvy moves, like selling a relatively affordable series of unlocked Black Edition processors, the Phenom II has had some success. In fact, AMD captured three of five spots in our most recent system guide‘s main configs.

That success is threatened by the imminent arrival of mainstream Core i7 derviatives, code-named Lynnfield, in the next month or so. Before that happens, AMD has decided to position itself a little bit better by introducing a new speed grade at the top of its Phenom II lineup—the reason we are gathered here today—and dropping its prices on some lower speed grades.

Thus we have the Phenom II X4 965 Black Edition, a processor that could compute Pi to an insane number of digits during the time it takes any one of us to pronounce its full name, in part because it runs at a rather high-strung clock rate of 3.4GHz. But mostly because of the name. To reach this lofty frequency, AMD has pushed into heady thermal territory; the X4 965 has a TDP rating of 140W, up a tad from the 125W rating of the X4 955 that preceded it. That’s not unfamiliar—some original Phenoms stretched the thermal envelope to 140W, as well—but it is at the limits of what we’d expect on the desktop.

Although the X4 965’s power requirements are on the extreme side, its pricing and performance are not. AMD tells us the 965 will list for $245, the same price rung that the 955 Black Edition occupied previously. That puts the 965 in more or less direct competition with the Core 2 Quad Q9550, which lists at $266 but is selling for as little as $220 at major online vendors like Newegg.

Intel still has the upper hand here in a couple of respects. Thanks to its higher per-clock performance, for instance, the Q9550 can compete with the X4 965 while running at only 2.83GHz. And at that lower clock speed, the Q9550 fits into a 95W thermal envelope. In fact, Intel even makes a low-power variant called the Q9550S that has a 65W TDP. (Of course, comparing Intel TDP ratings to AMD TDP ratings always incites a lovely controversy, so we’ll avoid putting too much stock in them and point you to our power consumption testing, instead.)

To make the Phenom II X4 965 a little more attractive versus the competition, AMD has uncorked a jug of mad marketing mojo and sloshed it around in various directions. Some has splashed onto major online stores, where AMD says you can expect to see CPU and motherboard discount combos that will save you up to 40 bucks versus full price—and possibly up to twice that with Radeon graphics cards and Corsair DIMMs added to the mix. A few ounces landed on AMD’s graphics driver team; they’re preparing a new Catalyst 9.8 driver revision that promises performance benefits for CrossFireX multi-GPU rigs on AMD chipsets, an illustration of the purported platform-level synergy the Phenom II has working for it. Along those same lines, the folks responsible for AMD’s OverDrive tuning utility have delivered a new round of profiles for popular games (and benchmarks) that will allow the utility to boost clock speeds on one or two cores in lightly multithreaded applications.

When I described this feature to one of my fellow editors, he called it a poor man’s version of the Core i7’s Turbo Boost, but I prefer to think of it as the oppressed, chain-smoking man’s version of Turbo Boost, for reasons that will become clear once we delve into it.

There’s also the unlocked upper multiplier to ease the work of overclocking the 965. And I believe there may be a giant, green, inflatable gorilla on the roof at Newegg right now, framed by several lines of plastic flags, but it’s hard to see from where I sit.

We’ll let you work out what exactly these measures do for the X4 965’s value proposition, but we can illustrate how this new processor compares to a whole host of competitors, siblings, and distant cousins thanks to our considerable cache of CPU performance data. In fact, this may be the last hurrah for our latest and rather extensive Vista x64-based set of performance results. Windows 7 is imminent, and I can feel a refresh coming on.

Test notes

In order to gauge the impact of memory type on performance and power use, we’ve tested the Phenom II X4 810 both with DDR2 memory on a Socket AM2+ board and with DDR3 memory on a Socket AM3 board. You’ll find the results in the following pages, labeled appropriately.

The Core 2 Quad Q8300 processor we used for testing came to us courtesy of the good folks at NCIX and NCIXUS. Thanks to them for making this comparison possible. We’ve underclocked our Q8300 to simulate a Q8200 for this review.

We’ve simulated several other speed grades via underclocking, too. Specifically, the Phenom II X4 920 is an underclocked 940, and the Core 2 Quad Q9550 is an underclocked Core 2 Extreme QX9650. We expect the performance of these “simulated” speed grades to be identical to the real things, but we sometimes exclude these processors from our power consumption testing because we do anticipate power use could vary slightly from the actual products.

Our testing methods

As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and the results were averaged.

Our test systems were configured like so:

Processor Core
2 Quad Q6600

2.4 GHz

Core
2 Duo E8400

3.00 GHz

Core
2 Duo E8600

3.33 GHz

Core 2 Quad Q8200 2.33 GHz

Core 2 Quad Q9300 2.5 GHz

Core 2 Quad Q9400 2.66 GHz

Core 2 Quad Q9550 2.83 GHz

Core
2 Extreme QX9770 3.2 GHz
Dual
Core
2 Extreme QX9775 3.2 GHz
Core
i7-940 2.66 GHz

Core i7-940 2.93 GHz

Core
i7-965

Extreme 3.2 GHz

Core
i7-975

Extreme 3.33 GHz

Athlon
64 X2 6400+

3.2 GHz

Phenom
X3 8750

2.4 GHz


Phenom II X4 920

2.8 GHz

Phenom II X4 940

3.0 GHz

Phenom
II X4 810

2.6 GHz


Phenom X4 9950

Black 2.6 GHz

Phenom
II X3 720

2.8 GHz

Phenom II X4 810

2.6 GHz

Phenom
II X4 955

3.2 GHz

Pentium
E6300 2.8 GHz
Core
2 Quad Q8400 2.66 GHz
Athlon
II X2 250 3.0 GHz
Phenom
II X2 550 3.1GHz
Phenom
II X4 965 3.4GHz
System bus 1066
MT/s

(266 MHz)

1333
MT/s

(333 MHz)

1600
MT/s

(400 MHz)

1600
MT/s

(400 MHz)

QPI
4.8 GT/s

(2.4 GHz)

QPI
6.4 GT/s

(3.2 GHz)

QPI
6.4 GT/s

(3.2 GHz)

HT
2.0 GT/s

(1.0 GHz)

HT
3.6 GT/s (1.8 GHz)
HT
3.6 GT/s (1.8 GHz)
HT
4.0 GT/s (2.0 GHz)
HT
4.0 GT/s (2.0 GHz)
HT
4.0 GT/s (2.0 GHz)
Motherboard Asus
P5E3 Premium
Asus
P5E3 Premium
Asus
P5E3 Premium
Intel
D5400XS
Intel
DX58SO
Intel
DX58SO
Gigabyte
EX58-UD3R
Asus
M3A79-T Deluxe
Asus
M3A79-T Deluxe
MSI
DKA790GX Platinum
Asus
M4A79T Deluxe
BIOS revision 0605 0605 0605 XS54010J.86A.1149.

2008.0825.2339

SOX5810J.86A.2260.

2008.0918.1758

SOX5810J.86A.2260.

