Table of Contents
With an MSRP of just $102, the new AMD Phenom II X4 840 CPU is one of the cheapest quad-core CPUs on the market. Value for money is usually a given with AMD CPUs, but can this processor push the envelope when it comes to budget performance?
AMD Phenom II X4 840 Processor Specifications
|Model Number/Core Frequency||Phenom X4 840 @ 3.2 GHz|
|L1 Cache||64 kB instruction/64 kB data per core (512 kB per CPU)|
|L2 Cache||512 kB per core (2 MB per CPU)|
|Memory Controller||Integrated 128 bit wide memory controller|
|Up to 2.0 GHz with Dual Dynamic Power Management|
|Supported Memory||DDR2 up to PC2-8500 (1066 MHz)|
|DDR3 up to PC3-10600 (1333 MHz)|
|HyperTransport 3 Specification||One 16/16-bit link at up to 4 GHz full duplex (2 GHz x 2)|
|Total CPU-to-system bandwidth||Up to 37.3 GB/s (21.3 GB/s memory, 16.0 GB/s HT3) with DDR3|
|Up to 33.1 GB/s (17.1 GB/s memory, 16.0 GB/s HT3) with DDR2|
|Packaging||938-pin Socket AM3 organic micro-pin grid array (mPGA)|
|Fab. Location||GlobalFoundries Fab. 1 module 1 in Dresden, Germany|
|Process Technology||45 nm DSL SOI (Silicon On Insulator)|
|Approximate Die Size||169 mm-squared|
|Approximate Transistor Count||ca. 300 million|
|Max. Thermal Design Power||95 Watts|
With such a low price-tag, costs have to be cut somewhere. The L3 cache is the one glaring omission from this CPU’s specifications. Basically, this is an Athlon II X4 645 plus 100 MHz. If the X4 645 could be purchased for less than $102, the X4 840 would receive an automatic “fail” from me, since it would be nothing more than sneaky marketing. However, the X4 840 is $16 less than the 645, making it a great value.
|Athlon II X4 640||$99.99||4/4||3.0 GHz||2 MB||None|
|Phenom II X2 560BE||$99.99||2/2||3.3 GHz||1 MB||6 MB|
|Phenom II X4 840||$102||4/4||3.2 GHz||2 MB||None|
|Phenom II X2 565||$109.99||2/2||3.4 GHz||1 MB||6 MB|
|Core i3-550||$114.99||2/4||3.2 GHz||512 kb||4 MB|
|Athlon II X4 645||$118.99||4/4||3.1 GHz||2 MB||None|
|Phenom II X4 920||$119.99||4/4||2.8 GHz||2 MB||6 MB|
|Phenom II X4 945||$134.99||4/4||3.0 GHz||2 MB||6 MB|
|Phenom II X4 955BE||$144.99||4/4||3.2 GHz||2 MB||6 MB|
|Core i3-560||$149.99||2/4||3.33 GHz||512 kb||4 MB|
|Phenom II X4 965BE||$164.99||4/4||3.4 GHz||2 MB||6 MB|
|Phenom II X6 1055T||$179.99||6/6||2.8 GHz||3 MB||6 MB|
|Core i5-650||$179.99||2/4||3.2 GHz||512 kb||4 MB|
|Phenom II X4 970BE||$185.99||4/4||3.5 GHz||2 MB||6 MB|
|Core i5-655k||$199.99||2/4||3.2 GHz||512 kb||4 MB|
|Core i5-750||$199.99||4/4||2.66 GHz||1 MB||8 MB|
|Phenom II X6 1075T||$199.99||4/4||3.0 GHz||3 MB||6 MB|
|Newegg.com Prices current as of Jan. 18, 2011|
In looking at the market, the X4 840 seems like a great value 4-core solution with 100 MHz higher clock speed and priced $17 less than the Athlon II X4 645. For a 2% increase in price, or $2 more than the Athlon II X4 640, the X4 840 offers a solid 200 MHz boost, or 6% increase in clock speed. It is pretty clear the X4 840 will out-duel similarly priced Intel offerings, like the i3-550. Those of us who have been following the pricing and performance trends over the past year know that without even looking at the above chart. The more interesting question is how the X4 840 compares low-end i5 CPUs (with two cores but four virtual threads) and the rest of the Phenom lineup with L3 cache.
Overclocking this chip was a breeze. Getting it stable at 3.8 GHz did not require too much juice or messing with additional settings. However, I did run into an issue with the motherboard: generally lower multipliers are available, which greatly helps the overclocking process. Despite this problem, I managed to achieve a stable 3.8 GHz to run all of the performance tests.
|Processor||Phenom II x4 840||Core i5 655k|
|Stock / Overclocked Speeds (GHz)||3.2 / 3.8||3.2 (Stock Only)|
|Motherboard||Biostar TA890GXE||Gigabyte GA-P55UD6|
|RAM||Kingston HyperX DDR3-2000 4GB|
|RAM Speed (overclocked)||DDR3-1600|
|GPU (for total 3DMark Score Only)||AMD HD 6950|
|Operating System||Windows 7 (x64)|
No surprises here: the four-core X4 840 trounced the two-core i5 655k in rendering performance. It is interesting to note that for more than double the price, the X6 1100T only yields a 63% performance increase over the X4 840. Also, despite its mere 200MHz additional clock speed, the X4 965 has a solid 12% performance gain over the X4 840. This is likely where the lack of L3 cache on the X4 840 comes into play. Whether this difference in performance is worth $50 is up to you. The value of this chip is really starting to take shape.
