One thing that caught my eye when I read the first email announcing these processors was that the Athlon II X4 has now reached 3.0 GHz. The Phenom II X4 940 & 945 processors also run at 3.0 GHz, making this an interesting crossing point where the value line is really starting to cross over into the performance line (in terms of raw clock speed). I checked the price on Newegg for the existing Athlon II 635 (2.9 GHz) processor and it's only $106. We can only assume that the new Athlon II 640 will retail around that same price, if it comes in at the list price of $122 (or higher) then it would be really hard not to justify bumping up to the Phenom II line.
As mentioned previously, all of these new processors will be the C3 revision, meaning they support hardware C1E and other little performance benefits over the previous C2 silicon. For some reason CPU-Z wouldn't show the revision, so I took a snip from the AMD Overdrive utility.
Too keep things consistent and try to maintain and "apples to apples" comparison, we try to use as much common hardware across the board. Each system's BIOS was set to "Optimized Defaults", with a verification of CPU & Memory speeds and timings.
We started out with a clean install of Windows 7 and performed all the Windows updates to bring it up to current levels as of when this review was written. Then we installed all out benchmark programs and ran a disk defrag. Afterwards we disabled some un-needed Vista features that could affect results such as SuperFetch, System Restore, Defrag, Indexing, and Remote Assistance. Furthermore we ran the Bapco Auto-Configuration Tool and used its "Level 3" settings which disables a bunch more stuff:
Each of the benchmarks were run at least five times, with the average being reported.
Processor Cost:
The table below prices out the various AMD & Intel processors that all fall within the same ballpark (and a few that aren't). One factor to consider is the additional cost for a motherboard and RAM (at minimum). Generally speaking, AMD boards are less expensive than Intel. I think it is a pretty good spread, and please keep these cost figures in mind when comparing performance numbers.
All the prices are taken from Newegg, with the exception being the new Athlon II X4 640. I would expect the final price to be closer to $105-$110, but we will just have to wait and see.
We can see from the price range, the biggest competitor for the Athlon II X4 640 would be Intel's Core i3 530 & 540 (dual core) processors, and possibly the Core 2 Quad Q8300. There are some Intel Core 2 Duo processors that fall within the price range (like an E7500), however I wouldn't expect overall performance to be able to compete.
Overall System Power Consumption:
Using a Kill-A-Watt meter we did a simple test to see how much a minimal system would consume while idle and under load. The "basic" system consisted of the motherboard, CPU, HSF, RAM, HDD, DVD, video card, and one spare 120mm fan. For idle testing C1E, EIST, and Cool n Quiet were enabled. For load testing we ran the Prime 95 (multithreaded version) stress test.
PCMark Vantage:
I'm usually not a fan of the big bundle type benchmarks (i.e. Sysmark & PCMark) because getting consistent numbers in the past was quite a pain. However, utilizing a SSD for the HDD benchmark helps to get better averages and we ran the tests three times. You can read more about the features of PCMark Vantage on the
Futuremark website.
Adobe Photoshop CS4:
To measure performance in Photoshop, we uses
Retouch Artists Speed Test. Photoshop was configured as per the included documentation, and we used a SSD for the scratch drive to ensure consistent numbers.
Everest CPU Queen's Problem:
This simple integer benchmark focuses on the branch prediction capabilities and the misprediction penalties of the CPU. It finds the solutions for the classic "
Queens problem" on a 10 by 10 sized chessboard.
At the same clock speed theoretically the processor with the shorter pipeline and smaller misprediction penalties will attain higher benchmark scores. For example -- with HyperThreading disabled -- the Intel Northwood core processors get higher scores than the Intel Prescott core based ones due to the 20-step vs 31-step long pipeline. However, with enabled HyperThreading the picture is controversial, because due to architectural bottlenecks the Northwood core runs out of internal resources and slows down. Similarly, at the same clock speed AMD K8 class processors will be faster than AMD K7 ones due to the improved branch prediction capabilities of the K8 architecture.
CPU Queen test uses integer MMX, SSE2 and SSSE3 optimizations. It consumes less than 1 MB system memory and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest CPU PhotoWorxx Processing:
This integer benchmark performs different common tasks used during digital photo processing.