2008.0918.1758

F5 0403 0403 11/25/08 0703
1.6
(1/21/09)
0902
0802 0802 0089
(5/15/09)
1103
0066
(6/30/09)
North bridge X48
Express MCH
X48
Express MCH
X48
Express MCH
5400
MCH
X58
IOH
X58
IOH
X58
IOH
790FX 790FX 790GX 790FX
South bridge ICH9R ICH9R ICH9R 6321ESB ICH ICH10R ICH10R ICH10R SB750 SB750 SB750 SB750
Chipset drivers INF
Update 9.0.0.1008

Matrix Storage Manager 8.5.0.1032

INF
Update 9.0.0.1008

Matrix Storage Manager 8.5.0.1032

INF
Update 9.0.0.1008

Matrix Storage Manager 8.5.0.1032

INF Update
9.0.0.1008

Matrix Storage Manager 8.5.0.1032

INF
update 9.1.0.1007

Matrix Storage Manager 8.5.0.1032

INF
update 9.1.0.1007

Matrix Storage Manager 8.5.0.1032

INF
update 9.1.0.1007

Matrix Storage Manager 8.5.0.1032

AHCI
controller 3.1.1540.61
AHCI
controller 3.1.1540.61
AHCI
controller 3.1.1540.61
AHCI
controller 3.1.1540.61
Memory size 4GB
(2 DIMMs)
4GB
(2 DIMMs)
4GB
(2 DIMMs)
4GB
(2 DIMMs)
6GB
(3 DIMMs)
6GB
(3 DIMMs)
6GB
(3 DIMMs)
4GB
(2 DIMMs)
4GB
(2 DIMMs)
4GB
(2 DIMMs)
4GB
(2 DIMMs)
Memory type Corsair
TW3X4G1800C8DF

DDR3 SDRAM

Corsair
TW3X4G1800C8DF

DDR3 SDRAM

Corsair
TW3X4G1800C8DF

DDR3 SDRAM

Micron
ECC DDR2-800

FB-DIMM

Corsair
TR3X6G1600C8D

DDR3 SDRAM

Corsair
TR3X6G1600C8D

DDR3 SDRAM

OCZ
OCZ3B2133LV2G

DDR3 SDRAM

Corsair
TWIN4X4096-8500C5DF

DDR2 SDRAM

Corsair
TWIN4X4096-8500C5DF

DDR2 SDRAM

Corsair
TWIN4X4096-8500C5DF

DDR2 SDRAM

Corsair
TW3X4G1600C9DHXNV

DDR3 SDRAM

Memory
speed (Effective)
1066
MHz
1333
MHz
1600
MHz
800
MHz
1066
MHz
1600
MHz
1600
MHz
800
MHz
1066
MHz
1066
MHz
1333
MHz
CAS latency (CL) 7 8 8 5 7 8 8 4 5 5 8
RAS to CAS delay (tRCD) 7 8 8 5 7 8 8 4 5 5 8
RAS precharge (tRP) 7 8 8 5 7 8 8 4 5 5 8
Cycle time (tRAS) 20 20 24 18 20 24 24 12 15 15 20
Command
rate
2T 2T 2T 2T 2T 1T 1T 2T 2T 2T 2T
Audio Integrated
ICH9R/AD1988B

with SoundMAX 6.10.2.6480 drivers

Integrated
ICH9R/AD1988B

with SoundMAX 6.10.2.6480 drivers

Integrated
ICH9R/AD1988B

with SoundMAX 6.10.2.6480 drivers

Integrated
6321ESB/STAC9274D5

with SigmaTel 6.10.5713.7 drivers

Integrated
ICH10R/ALC889

with Realtek 6.0.1.5704 drivers

Integrated
ICH10R/ALC889

with Realtek 6.0.1.5704 drivers

Integrated
ICH10R/ALC888

with Realtek 6.0.1.5704 drivers

Integrated
SB750/AD2000B

with SoundMAX 6.10.2.6480 drivers

Integrated
SB750/AD2000B

with SoundMAX 6.10.2.6480 drivers

Integrated

SB750/ALC888

with Realtek 6.0.1.5704 drivers

Integrated

SB750/ALC1200

with Realtek 6.0.1.5704 drivers

Hard drive WD Caviar SE16 320GB SATA
Graphics Radeon
HD 4870 512MB PCIe with Catalyst 8.55.4-081009a-070794E-ATI
drivers
OS Windows Vista Ultimate x64 Edition
OS updates Service
Pack 1, DirectX redist update August 2008

Thanks to Corsair and OCZ for providing us with memory for our testing.

Our single-socket test systems were powered by OCZ GameXStream 700W power supply units. The dual-socket system was powered by a PC Power & Cooling Turbo-Cool 1KW-SR power supply. Thanks to OCZ for providing these units for our use in testing.

Also, the folks at NCIXUS.com hooked us up with a nice deal on the WD Caviar SE16 drives used in our test rigs. NCIX now sells to U.S. customers, so check them out.

The test systems’ Windows desktops were set at 1600×1200 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled.

We used the following versions of our test applications:

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

Before we get into application performance, here’s a quick look at the memory subsystems of the X4 965 and its competition. The graph above shows the bandwidth delivered at each stage of the memory hierarchy. Superficially, the X4 965 isn’t too different from its closest rival, the Core 2 Quad Q9550. The Q9550’s L1 cache tends to deliver higher bandwidth at block sizes below 256KB, but the Phenom II’s cache hierarchy achieves higher throughput at intermediate block sizes—and its integrated memory controller delivers the goods at the largest block sizes, where main memory bandwidth comes into play.

Here’s a closer look at the largest block size. The X4 965’s dual channels of DDR3 memory perform nicely, while the Q9550’s bandwidth is limited by its front-side bus. Then again, Intel has incorporated a triple-channel memory controller into the Core i7, and wow.

Integrated memory controllers reduce the latency caused by chip-to-chip communication, and so the Phenoms and Core i7s occupy the top rungs here. The Phenom II X4 965 slots in right where one would expect.

Crysis Warhead

We measured Warhead performance using the FRAPS frame-rate recording tool and playing over the same 60-second section of the game five times on each processor. This method has the advantage of simulating real gameplay quite closely, but it comes at the expense of precise repeatability. We believe five sample sessions are sufficient to get reasonably consistent results. In addition to average frame rates, we’ve included the low frame rates, because those tend to reflect the user experience in performance-critical situations. In order to diminish the effect of outliers, we’ve reported the median of the five low frame rates we encountered.

We tested at relatively modest graphics settings, 1024×768 resolution with the game’s “Mainstream” quality settings, because we didn’t want our graphics card to be the performance-limiting factor. This is, after all, a CPU test.

A small clock speed bump for the Phenom II produces a small performance bump, as well, putting the X4 965 right in league with the Core 2 Quad Q9550.

Far Cry 2

After playing around with Far Cry 2, I decided to test it a little bit differently by recording frame rates during the jeep ride sequence at the very beginning of the game. I found that frame rates during this sequence were generally similar to those when running around elsewhere in the game, and after all, playing Far Cry 2 involves quite a bit of driving around. Since this sequence was repeatable, I just captured results from three 90-second sessions.

Again, I didn’t want the graphics card to be our primary performance constraint, so although I tested at fairly high visual quality levels, I used a relatively low 1024×768 display resolution and DirectX 9.

Not only does the X4 965 basically match the Q9550 here, but both chips outperform two flavors of the Core i7, including the 940. We’re only talking about a few frames per second, though, and notice that the Core i7 has a higher minimum frame rate, which should mean smoother-feeling gameplay.

Unreal Tournament 3

As you saw on the preceding page, I did manage to find a couple of CPU-limited games to use in testing. I decided to try to concoct another interesting scenario by setting up a 24-player CTF game on UT3’s epic Facing Worlds map, in which I was the only human player. The rest? Bots controlled by the CPU. I racked up frags like mad while capturing five 60-second gameplay sessions for each processor.

Oh, and the screen resolution was set to 1280×1024 for testing, with UT3’s default quality options and “framerate smoothing” disabled.

Yep, the X4 965 finished behind a couple of other Phenom IIs, not because it’s really slower, but because there’s some undeniable variability in this method of testing. What we can say with confidence is that the X4 965 performs similarly to the Q9550 and the Core i7-920. This is one game, by the way, where the dual-core processors tend to cluster in the lower third of the performance results. Yes, more cores can bring you more performance in some games; it’s just a rare event right now.

Half Life 2: Episode Two

Our next test is a good, old custom-recorded in-game timedemo, precisely repeatable.

The X4 965 matches the 3.2GHz Core 2 Extreme QX9770 here and is well ahead of the Q9550—for what it’s worth, that is. All of these processors can run this game at more-than-acceptable frame rates, obviously.

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.

The Q9550 takes it by a hair.

WorldBench

WorldBench’s overall score is a pretty decent indication of general-use performance for desktop computers. This benchmark uses scripting to step through a series of tasks in common Windows applications and then produces an overall score for comparison. WorldBench also records individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results alongside the results from some of our own application tests.

