Here's a list of the parts used for our test setup.

 

Processor Intel Core 2 Extreme QX9650 (3GHz, 12MB L2) @ 4GHz
Motherboard ASUS P5E3 Premium (Intel X48 Express)
Memory Corsair Dominator 2GB 1800MHz DDR3 Kit
Graphics Cards / Drivers ASUS EAH4850 TOP 512MB (680/1050MHz) - 8.6 Hotfix
PowerColor Radeon HD 4850 512MB (625/993MHz) - 8.6 Hotfix
Sapphire Radeon HD 4850 512MB (625/993MHz) - 8.6 Hotfix
Force3D Radeon HD 4870 512MB (750/900MHz) - 8.6 Hotfix
Sapphire Radeon HD 4870 512MB (750/900MHz) - 8.6 Hotfix
XFX GeForce GTX 260 896MB (576/1242/999MHz) - 177.41 WHQL
NVIDIA GeForce GTX 280 1GB (602/1296/1107MHz) - 177.41 WHQL
Cooler Scythe Zipang
Storage Seagate Barracuda 7200.7 80GB SATA x 2
Power Supply Enermax Galaxy 1000W
Operating System Microsoft Windows Vista Ultimate 32-bit w/ SP1

 

 

 

Benchmarking notes

 

Futuremark 3DMark Vantage - Two sets of tests were run on 3DMark Vantage. The first is the 'standard' Performance preset, followed by the highest option available, the Extreme preset. The latest 3DMark Vantage version was used for this review.

Enemy Territory: Quake Wars - Testing was carried out at three resolutions of 1680 x 1050 (on 22-inch widescreens), 1920 x 1200 (on 24-inchers) and card-killing 2560 x 1600 (on 30-inchers). We dropped our custom timedemo test and opted for HOC's ETQW benchmark.

Crysis - The same three resolution settings were used for Crysis. We used Crysis Benchmarking Tool. Game quality was set to either completely High or completely Very High. No AA or forced AF was used.

World in Conflict - Game quality was set at High and Very High. The differences are the level of AA/AF as well as some extra eye-candy processing. On a side note, Medium triggers DX9 rendering instead of DX10.

Unreal Tournament 3 - Processing was set to Intense, while we forced 4x AA and 16x AF from the the ATI Catalyst Control Centre or the NVIDIA Control Panel. UT3Bench was used for this test, and all imaging options were set to the highest available (level 5).

 

 

Temperature and noise levels testing notes

 

To test load temperature, we left HWMonitor opened while looping Futuremark 3DMark06 game tests for about 45 minutes to mark the highest operating temperature. After that, we allowed the system to idle for about 20 minutes to note the idle temperature.

For our noise level tests, we removed the noisy Delta sitting on top of our Scythe Zipang and replaced it with a 1600rpm Scythe S-Flex. The only other audible peripheral would be the fan in the Enermax Galaxy.

We left the graphics card fan in automatic mode first, looping Futuremark 3DMark06 game tests for about 45 minutes while taking note of any changes in fan noise.

Next, we slowly cranked up the fan speed until it is audible and note the percentage.

Finally, we listened to the noise levels for all the cards at 50% fan speed and 100% fan speed. Changing the fan speed on the ATI cards was done by creating a profile and editing the XML file. Changing the fan speed on the NVIDIA cards was much easier with the help of Rivatuner.