Performance
All 3 boards follow the same design principle: they are built around the AMD 761 Northbridge and the VIA VT82C686B Southbridge. The Northbridge commands the communication between the CPU, the DDR SDRAM, and the AGP subsystem, whereas the Southbridge governs the IDE and floppy channels, the LPT, USB and COM ports. Interestingly there is not a single motherboard in the market, which uses the complete AMD 760 chipset, means an AMD 761 NB and AMD 761 SB. This is apparently due to cost reasons. Now: when all of these boards follow the same design principle, we shouldn't see much of a performance difference, should we?

The installation of all 3 boards was without any problem. Within 45 minutes each of the 3 candidates was up and running. It goes without saying that all 3 boards were tested with exactly the same CPU, the same memory, the same components, and software/driver environment to ensure comparability. All tests were done in a monitor resolution of 1024 x 768 pixels, 16 bit, and a monitor frequency of 85 Hertz. We mention this in reference to the recently described considerable differences in benchmarking scores in dependency from the monitor resolution. We also want to mention that we did use only the "official" retail BIOS'. We did not use any tuned-up performance BIOS, which manufactures sometimes offering for reviewing purposes.

We used the following benchmarking software:

These boards are currently the best money can buy. Wouldn't it be boring to run them at the stock speed of 1333 MHz? We decided to unleash the full potential of these powerful mobo/CPU combos and to benchmark right away at 1570 MHz. We guess that this is what most of our readers will do anyway. Why 1570 MHz? Because this was the maximum speed under which all 3 boards would perform all tests stable for at least 1 hour. Initially we intended to use again our trusted Alpha PAL 6035 with a Delta 60mm fan, which is more or less representing mainstream air-cooling. However, it turned out that this HSF did not provide enough cooling power at 1570 MHz and we decided to change to the heavy duty Taisol CGK761CU with a 92mm Delta fan combo mounted on a 60cm to 92cm aluminum adaptor. This combo is providing up to 8C lower core temperatures when under load than the PAL 6035 as shown in this previous review. The Taisol CGK761CU is one of the biggest heatsinks around, only slightly smaller than the Swiftech MC462. The Taisol fit without problem on all boards. All performance settings were disabled. We set the CPU voltage on all boards to the respective maximums. The CAS setting of our NANYA memory was kept at "2". For reference we included in the Passmark "Performance Test 3.4" result table a "P4 generic" base line provided by the author of this software. We don't know what board was used and about the test settings. The P4 scores are therefore just an indication and should not be taken too serious.

Click to Enlarge - Result Table PassMark "Performance Test 3.4

Sisoft Sandra Memory Results - 1333 @ 1570MHz (10x157MHz)

Highest - GA-7DXR Sisoft Sandra Memory Scores at 1333@1570 (10 x 157 MHz)

AMD N-Bench and MadOnion 3DMark2001

Highest - EP-8K7A AMD N-Bench Result (The CPU was actually running at 1570 MHz)

Game and Application Benchmarks - 1333 @ 1570MHz

The "Performance Test" scores are exhilarating. All 3 candidates are leaving the Pentium 4 1400 MHz far behind. At 10 x 157 MHz the EP-8K7A is even overtaking the Pentium's P4 memory scores (RAMBUS!). We see, that basically all 3 candidates are performing similar, with slight advantages here and there for one or the other board. A clear trend, where one board would shine or fall back, is not visible. The EP-8K7A showed the best scores in CPU performance, we can see the board slightly ahead in the mathematics department of the Performance Test as well as leading the AMD N-Bench result. But that did not translate into better scores in the application-oriented benchmarks: The 2 Winstone tests see the Epox even slightly behind. That show that these 3 boards are too close together to declare one the winner.

We were surprised about the good performance of the A7M266. We expected the 2 younger contenders to outpace the A7M266 taking advantage of the fact that the A7M266 is already more than half a year in the market. A similar comparison of AMD 760 boards here has the A7M266 even more ahead of the EP-8K7A. Other comparisons at ZDnet (GA-7DXR: "AMD 760 Performance leader") and at Amdmb.com (which has the Epox slightly ahead) confirm as well, that there is little difference between these 3 boards. We think that differences of up to 5% are well within the margin of error or just fluctuations within repeated test runs. And we definitely will not enter into a discussion about such minuscule differences, because the tests we used and the testing environment are in no way "scientific". A slightly more fragmented drive, a bit lower temperature, a little application or service running in the background, just 1 mouse movement during a test (!) and other, similar minor factors can easily offset such small differences, not to talk about different BIOS or driver versions.