AMD has chosen its target well. The X4 965 exactly matches the Q9550 in WorldBench’s composite score.

Oh, and you may be wondering how the Core i7-975 Extreme finished behind the Core i7-965. Simple: for whatever reason, our new Core i7 test rig didn’t play well with the Nero 7 Ultra test. The application would hang on each attempt through the test. As a result, we couldn’t include results for Nero in the WorldBench composite score for the 975 Extreme, which held it back. The Nero test has long been a sticking point in the WorldBench suite, although it usually runs successfully for us after multiple tries.

Productivity and general use software

MS Office productivity

Firefox web browsing

Multitasking – Firefox and Windows Media Encoder

WinZip file compression

Nero CD authoring

The new Phenom II pretty much creams the Q9550 in the Office, Firefox, and Firefox/Media Encoder multitasking tests, finishing the job 20 to 30 seconds sooner in each case. The Q9550 makes up that lost ground in the WinZip and Nero tests. The Nero results are stratified by disk controller type, you’ll notice.

Image processing

Photoshop

I believe disk I/O performance is also a bit of a factor in WorldBench’s Photoshop test, and it’s not pretty for the AMD chips. The X4 965 does well to break into the ranks of Intel processors, at least.

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. So this time around, 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 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 MMX, SSE2, and Hyper-Threading. Naturally, he’s ported picCOLOR to 64 bits, so we can test performance with the x86-64 ISA. Many of the individual functions that make up the test are multithreaded.

Our two remaining image manipulation tests show us nearly the same thing: that the X4 965 is just slightly faster than the Q9550 when doing work of this nature.

Media encoding and editing

x264 HD benchmark

This benchmark tests performance with 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. These scores come from the newer, faster version 0.59.819 of the x264 executable.

Windows Media Encoder x64 Edition video encoding

Windows Media Encoder is one of the few popular video encoding tools that uses four threads to take advantage of quad-core systems, and it comes in a 64-bit version. Unfortunately, it doesn’t appear to use more than four threads, even on an eight-core system. For this test, I asked Windows Media Encoder to transcode a 153MB 1080-line widescreen video into a 720-line WMV using its built-in DVD/Hardware profile. Because the default “High definition quality audio” codec threw some errors in Windows Vista, I instead used the “Multichannel audio” codec. Both audio codecs have a variable bitrate peak of 192Kbps.

Windows Media Encoder video encoding

Roxio VideoWave Movie Creator

Across all of our pure video encoding tests, the X4 965 is a little quicker than the Q9550. WorldBench’s VideoWave test emphasizes various video editing and batch file manipulation tasks, in addition to encoding, and the Q9550 has the edge in it.

LAME MT audio encoding

LAME MT is a multithreaded version of the LAME MP3 encoder. LAME MT was created as a demonstration of the benefits of multithreading specifically on a Hyper-Threaded CPU like the Pentium 4. Of course, multithreading works even better on multi-core processors. You can download a paper (in Word format) describing the programming effort.

Rather than run multiple parallel threads, LAME MT runs the MP3 encoder’s psycho-acoustic analysis function on a separate thread from the rest of the encoder using simple linear pipelining. That is, the psycho-acoustic analysis happens one frame ahead of everything else, and its results are buffered for later use by the second thread. That means this test won’t really use more than two CPU cores.

We have results for two different 64-bit versions of LAME MT from different compilers, one from Microsoft and one from Intel, doing two different types of encoding, variable bit rate and constant bit rate. We are encoding a massive 10-minute, 6-second 101MB WAV file here.

The ridiculously close contest between our contenders continues, with only a second separating them in each case.

3D modeling and rendering

Cinebench rendering

Graphics is a classic example of a computing problem that’s easily parallelizable, so it’s no surprise that we can exploit a multi-core processor with a 3D rendering app. Cinebench is the first of those we’ll try, a benchmark 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. Some of the beta 64-bit executables have been quite a bit slower than the 3.6 release, but this should give us a decent look at comparative performance, regardless.

3ds max modeling and rendering

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.

Wow, looks like our 3D rendering tests are a clean sweep for the Phenom II X4 965 over the Q9550. Combine that with the near-sweep of the video encoding tests, and the new Phenom II is beginning to look like it has a claim to better overall performance. But I’m getting ahead of myself.

[email protected]

Next, we have a slick little [email protected] benchmark CD created by notfred, one of the members of Team TR, our excellent Folding team. For the unfamiliar, [email protected] is a distributed computing project created by folks at Stanford University that investigates how proteins work in the human body, in an attempt to better understand diseases like Parkinson’s, Alzheimer’s, and cystic fibrosis. It’s a great way to use your PC’s spare CPU cycles to help advance medical research. I’d encourage you to visit our distributed computing forum and consider joining our team if you haven’t already joined one.

The [email protected] project uses a number of highly optimized routines to process different types of work units from Stanford’s research projects. The Gromacs core, for instance, uses SSE on Intel processors, 3DNow! on AMD processors, and Altivec on PowerPCs. Overall, [email protected] should be a great example of real-world scientific computing.

notfred’s Folding Benchmark CD tests the most common work unit types and estimates performance in terms of the points per day that a CPU could earn for a Folding team member. The CD itself is a bootable ISO. The CD boots into Linux, detects the system’s processors and Ethernet adapters, picks up an IP address, and downloads the latest versions of the Folding execution cores from Stanford. It then processes a sample work unit of each type.

On a system with two CPU cores, for instance, the CD spins off a Tinker WU on core 1 and an Amber WU on core 2. When either of those WUs are finished, the benchmark moves on to additional WU types, always keeping both cores occupied with some sort of calculation. Should the benchmark run out of new WUs to test, it simply processes another WU in order to prevent any of the cores from going idle as the others finish. Once all four of the WU types have been tested, the benchmark averages the points per day among them. That points-per-day average is then multiplied by the number of cores on the CPU in order to estimate the total number of points per day that CPU might achieve.

This may be a somewhat quirky method of estimating overall performance, but my sense is that it generally ought to work. We’ve discussed some potential reservations about how it works here, for those who are interested. I have included results for each of the individual WU types below, so you can see how the different CPUs perform on each.

The Phenom II’s dominance in the Amber and Tinker WU types gives it a sound overall victory in Folding performance. However, the two Gromacs types are arguably the most common, and the Q9550 handles those just as well as the X4 965.

Power consumption and efficiency

Our Extech 380803 power meter has the ability to log data, so we can 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. (We plugged the computer monitor into a separate outlet, though.) 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.

All of the systems had their power management features (such as SpeedStep and Cool’n’Quiet) enabled during these tests via Windows Vista’s “Balanced” power options profile.

I’ve only shown verbose results for some of the most relevant competitors here. You can see the results for the other processors in prior reviews.

We can slice up this 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.

TDP ratings deal with maximum power draw, but they don’t account for what happens when a CPU is unoccupied. Interestingly enough, the systems based on Phenom II quad-cores (including the X4 965) draw quite a bit less power at idle than our Q9550-based test system.

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, during which the processors were rendering.

The X4 965-based system draws roughly 20W more power at the wall socket than a similarly configured system with a Phenom II X4 955 processor. AMD must be at the edge of what it can get from its 45nm quad-core chips with the X4 965 at 3.4GHz. That said, the X4 965-based system draws only 15W more than the Q9550-based one. The gap between the Q9550- and X4 965-based systems is thus smaller than the processors’ TDP ratings alone suggest.

Another way to gauge power efficiency is to look 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.

We can quantify efficiency even better by considering specifically the amount of energy used to render the scene. 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.

By virtue of its lower system power draw at idle and its ability to finish the rendering task sooner, the Phenom II X4 965 fares better than the Q9550 in our two most important measures of energy efficiency. The bottom line here is that the X4 965 may require slightly better cooling to dissipate the heat it produces under load, but it can still be more power efficient overall than the Q9550.