It performs the following tasks on a very large RGB image:
- Fill
- Flip
- Rotate90R (rotate 90 degrees CW)
- Rotate90L (rotate 90 degrees CCW)
- Random (fill the image with random coloured pixels)
- RGB2BW (color to black & white conversion)
- Difference
- Crop
This benchmark stresses the integer arithmetic and multiplication execution units of the CPU and also the memory subsystem. Due to the fact that this test performs high memory read/write traffic, it cannot effectively scale in situations where more than 2 processing threads used. For example, on a 8-way Pentium III Xeon system the 8 processing threads will be "fighting" over the memory, creating a serious bottleneck that would lead to as low scores as a 2-way or 4-way similar processor based system could achieve.
CPU PhotoWorxx test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest CPU ZLib Compression:
This integer benchmark measures combined CPU and memory subsystem performance through the public
ZLib compression library Version 1.2.3.
CPU ZLib test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest CPU AES Encryption:
This integer benchmark measures CPU performance using AES (a.k.a. Rijndael) data encryption. It utilizes Vincent Rijmen, Antoon Bosselaers and Paulo Barreto's
public domain C code in ECB mode.
CPU AES test uses only the basic x86 instructions, and it's hardware accelerated on VIA PadLock Security Engine capable VIA C3 and VIA C7 processors. The test consumes 48 MB memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest FPU Julia Fractal:
This benchmark measures the single precision (also known as 32-bit) floating-point performance through the computation of several frames of the popular "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87, 3DNow!, 3DNow!+ or SSE instruction set extension.
FPU Julia test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest FPU Mandelbrot Fractal:
This benchmark measures the double precision (also known as 64-bit) floating-point performance through the computation of several frames of the popular "Mandelbrot" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87 or SSE2 instruction set extension.
FPU Mandel test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Everest FPU SinJulia Fractal:
This benchmark measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing trigonometric and exponential x87 instructions.
FPU SinJulia test consumes less than 1 MB system memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
CINEBENCH:
CINEBENCH is based on MAXON's commercial animation software, CINEMA 4D, which is used extensively by studios and production houses worldwide. The program is capable of scaling up to 16 CPUs, is multithreaded, and has 64bit executables.
3DS Max Design 2010:
For this benchmark we rendered the "Lounge Bar Scene Final" from the included demonstration files.
POV-Ray:
POV-Ray is a free raytracing program. Version 3.7 beta has 64bit support and native SMP capability. We used the built-in benchmark to test multithreaded performance. An interesting note, the latest b37a gives slightly better numbers for AMD processors over b36 (I saw consistently a couple seconds of improvement), however the Intel numbers stayed relatively the same.
x264 HD Benchmark:
x264 is a free library for encoding H264/AVC video streams.
Tech ARP has a nifty benchmark app to bring some automation and standardization to the encoder. Simply execute a batch file and it will run three tests using two different encoder versions. We are only reporting the newest version since its scores are a little better. This is multithreaded and maintained very high CPU usage, especially during the second pass.
WinRAR:
WinRAR has a built in benchmark tool that offers both single and multithreaded tests. We are just reporting the multithreaded test results.
Mathematica:
Wolfram Mathematica does computation, modeling, simulation, visualization, and so much more. It is a very robust product that has been around for as long as I can remember. The built-in benchmark program runs the following tests and reports a score along with the time it took to process.
- Data Fitting
- Digits of Pi
- Discrete Fourier Transform
- Eigenvalues of a Matrix
- Elementary Functions
- Gamma Function
- Large Integer Multiplication
- Matrix Arithmetic
- Matrix Multiplication
- Matrix Transpose
- Numerical Integration
- Polynomial Expansion
- Random Number Sort
- Singular Value Decomposition
- Solving a Linear System
wPrime:
wPrime is a perfectly threaded benchmark that gives very consistent scores and uses 100% of the CPU. The name might be a little misleading, it actually has nothing to do with prime numbers, it actually does square roots calculations.
Folding @ Home Benchmarks:
We at EXTREME Overclocking are Folding @ Home nuts! If you fold then you have probably seen the
EOC Folding Stats pages. Using the Folding@Home benchmark CD created by
notfred, we ran the benchmark64 app to get these numbers. The benchmark doesn't run any of the new SMP work units, but it does give general relative performance of each processor.
