Testing Methodology:

System Configuration

Motherboard:	ECS KN1 Extreme Socket 939
Processors:	3000+
Contenders...	ATI X700 Pro 128MB (425/860) Cat. 5.6 EVGA 6800 vanilla 128MB (325/700) Gigabyte 6600GT 128MB (500/500) Leadtek 6600GT 128MB (550/560) Forceware 77.72
Hard Drive:	Seagate 7200.7 SATA
Power Supply Unit:	Ultra 520W
Case:	N/A
Operating System:	Windows XP Professional SP1 (with latest updates)
Memory:	Corsair PC3200XL (2-2-2-6)

We reinstalled Windows XP SP1 on our test bench before performing our performance tests. The hard drive was reformatted and windows updates were installed to keep our system up-to-date. We then upgraded our system from DirectX 9.0b to 9.0c version. Everything in the BIOS was set to default expect for memory timings and on-board LAN was disabled. Please note the clock speeds of these cards again to clear any confusion you may have.

AA and AF settings were changed globally as necessary through control panels and the image quality was set to "High" for all our tests. We decided to leave the games at their default settings for the most part; however, we did make some changes during certain points. Please refer to the list below for information on these changes:

Half-Life 2: The game was tested with internal AA and AF settings with everything set to high.

Far Cry: For this game, we used Ubisoft?s recalled 1.31v patch for the "Training" demo. The image quality was set to "High".

Unreal Tournament 2004: For this game, we used an open-source benchmarking utility, UMark, and BR-Colussus map for our testing. All game settings were set to either normal or high.

Doom III: We used Doom III?s Demo 1 for our testing at "High" settings. As you may or may not know, Doom III requires two runs of the demo to yield correct scores. The first time it caches the textures of the game while the second time it measures the performance more accurately with all the textures already pre-loaded. Obviously we used the pre-loaded, second run scores for our comparison purposes.

Also, v-sync was turned off in all our tests and each test was run three times to confirm the numbers. The average of the three runs was taken for competitive analysis.