Overclocking, Smart Profiles, and ceramic impellers

As I mentioned earlier, AMD’s OverDrive utility has a new Smart Profiles feature that, to some extent, mirrors the functionality of the Turbo Boost capability built into the Core i7. Smart Profiles are, in some ways, more like the game profiles built graphics drivers for multi-GPU schemes such as SLI and CrossFireX, because they’re based in software.

Here’s how it works: the OverDrive utility has a set of profiles for many of the most popular current games. Updates to the list are downloaded automatically once a week by the OverDrive utility when it’s first run. The profiles contain information about how many threads a game is likely to use and how the CPU should behave in it. Here’s an example, a profile for 3DMark05 taken from AMD’s website:

If Smart Profiles are enabled, several things happen automatically when the profiled application is launched. The OverDrive helper service sets the thread affinity for the application to cores 0 and 1. Those two cores are overclocked slightly, in this case by one full multiplier tick beyond the stock speed. Cool’n’Quiet dynamic clock speed scaling is disabled, as well, and the cores not used by the application are clocked down, core 2 by one tick and core 3 by two ticks. The profile also has settings for custom CPU voltage, process affinity, and whatever “Boost State” happens to be—sorry, not sure on that one.

As you might imagine, invoking a profile can provide a performance boost. AMD estimates those gains rather conservatively at 2-4%, depending on the application. Given the way it works, though, the Smart Profiles feature has some big limitations. For one thing, you’ll need a Black Edition processor in order to take advantage of the multiplier-based clock speed increases. Also, since this is technically overclocking, stability isn’t guaranteed. And this function won’t even work without a profile for the game or application you’re using. AMD maintains those profiles, determining what’s best on a per-application basis by while attempting to honor its CPU TDP targets and avoid instability.

Right now, profiles are very conservative; they don’t go beyond one tick up from the default multiplier, in part because profiles don’t include any processor-specific instructions and so have to be very generally applicable. AMD’s Sami Makinen told us the company is considering expanding the information included in profiles to provide CPU model-specific instructions that might grant it more leeway in cranking up the clocks. He also said the firm is considering including an option to enable Smart Profiles by default when the OverDrive application is installed. That could prove helpful because, right now, one must switch the utility into Advanced mode and then click a cryptic checkbox labeled “Application list” under the Smart Profiles tab in order to enable it.

Given all of their limitations, Smart Profiles in their present form aren’t likely to be much of a help for the Phenom II against what we expect to be a very aggressive Turbo Boost implementation in Intel’s Lynnfield processors. Still, this is an interesting feature that holds some potential as a tide-me-over until AMD can implement a Turbo Boost-like capability in future silicon. The fact that users can create their own profiles opens up some intriguing possibilities, as well.

Curious to see what sort of gains Smart Profiles could achieve, I enabled them and tested Far Cry 2. Check it out:

Yeah, so two frames per second won’t set the world on fire. I’m still curious to see how far AMD takes this capability in software.

Using more traditional methods, I was able to get our Phenom II X4 965 chip running stable at 3.9GHz, courtesy of beaucoup cooling and 1.5V of juice. That produced more dramatic performance results.

Not bad, eh? I don’t think the CPU temps went beyond 60°C with our Cooler Master Hyper tower air cooler.

The value proposition

We’ve taken a long and meandering route through several truckloads of performance data, and in order to help you make sense of it all, we have ripped a page from one of our CPU value articles.

To create a synthetic “overall performance” score, we computed an unweighted average of the results for a subset of our tests consisting of the benchmarks used in the CPU value article. Our formula includes 22 different benchmarks, but since our aim is practicality, it excludes a few more esoteric ones like the scientific computing applications. As our baseline, the Athlon X2 6400+ gets a 100% score. Other scores are all relative to it.

We took our prices from Newegg, since list prices and street prices on desktop CPUs from both Intel and AMD have diverged quite a bit lately. Note that those prices include reductions on the lower-end Phenoms, presumably made in anticipation of this product launch, even though they’re not yet reflected on AMD’s price list.

Of course, what you see below is a crazy experiment and probably meaningless, but some folks may find it a worthwhile thought exercise, at least. These scatter plots show price versus performance in a fairly intuitive way. To oversimplify slightly, the best CPU values tend to be located closer to the top and left edges of the plot.

The Phenom II X4 965 doesn’t look too bad here. It costs a little more, and performs a little better than, the Core 2 Quad Q9550. Then again, the Phenom II X4 955 is down to 200 bucks, and it looks like a bit better value than the new guy.

Now, here’s another crack at the same issue with total system cost taken into account. To get our pricing numbers for the X axis, we’ve added the cost of a motherboard, memory kit, graphics card, and hard drive to that of our processors. Wherever it made sense, we picked components from our latest system guide. Again, we got our prices from Newegg. Here’s a complete breakdown:

Intel LGA775 platform AMD Socket AM2+ platform Intel Core i7 platform
Gigabyte GA-EP45-UD3P $135 Gigabyte GA-MA790X-UD4P $110 Gigabyte GA-EX58-UD3R $190
4GB Corsair DDR2-800 $50 4GB Corsair DDR2-800 $50 6GB OCZ DDR3-1600 $125
Sapphire Radeon HD 4870 512MB $145 Sapphire Radeon HD 4870 512MB $145 Sapphire Radeon HD 4870 512MB $145
Western Digital Caviar Black 640GB $75 Western Digital Caviar Black 640GB $75 Western Digital Caviar Black 640GB $75
$405 $380 $535

Notice that we are making some assumptions here that may not be entirely valid. For instance, we’ve priced the Socket AM3 Phenom II processors on a Socket AM2+ motherboard with DDR2 memory, though we tested most of them with DDR3 memory. As you may have noticed, memory type didn’t make much difference at all to the performance of the Phenom II X4 810, and we expect the story will be similar for the rest. In the same vein, we priced the Core 2 processors with DDR2 memory, though we tested them with DDR3. Our goal in selecting these components was to settle on a standard platform for each CPU type with a decent price-performance ratio, not to exactly replicate our sometimes-exotic test systems.

The slightly lower cost of the AMD platform clarifies things for us. A nice line of green runs along the left side of the plot, signifying slightly better value for the Phenom II series than for the Core 2 alternatives. Then again, the Core i7-920’s breakout performance sure looks tempting up there.

Conclusions

So, this is a minor speed grade bump. I hope you weren’t expecting lightning and thunder.

My sense is that AMD has taken a bit of a risk here by introducing a CPU with a 140W TDP at this point in time. They may take flak from some quarters for doing so, and given the fact that Intel’s competing product has a 95W TDP, it’s not hard to see why. Still, our testing showed that the Phenom II X4 965 isn’t a total embarrassment in the peak power draw department, and it actually managed to beat out the Core 2 Quad Q9550 in our key power efficiency metrics thanks to strong performance and low system-level power draw at idle. The X4 965 is also a little faster overall than the Q9550, particularly in floating-point heavy applications like video encoding and 3D rendering. Yes, you will want a relatively large cooler to keep the X4 965 quiet under load, but that’s also true of any number of high-end CPUs from Intel and AMD at present.

Thanks to a combination of 45-nm process technology, guile, and giant, inflatable gorillas, AMD has managed to grant the Phenom II X4 965 Black Edition a reasonably attractive value proposition, as well. I’m not sure I could bring myself to buy anything right now, with Intel’s Lynnfield poking up over the horizon. But this new CPU at least gives AMD every bit of hope of remaining competitive it can squeeze out of its current technology portfolio. We’ll soon find out whether that’s enough.

Comments closed
    • ish718
    • 13 years ago

    Now I’ll wait for the stock 4.0ghz 32nm phenom II….

    • swaaye
    • 13 years ago

    Depends on the airflow really. I’ve gotten my PIIX4 940 up to 70C. I do think it’s hotter than my Q6600 @ 3.0 on stock V. The Phenom IIs idle very well, but when loaded down they are quite warm.