Far Cry 2:
Far Cry 2 has a nice built-in benchmark tool. Essentially we ran the Ranch Small test using the Very High settings overall at two difference screen resolutions.
With very high settings and high resolution, the numbers become much closer together. This is a situation where the video card is going to have more of an impact than CPU speed or the number of cores. Pretty much all processors performed identical, except for the dual-core which is at a slight disadvantage.
Crysis Warhead:
We ran the HardwareOC Crysis Warehead benchmark front end to maintain consistency in our tests. There's not a lot of options in the tool, mostly just setting the screen resolution, graphics quality, and choosing the map.
DiRT 2:
My little nephew loves race car sims, especially ones where the cars can take damage. Dirt 2 is pretty new and supports up to DirectX 11, and it also has a built-in benchmark making results easily comparable and reproducible. We made sure to install the 1.1 patch as it fixes a few configuration defaults and such.
For the life of me I could not get DiRT 2 to run on the Q8300 system, it simply kept crashing at the beginning while loading. I tried all the fixes I could find on the Internet (that wouldn't affect benchmark numbers) but nothing helped. However, from the numbers below, it's pretty apparent that CPU power doesn't mean a whole lot (unless it's a really really old CPU).
TrackMania Nations Forever:
The TrackMania series of games are extremely popular world-wide. TM Nations Forever is a FREE downloadable game (the rest in the series you have to buy) that allows single player and multiplayer action. You can even build your own tracks! It's a fun game that you can play just taking a little break, or get lost in the action for hours.
Valve Particle Simulation:
Using the source engine as the core, Valve wrote a little particle simulation map / benchmark. This benchmark is heavily multithreaded, maintaining around 90% CPU usage.
Valve VRAD Map Compilation:
Another nifty benchmark from Valve uses their VRAD lighting tool to pre-compute lighting in games. When compiling the map, CPU usage stayed at a constant 100%.
Product Summary & Final Conclusions:
As I stated in my opening paragraph, these CPUs for the most part are simply speed bumps in existing product lines along with using the latest C3 revision silicon. These new processors should replace existing models at the same price point, thus pushing down the cost of all existing models.
From an upgrade perspective, AMD still reigns supreme by having all their Athlon II & Phenom II AM3 processors backwards compatible with the AM2/AM2+ socket.
When you put the AMD Athlon II X4 640 and Phenom II X4 945 side by side, both running at 3.0 GHz, it's clear there is a slight performance advantage to the Phenom II line with the 6MB of L3 cache (which is one reason the 945 costs more). Even with the existing overlap between the two lines, one still has to realize that the 640 is the top of the Athlon II line, while the 925/940/945 processors are at the bottom of the Phenom II line.
Comparing Intel products, the cheapest quad-core (which we included in this review) is the Core 2 Quad Q8300, costing ~$150. The cheapest quad-core i5 processor is ~$200, which for that cost you can get a Phenom II X6 1055T.
If you would like to see some extra comparative numbers with X3 & X2 processors, you can check out some earlier reviews on the
Athlon II X4 635 & X3 440 processors or
Athlon II X3 435 & X2 240e processors. Please remember that those reviews used different benchmark programs and hardware setups, so don't try to compare them to numbers in this review.
| AMD Model | Frequency | Speed Bump | Cores | L2 Cache | TDP | Price | Tray OPN |
|---|
| Athlon II X4 640 | 3.0 GHz | +100 MHz | 4 | 2 MB | 95 W | $122 | ADX640WFK42GM |
| Athlon II X3 445 | 3.1 GHz | +100 MHz | 3 | 1.5 MB | 95 W | $87 | ADX445WFK32GM |
| Athlon II X2 260 | 3.2 GHz | +100 MHz | 2 | 2 MB | 65 W | $76 | ADX260OCK23GM |
| Athlon II X4 610e | 2.4 GHz | +100 MHz | 4 | 2 MB | 45 W | $143 | AD610EHDK42GM |
| Athlon II X3 415e | 2.5 GHz | +200 MHz | 3 | 1.5 MB | 45 W | $102 | AD415EHDK32GM |
| Athlon II X2 245e | 2.9 GHz | +100 MHz | 2 | 1 MB | 45 W | $77 | AD245EHDK23GM |