Once again, there i5 655k barely puts up a fight against the X4 840 in this multi-threaded file compression benchmark. For $100 less, users get over 20% performance boost with the X4 840. When overclocking this chip, performance scales almost linearly: a 19% increase in clock speeds yields almost 21% in performance. If the computer was only used for file compression and rendering, the answer would be simple: more cores equals more performance.
The i5 655k and X4 840 are pretty much on par performance-wise with 3DMark Vantage. There is about a 5-12% difference depending on which score is analyzed. As for the CPU score: the 5% is relatively small and is potentially due to differences in chipsets/platforms between Intel and AMD.
When looking at total Vantage score, the i5 does have a 12% edge in performance. This result is quite surprising considering the CPU score showed the exact opposite result, with the X4 840 coming out on top. The only possible explanation that comes to mind is the PCIe link. On the i5 the PCIe link is on-die and linked by the QPI, while on the AMD system it has it go through the northbridge. This could account for such a performance difference between the i5 and X4 840. For games and 3D performance, it looks like the i5 can compete, even with less real cores.
Single-threaded SuperPi 1M is where Intel is king. The i5 655k was able to calculate 1 million digits of in half the time that it took the X4 840. Although we are only talking about a ~13 second difference, in the world of extreme benchmarking a lot of boints might be hanging in the balance. It looks like the lack of L3 cache might play a role here as well as the X4 840 overclocked to 3.8 GHz is slightly slower than the stock X4 965. The 400MHz clock advantage does not seem to outweigh the lack of L3 cache in this benchmark.
Results here should mirror the SuperPi 1M results above. It is interesting that in SuperPi 32M, sheer MHz seemed to have more of an effect than in 1M. It is clear though that single-threaded applications see a huge performance gain from the i5 655k vs. X4 840. For half the price, users do actually get half the performance in this example.
AIDA64 has replaced Everest: I’ve not got a feel about this suite of synthetic benchmarks yet as I have not been able to test it on a wide-range of different CPUs. The results vary significantly and scores are graphed here, not percentages.
CPU Queen (Courtesy AIADA64):
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…CPU Queen test uses integer MMX, SSE2 and SSSE3 optimizations.
In looking at the results, this test seems to be closely linked to CPU MHz instead vs. number of cores. The Phenom II X4 840 edges out the i5 655k by 6%, not a huge difference, but a win for the AMD.
CPU Photoworxx (Courtesy AIADA64):
This benchmark performs different common tasks used during digital photo processing…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…CPU PhotoWorxx test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Again the results don’t vary by much, but it seems like X4 840 gets the slight advantage here once again, but only by 7%.
CPU Zlib (Courtesy AIADA64):
This integer benchmark measures combined CPU and memory subsystem performance through the public ZLib compression library. CPU ZLib test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Finally, some conclusive results! The Phenom II X4 840 outperforms the i5 655k by 30% in this benchmark. These results are likely due to the extra cores, stronger memory controller and additional L2 cache on the AMD CPU.
CPU AES (Courtesy AIADA64):
This benchmark measures CPU performance using AES (Advanced Encryption Standard) data encryption…AES is used in several compression tools today, like 7z, RAR, WinZip, and also in disk encryption solutions like BitLocker, FileVault (Mac OS X), TrueCrypt. CPU AES test uses only the basic x86 instructions, and…is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.
Honestly, I do not even know what to say here. This graph makes me question the entire benchmarking suite from AIDA. If we assume the results are accurate, then the i5 655k outperforms the Phenom II x4 840 by about 84% in data encryption tasks. I somehow have a hard time accepting that this test is a reflection of data-driven tasks since the x4 840 seemed to beat the i5 655k by about 20% in the 7zip benchmark.
Above is my attempt at a maximum overclock. I did not spend hours upon hours tweaking my setup, but this is a pretty good indication of where you can get this chip with a decent amount of effort. I’d expect another few MHz depending on the batch of processor you receive. Mine was an engineering sample, which tend to be slightly weaker than later batches. 4GHz is, however, pretty respectable for a $100 chip.
For $102 can you really complain? This chip stood up pretty well to the i5 655k in a lot of tests, which is about $100 more expensive. For the price and the number of cores, you really can’t beat the Phenom II X4 840 CPU. My only complaint is the lack of L3 cache. Though this did not affect performance in all tests, naming this chip Phenom instead of Athlon does seem kind of strange. AMD did not jack up the price though, so I do not see this as an act of deception: just be aware of the differences when choosing between a Phenom II X4 8xx and 9xx series chip. With the sort of performance you can see above, there are not a lot of negative things I can say about a $100 product that performs not too dissimilarly from a $200 product.– Matt Ring (mdcomp)