    • albundy
    • 13 years ago

    holy bejesus! 60 degrees with 3rd party cooler? I dont think i have ever gone past 54 degrees on max load with stock cooling on my 955be at 3.8ghz. then again, your working with a higher tdp.

    • SubSeven
    • 13 years ago

    You forgot a portion… its “Zorg oldies but goldies.”

    • XaiaX
    • 13 years ago

    Yes, depending on what’s happening on the screen. That’s the real issue here, though. Depending on how much motion is going on in a given scene, the human eye can recognize discontinuity at up to around three /[

    • XaiaX
    • 13 years ago

    That’s a great idea, I hadn’t meant to imply that log-based scales should be exclusive.

    That’d make a good comparison, though. “Here’s your linear performance for the price, here’s what you’re likely to notice.”

    • derFunkenstein
    • 13 years ago

    Sorry, it’s not very often I get mid-high performance for budget $$. The last time I even came close to that I was running a Thoroughbred rev B 1700+ at 2.3GHz on an Asus A7N8X right after they got released.

    • VILLAIN_xx
    • 13 years ago

    …you keep rubbing it in ya bastard!

    🙂

    • Meadows
    • 13 years ago

    Response times really aren’t a good example for using it.

    • vikramsbox
    • 13 years ago

    This is an article of interest where the 965BE when undervolted to 0.832V/1.240V gives a load output down by 30-50W as compared to the stock voltages of 1.000V/1.392V
    §[<http://www.anandtech.com/mb/showdoc.aspx?i=3621<]§

    • Mr Bill
    • 13 years ago

    Um, many systems, especially those incorporating diffusion in some form, have log normal responses. Our sight, hearing and sense of touch to name a few. Electronics often fall into this category. Log normal axis on some plots would split out the curves. IOMeter File Server Response time plots are a perfect example…
    §[<http://www.techreport.com/articles.x/15079/11<]§

    • toyota
    • 13 years ago

    babelfish has crapped itself…

    • derFunkenstein
    • 13 years ago

    Now you’re catching on. I know a good deal when I see it from any tech company. 😉

    • Game_boy
    • 13 years ago

    Yeah, an AMD server division employee (John Fruehe) says Istanbul is unlikely to have a desktop version because of research AMD has done suggesting the market is too small. The validation and testing costs of bringing it to a desktop platform are much higher than you would think.

    • sparkman
    • 13 years ago

    What does the rumor mill say about 6-core Phenom II X6’s? Anything? Cause I’m about to buy a X4 965 but I’m wondering about my upgrade options.

    • Meadows
    • 13 years ago

    Then again, flip-mode is on point once more, since we’re not interested in /[

    • Tamale
    • 13 years ago

    Indeed, but the OP’s suggestion of the importance of 60 to 70 being less significant than the importance of going from 30 to 40 is indeed interesting.

    • maxxcool
    • 13 years ago

    Samual l Jackson: What? What ain’t no country ever ever heard of? Do they speak English in what?

    Slacker on couch : What !?!?

    Samual l Jackson: ENGLISH [email protected]#$#$? DO…. YOU….. SPEAK…. IT!!????

    Slacker on couch : <shrieking> WHAT!?!?!?

    /Samual l jackson levels his handgun at the Slacker on the couch./

    Samual l Jackson: Say What one more time, I @#[email protected]#$ Dare you!!

    • maxxcool
    • 13 years ago

    Love that movie…. 😉

    • Freon
    • 13 years ago

    Yeah it does seem like everything is slowing down, but I do think the dollar is compelled to move away from faster and faster CPUs and GPUs and to other areas like SSDs, or HTPCs, home servers, etc. The race is more implementing useful products. I’m more excited about building or buying a home server, or PC DVR system than I am about building a killer i7 gaming rig. Even my lowly C2D system really isn’t sweating any game or app these days.

    Games are definitely a factor, as is the economy in general. Games are not going to push PCs much until the next generation of game consoles, now that PC exclusive games are so rare.

    Definitely excited about SSD. There is a lot of room for prices to drop and it seems to be happening. I’m not sure where the floor will be found, but it doesn’t seem like we’re seeing a corner in the price/size curve yet.

    • Freon
    • 13 years ago

    Yeah, i7 kinda pushes you to 6GB, so I’m totally fine with the i7 system being priced with 6GB. In the same vein, I think it is far more likely someone building a new X4 965 system would buy a newer AM3 board.

    I just think the cost should be cost of what was tested, whatever that may be. If what was tested isn’t perceived as exactly fair or the best choice otherwise, then accept that critique and end it there. It’s hard to pick when you have some choices like DDR2 vs DDR3, but a definitive decision should be made and held throughout the test and conclusions.

    The one data point of X4 810 at a ~30% lower clock rate doesn’t seem sufficient to say memory type doesn’t matter, either, but that’s a separate point.

    • Meadows
    • 13 years ago

    70 versus 60 is still well within the scope of your perception, of course depending on the game (RTS won’t show jack).

    • flip-mode
    • 13 years ago

    Bad move IMO. The log scale shows perception, not performance; when you review hardware it is generally the performance of the hardware that you are interested in. Now from a value perspective, a log scale would make a lot of sense – yes, this hardware is faster than that, but how perceptible is that increase in performance?

    • poulpy
    • 13 years ago

    Lol nice quote 🙂

    • maxxcool
    • 13 years ago

    Voila… the ZF1.

    …It’s light… the handle’s adjustable for easy carrying… good for righties and lefties…. Breaks down into four parts, undetectable by X-rays.. It’s the ideal weapon for quick, discreet interventions.

    A word on fire power: Titanium recharger. 3000 round clip with bursts of 3 to 300. With the replay button, another Zorg innovation, it’s even easier… one shot……. and replay sends every following shot to the same location… then there are the Zorg oldies…

    …Rocket launcher. The always efficient flame thrower… My favorite. Our famous net launcher, the arrow launcher, with exploding or poisonous gas heads – very practical.And for the grand finale, the all-new ice-cube system! …Four full crates, delivered right on time!

    • maxxcool
    • 13 years ago

    Peh! pi~ [email protected] ghz, 65$, 4ggskill 75$, Abitp35e….75$ :8)~

    pantz that ida meh haze!

    • FireGryphon
    • 13 years ago

    Wow, that would make TR the undisputed, most hardcore tech site on the ‘net. It might be hard to get used to a new scale. It’s easy to look at 60 vs. 70 fps and get an idea of how each card will scale with a more demanding workload, if only because we’re all so used to using a linear scale.

    • Mr Bill
    • 13 years ago

    The AMD 955BE on an MSI 790FX-GD70 scores 11.69 GB/sec with 2GB DDR3 1333 @ 7-7-7-20-1T, in the Tech Report review…
    §[<http://www.techreport.com/articles.x/16702/9<]§ But it scores 13.6 GB/sec with 8GB DDR3 1600 @ 8-8-8-24-2T and NB @ 2.0GHz. It scores 15.6 GB/sec with 8GB DDR3 1600 @ 8-8-8-24-2T and NB @ 2.4GHz. Its fairly easy to overclock even with 4 banks of DDR3.

    • adisor19
    • 13 years ago

    Is this why you hate Apple ?! Cause they don’t use AMD processors ?!

    tsk tsk tsk

    Adi

    • snakeoil
    • 13 years ago

    you are a winner!!!!!

    • derFunkenstein
    • 13 years ago

    That’s OK, I only paid about 2/3 of what you did and have a similar config – MA790X-UD4P and X2 550BE unlocked. 😉

    • equivicus
    • 13 years ago

    @maxxcool, Sprechen sie Englisch?

    • poulpy
    • 13 years ago

    Now now how dare you contradict an honest grammar nazi merely trying to get on with his job?! 🙂

    • OneArmedScissor
    • 13 years ago

    Sorry, I’m not intentionally trying to rain on everyone’s parade, but even those things aren’t going to amount to much. We’re not going to have actual USB 3.0 devices for farkin ever (eSATA is overkill as it is), DX11 cards are largely going to be more of the same, and the “cheaper SSDs” of the next year or so are pretty much already here (though the prices will drop little by little).

    It’s not CPUs, it’s computers in general. You have to consider that there’s a point for everything where enough is enough. It took a lot longer than for many other things to get there for high-resolution video games, which I’m sure is the primary interest of most people here.

    That doesn’t mean it’s a never ending race, though. Things aren’t even really getting that much cheaper anymore. The cost of new manufacturing technologies is exploding.

    Prepare for more boredom until we get some radical new type of OS or something.

    Hopefully, what will change most drastically is efficiency. It makes no sense that many cheap laptops operate on about 10w, while desktops have been sitting at generally 60w, minimum, for years and years.

    • Kaleid
    • 13 years ago

    I agree. Its all a bit unexciting isn’t it?
    Give us USB3, cheaper SSD and DX11 graphic cards instead.

    • DancinJack
    • 13 years ago

    Way too many flame wars on TR lately. Getting pretty obnoxious.

    • derFunkenstein
    • 13 years ago

    ENGLISH DO YOU SPEAK IT!

    • Krogoth
    • 13 years ago

    CPU world has been so stagnant.

    Core i7 975 and 965 are the only units that yield a significantly performance gap over 65nm Core 2 Quads that came out two year ago. I can’t remember the last time development was this slow.

    Anyway, the bottom line is that you cannot go wrong picking any of these chips. The budget, mid-range and high-end have of all their shining points.

    X4 965 is just a minor boost over its predecessors. It manages to pull itself away from its Core 2 45nm counterparts, yet still falls short of current i7 units. The upcoming i5s are expected to yield almost the same performance as the current i7s. Platform cost is a different matter though. I expect i5 to pricey for a few months until demand and supply stabilize.

    • maxxcool
    • 13 years ago

    Lies! Amd recmanded here on site! : §[<http://www.techreport.com/articles.x/9262/17<]§ Ja! :P ;/) swe that last time AMD pantz, now not so pantz in the milk :<9 ja!

    • thermistor
    • 13 years ago

    #50…hold on a minute – and if the whole grammar thing is snark, then I obiously missed it.

    You can say, he is taller than me. OR He is taller than I (am). Drop the am if you’d like, no big deal. ‘Than I’ sounds too formal to some listeners and is avoided. I try to say the whole thing – He is taller than I am. That way, it sounds natural, to me at least. Any time I don’t match subject/object it grates on my own ears. I wouldn’t be caught dead saying ‘he is taller than me’ but again that is only personal preference.

    I’ve looked through some grammar blogs, which proves i have no life.

    • flip-mode
    • 13 years ago

    That part of the baggage of the i7 platfrom IMO, and it would be disingenuous not to fill all the channels and not to have at least 4GB.

    • data8504
    • 13 years ago

    I agree. +1.

    • SomeOtherGeek
    • 13 years ago

    You could simply do the math yourself and know that all the dots would be more closer together. It doesn’t take a scientist to see that.

    TR did a great job either way. It might ultimately come down to personal preference cuz for all practical purpose they are both about the same without the i7 940/965/975 (those last 3 are for bragging rights)

    • Atradeimos
    • 13 years ago

    Intel’s marketing, especially in the HPC area, seems extremely anti-RISC. I don’t want to see the day my tiny little hobbyist microcontroller (or smartphone) has to lug around an x86 decode block.

    Nehalems and Phenoms are both RISC-like under the hood. I don’t get Intel’s obsession with x86 when they could expose Nehalem’s ‘u-ops’ and let compilers optimize for them. I think it would be way more flexible than the hardware decode arrangement they have going, and save a decent chunk of power.

    • wiak
    • 13 years ago

    there are two Phenom II 965 one is 140W (HDZ965FBK4DGI) and one is 125W (HDZ965FBGIBOX) much like Phenom II 945 has one 125W (HDX945FBGIBOX) and one 95W (HDX945WFGIBOX)

    • XaiaX
    • 13 years ago

    In a CPU test, yeah, it’s probably not such an issue. I was thinking more for general GPU performance comparisons.

    Also, yes, I hate when people truncate the lower values to exaggerate differences. That’s just plain manipulation.

    • asdsa
    • 13 years ago

    Thank you from this enlightening comment.

    • derFunkenstein
    • 13 years ago

    (edited)

    4GB on the Intel setup means asymmetrical data sizes and a performance penalty. Most TR readers who would either buy or recommend such a setup know that and therefore are less likely to buy it with 4GB than they are to buy ti with 6GB.

    they could also recommend 4GB of DDR2-1066 for the AM2+ platform because it’s capable of using such memory.

    • flip-mode
    • 13 years ago

    Phenom 2 X4 inferior, Core i7 superior.

    • ironoutsider
    • 13 years ago

    Intel and AMD both suck. ARM FTW!!!

    • dpaus
    • 13 years ago

    AMD could save a lot of money in the marketing department by just printing out Damage’s “value” charts and using a ruler to pick the right price point for their latest CPU!

    • SubSeven
    • 13 years ago

    I whole-heartedly agree with you. And again, i am not raising this to slap TR because i think the reviews/reports here are first rate, (for the most part) but because it seems illogical to me to analyze value proposition when comparing apples to oranges. It is unfair to say that one has a “better” value than another when the other has more of something (I think I outdid myself on this sentence here). Granted it doesn’t make much sense to invest in a high end i7 system and limit it by getting 4GB (2×2 dimms). The only feasible option i see here is to adjust the cost mathematically so that costs are represented for identical quantities of perifierals. For example taking the 6gb kit for $125 and applying only 2/3 of the cost in the value analysis, so that the comparison compares equal units. Maybe it’s just me, but to me this is like fractions, it’s hard to quantify which fraction is bigger (or in this case has a better value proposition) without a common denominator.

    • MadManOriginal
    • 13 years ago

    Yeah although you could put 2x2GB in to an i7, some tests indicate the dropoff in desktop use is minor. Doing that is a bit like getting an AM2+ platform for an AM3 CPU though 😉 Overall the platform value charts are less useful than the CPU ones because platform cost can vary a lot more by motherboard and RAM chosen, if you get deals, and whether you need more for the system than the basic parts they decide to include (the absolute price difference would remain the same but not the percent difference.) Nonetheless the value charts are still a nice metric that TR made.

    • albundy
    • 13 years ago

    i can say now i am truly happy with my amd 955be / msi 790fx-gd70 combo that i got from newegg last month for $300. got a stable overclock to 3.8ghz on stock cooling. an decent x58 board w/ i7 920 would have cost me more than double!

    • Meadows
    • 13 years ago

    At least I know something about grammar, and don’t say things like “… than I.”

    This is a message to you, Danny mah boi – it’s “than me”.

    • mattthemuppet
    • 13 years ago

    ha, one describes oneself as a kettle!

    • clhensle
    • 13 years ago

    I believe that the bus speed is 200mhz, so the +1 multi will give you an additional .2ghz (so 3.6ghz).

    Also, my roommate has a 940, we were able to push his to 3.2 on stock voltage and 3.5 on +.1v, 24hr prime 95 stable, then backed it off to 3.4 where it still sits, not a whole lot of headroom, but not bad for the cost.

    I have a Xeon 3350 2.66 (Q9450 equivalent) and it has massive stock voltage headroom, I hit a bus limit of 440 on my dfi x48 (3.52ghz) on the stock voltage, so I lowered it to 1.8125v 31hrs prime stable, stayed there for a while until I installed L4D, had some stablility issues, dropped it to 3.4ghz and never looked back.
    §[<http://splak.net/computer/watercooled/benchmarks/uncompressed/prime2.JPG<]§

    • thermistor
    • 13 years ago

    #46…The i7’s use a triple channel controller. To populate it with something other than equal quantities in each of the 3 banks would artificially limit it. We could go back to 3 x 1 Gb dimm, which may handicap i7, and would anyone buy i7 and put in less memory than they a likely using with AM2+/AM3/775? Seems like there is no good way to make everyone happy.

    • SubSeven
    • 13 years ago

    I think you gentlemen are both missing a more important issue. I personally think that using the ddr2/am2+ setup in the value propostion is just fine, after all, it is one of the benefits of an AMD setup. What i have a big problem with is comparing a setup with 6gb DDR3 ($125) to a setup with 4gb DDR2 ($50). This skews results even further to AMD on the value proposition. Why not use 4gb DDR3 instead, which can be had for $70-$75 (or less)? I never really understood that.

    • Meadows
    • 13 years ago

    Which is an achievement in itself.

    • danny e.
    • 13 years ago

    everybody knows you’re a moron & more obnoxious than i.

    • snakeoil
    • 13 years ago

    everybody knows that anandtech is in intel’s payroll, so please don cite them as a source.

    • cygnus1
    • 13 years ago

    But wouldn’t that line still be there if they’d priced the Core 2 and AM3 with DDR3? The ram would cost the same either way.

    Both the Core 2 and the AM2+ with DDR2 platforms are dead ends. We all agree with that. But most non-geek people don’t care about platform longevity because they don’t plan on upgrading a CPU. They might throw in more ram, a bigger hard drive or a faster video card. But, very few people actually buy CPUs outside of full system purchases.

    I very much agree that the performance difference between DDR2 and DDR3 are minuscule for the Core 2 and Phenom II, therefor a cost performance chart should be based on the reasonably cheapest system you could possibly build around either CPU.

    I think the graphs are great as they are.

    • flip-mode
    • 13 years ago

    What’s wrong with snakeoil?

    §[<http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3619&p=8<]§ AMD x4-965 uses _[

    • maxxcool
    • 13 years ago

    Tens! London! Snakeoil under the northbridge! Sunday! Sunday! Sunday! Ja! hada to metso Snakeoil!

    965 50% slower in pantz! §[<http://www.techreport.com/r.x/phenom-ii-x4-965/x264-2.gif<]§ iaia! who dont play winner not in capsasi

    • cobalt
    • 13 years ago

    My apologies if I missed it, but could you clarify the clockspeeds the Smart Profile nets you? (Not enough coffee to find the base numbers and multiplier to calculate it myself….)

    Also, any idea what speed you can hit on the Phenom 965 without touching the voltage? The Q9550 E0’s are apparently pretty impressive in that regard for those of us paranoid about upping those V’s….

    • Freon
    • 13 years ago

    I’m not trying to defend a lazy reader, but it seems reasonable to critique this. It affects the conclusions like the, ‘A nice line of green runs …’. It’s starting to stretch credulity to pile on conclusions based on some sketchy assumptions. Disclosure is one thing, making further conclusions or observations on top of flawed data is another. These sort of assumptions on price are at the core of the idea of these value graphs. It compounds errors.

    I have to agree with MadManOriginal on this one, and say it is something TR should consider rectifying in the future by either providing more accurate cost estimates or changing their test rig back to an AM2+.

    • Damage
    • 13 years ago

    An update: The only newer stepping of a full-cache Core 2 Quad that I have on hand is a Core 2 Quad Q9550S, funnily enough. I can attempt to simulate a higher-power Q9550 with it via overvolting, but in order to do so, I have to know:

    1) That I can expect similar idle power draw from a Q9550 and a Q9550S when they are throttled down at idle via SpeedStep.

    2) What a typical peak core voltage is for a recent-stepping Q9550.

    I can’t seem to find these answers online. Question 2 is slippery because voltages vary from chip to chip, set at the factory. I’ve asked Intel if they can supply me with this info or, alternately, with a real and recent Q9550.

    If any of you guys think you have clear and authoritative answers to those two questions, though, by all means let me know.

    • Freon
    • 13 years ago

    As long as the left edge is zero, I don’t see an issue. I only have an issue when reviewers crop off and show data normalized to some arbitrary value. Like if you have three systems that produce 30hz, 40hz, and 50hz, a bad review would set the x-crossing at 20hz, making the 50hz system look 3 times faster than the 30hz system.

    I would only argue some of the tests are kinda pointless. HL2 for instance is ridiculously fast on any of these processors, around double the 60hz LCD most people are probably using anyway. It’s the apps that push the envelope, where not all systems are well above the “faster than needed” threshold.

    • indeego
    • 13 years ago

    /[

    • flip-mode
    • 13 years ago

    Stroke your science training to someone else, please. I understand scientific methond, so there is no reason to condescend. It is a *[

    • MadManOriginal
    • 13 years ago

    He says ‘they suspect’ there’s no performance difference not that ‘it’s a fact.’ It’s a reasonable assumption but it’s still an assumption. What is a fact is that the platform cost chart is skewed by putting the AM3 CPUs on an AM2+ platform if only because it would be silly for people buying a new platform to get an AM2+ platform for an AM3 CPU – after all one of the draws of AM3 is that it will be AMDs main platform for some time to come. I guess sloppy methods are acceptable to you but my science training makes me balk at such errors. *Now that I look at the test methods many of the LGA775 CPUs are on a DDR3 platform too. *sigh* I guess the value charts are just something to overlook then.

    It’s the same type of problem with using a QX9650 to simulate a Q9550, at least for power draw especially after comparing TR’s results in that area with other sites. Is it one answer? Sure, but is it accurate? Not really, or at least not without using a non-siumulated CPU to know.

    • Mr Bill
    • 13 years ago

    I like this idea and supporting observations. But it won’t happen.

    • Lazier_Said
    • 13 years ago

    Thanks for clearing that up, Damage.

    • Corrado
    • 13 years ago

    Also, at least in the past, Intel has rated TDP as the average usage, whereas AMD has rated it as the absolute top usage. This is why you have chips that rate @ 140W that go from 2.0ghz to 3.4ghz. They are the same stepping and chip with different multipliers, and will MAX OUT at 140W under complete and utter meltdown mode.

    • flip-mode
    • 13 years ago

    It would be misleading if they did so without mentioning it, and without also mentioning the fact that there is negligible performance difference. But the manner in which TR did it was not at all misleading. It is nothing more than a method, and methods are various, and it is possible for various methods to all produce meaningful results.

    • Corrado
    • 13 years ago

    The stock cooler cools fine, its just a bit louder than some may want. He said to keep it QUIET, not to keep it COOL.

    • XaiaX
    • 13 years ago

    Have you considered displaying frame rates on a logarithmic scale?

    The problem with the linear scale is that it presents 70 vs. 60 fps as if it were the same kind of performance increase as 30 vs. 40 fps. Due to the way the human (well, any organism’s) vision system works, perception of smoothness is non-linear. I’m not sure of the just-noticeable-difference for for frame rates, (or if anyone has actually studied that) but you could use pitch as a model and go with log2.

    I think it’d be interesting to see, since most people are intuitively aware of the idea of “diminishing returns”, but to explicitly graph out the difference logarithmically could be illuminating.

    • MadManOriginal
    • 13 years ago

    It’s misleading or at least inaccurate for the anal retentive scientist type to test a CPU with one platform (AM3/DDR3) and then calculate its platform cost with another platform (AM2+/DDR2) for the price/performance charts even if the performance numbers between the two platforms are close.

    • derFunkenstein
    • 13 years ago

    Did you read the test setup page?

    AHCI controller 3.1.1540.61

    • derFunkenstein
    • 13 years ago

    3.9GHz is only slightly higher than the 3.8 I can hit as my max, and right in line with what other folks were getting on air with Phenom II X4s. That means that this wasn’t a significant revision/stepping, which is too bad. I was hoping they might be able to find a way to rearrange the silicon to fit in a lower thermal envelope and hit higher clocks – AMD is going to need them if they want to keep up with Intel.

    • Meadows
    • 13 years ago

    So according to your performance graphs, both the X4 955 and 965 are better than the Q9550. That’s not a bad showing.

    (The 955 because it’s better value, and the 965 because it’s faster by a manatee whisker.)

    • poulpy
    • 13 years ago

    I totally agree, those TDP are calculated differently and do not represent the exact same thing so comparing them too seriously doesn’t make sense.

    • Meadows
    • 13 years ago

    Look at progress.
    Back in my day,…

    • Freon
    • 13 years ago

    ” Yes, you will want a relatively large cooler to keep the X4 965 quiet under load, but that’s also true of any number of high-end CPUs from Intel and AMD at present.”

    Are the retail fans not sufficient at stock speeds? I imagine the value proposition could change if the X4 965 required a $40 HSF whereas the Q9550 is happy enough with the Intel cooler.

    • AlvinTheNerd
    • 13 years ago

    Heres a good conclusion: The 965 competes well with the 9550 both in performance and heat. Just let the total system cost determine your choice.

    Again and again it is brought up that the 9550 is 95W and the 965 is 140W. Even the author seems to be be fooled by these numbers. But the TDP is a motherboard spec, nothing more, nothing less.

    Look at the numbers again: At peak, the 965 is 15W hotter, at idle 20W cooler. Overall I think the 965 is the cooler of the two processors considering the vast majority of desktop users.

    • snakeoil
    • 13 years ago

    the phenom 2 965 idle power consumption is 30 watts less than i7 and 40 watts less than the old core 2 quad.
    impressive.

    §[<http://forums.vr-zone.com/news-around-the-web/469832-phenom-2-965-idle-power-compsumtion-impressive.html<]§ overclocks to 4.2 ghz on air. §[<http://forums.vr-zone.com/news-around-the-web/469766-phenom-2-965-4-2ghz-air-stable-review.html<]§

    • wingless
    • 13 years ago

    You are spot on with that assessment sir. Clock for clock AMD isn’t with Core i7, but product-to-product and price-to-price it is! We have a perf/$ champ in the 965. Air cooled it clocks alright, but with water cooling it would murder.

    • armistitiu
    • 13 years ago

    Hey, all i see is a great CPU at a perfect price point (don’t lower it AMD! it’s good where it is) that sometimes performs even better than extreme edition intel CPU’s or even i7. Sure the architecture isn’t the best one around. But who cares about that when you buy a computer? I’ve always liked TR’s graphics (the one where they show you perf/price) and it’s clear to me that PhII 965 is a perf/$ monster.
    One thing about the TDP rating: AMD measures it as the absolute max unlike Intel…so maybe that’s why it’s reported at 140W TDP. But the power consumption stayed in normal parameters. Also i hate that intel and AMD don’t use the same thermal measuring technique.
    Good review btw.

    • sparkman
    • 13 years ago

    Ewwwww I got mojo on me…!

    • juampa_valve_rde
    • 13 years ago

    few ticks left for deneb… what’s next on desktop amd? 6 core?

    • flip-mode
    • 13 years ago

    I am fully impressed with the idle power consumption of Phenom 2 – what a vast improvement it is over Phenom 1. And it’s even better than Core 2 Quads.

    I do think that if I was buying today I’d get an X4-965 on the MSI GD70 or the Gigabyte 790FX UD5 (mangled the names of those mobos a little, I’m sure).

    But I am most certainly waiting for Lynnfield.

    • OneArmedScissor
    • 13 years ago

    Wow. I can’t believe Intel tells you to simulate them!

    Thank you for being open about that. Otherwise, I don’t think anyone ever would have known that’s going on.

    Now what else don’t we know? :p

    • Damage
    • 13 years ago

    The differences you point out are definitely due to different steppings of the chips. In fact, as the test notes state, the Q9550 in our tests is an underclocked and undervolted QX9650, made to simulate the Q9550. That makes it pretty old, since the QX9650 was a very early Penryn.

    Now that you mention it, I should have pointed that out more prominently. I will look around in the morning to see if I have any newer full-cache Penryns that might produce lower idle power numbers.

    I’m not sure I do, though. I’ve spoken with Intel repeatedly about the problem of them sending out ultra-high-end chips and asking us to use them to simulate lower-end parts, specifically in reference to power consumption testing. They have, to date, not been especially receptive or willing to sample additional SKUs or newer steppings with any regularity. At least, not to us.

    I need to sleep now, but I’ll try to address this more fully in the (later) A.M. Thanks.

    • OneArmedScissor
    • 13 years ago

    Look at page 2, test notes. They have a bazillion different setups. It’s the motherboards. Check out the difference in power between the Core i7 965 and 975 and setups.

    It annoys me that there’s not terribly much consistency or emphasis put on the SYSTEM power use, but then again, all of the enthusiast sites seem to use motherboards that are the flagships of enthusiast product lines, likely because they’re given to them for the advertising. They have so many bells and whistles, the power use gets jacked way up.

    My brother’s computer has a middle of the road P45 and a Radeon 4850, but it only uses 90w idle, and its power draw while playing games, or at a full CPU load, is about 140w.

    And yet, most sites show that their Core 2 quad setups idle at what my brother’s computer sees at its peak. Most of us aren’t using that sort of hardware, and it doesn’t really have much relation to how much power our computers are actually using.

    I pretty much ignore it. It doesn’t even give you a general idea. For example, many sites show Core 2 and Phenom II as basically even, with Core i7s using drastically more power, even at idle. And yet, Anandtech shows Core i7s beathing them both. There’s no consistency.

    • Lazier_Said
    • 13 years ago

    TR’s power measurements are not internally consistent.

    Anandtech for example measured their testbed with the Q9400, 9450, 9650 within 2.5 watts of each other at idle, and the QX9770 is up 4 watts. About what you’d expect considering power management features.

    TR shows their idle Q9550 and QX9770 as 23 watts higher than the ought-to-be-almost-identical Q9400.

    At load AT measured the speedbumped Q9650 and QX9770 as 17 and 24 watts over the Q9400.

    TR shows a 39 and 45 watt spread, respectively.

    I suspect either some aspect of TR’s testbed was changed between measurements or else you’re mix and matching current and last year’s steppings.

    Idle cache doesn’t use 20 watts.

    • OneArmedScissor
    • 13 years ago

    AMD has made multiple versions of the same CPU with different TDP ratings for years, without changing the stepping. Their 45nm CPUs have all been the same stepping since the Opterons came out. They’ve said that the increases in clock speed over time are because their yields have been better than they originally expected.

    I believe the TDP variations are more of a binning thing. Even fairly recently, they were still selling 89w and 125w X2 6400+s.

    • toyota
    • 13 years ago

    so the Phenom II X4 945 95 watt version wasnt a new stepping? it seems that AMD would have at least waited till they could do something similar with the 965 before releasing it.

    • Ushio01
    • 13 years ago

    Did you use AMD AHCI drivers or Vista for the winzip, nero and photoshop tests?

    Because in your High-end Socket AM3 chipsets review(https://techreport.com/articles.x/17061/9) you showed that the vista drivers are superior and I would love to see if they improve phenom performance in the above 3 tests.

    • OneArmedScissor
    • 13 years ago

    They haven’t moved to a new stepping yet. It’s been 10 months since the 45nm Opterons came out. Both Intel and AMD generally take about that long.

    They could very well be gearing up to do that shortly after the Lynnfields are out. I’m sure they could squeeze out 3.5 GHz, if not 3.6 GHz, if they’re willing to stick to the 140w TDP.

    What I’m wondering isn’t if they’ll be able to push it any further, but if we’re starting to see the end of overclockable CPUs. We’ve become accustomed to having a huge amount of headroom, but there’s been a 4 GHz-ish wall for years now. AMD is slowly but surely climbing up to that with base clock speeds, and Intel will hit it with the turbo boost on later Westmeres.

    • ssidbroadcast
    • 13 years ago

    q[

    • Prion
    • 13 years ago

    Hmm, being rated at a 140W TDP means we probably won’t be seeing much in the way of higher clocked Phenom II X4s for a while, huh? Probably another reason for these TurboBoost-style AOD profiles, gotta look competitive against the upcoming Core iX barrage until Bulldozer can get out the door.

    • UberGerbil
    • 13 years ago

    Nice work.
    g[

Pin It on Pinterest

Share This

Share this post with your friends!