by Josh Walrath

Quake III Arena

     At the time of testing, Doom 3 was not available, so I relied on Quake III for some OpenGL tests. This is not the average Quake III demo1 though, this is a recorded demo of 5 bots and myself playing in the very stressful NV15 map (very, very high polygon count). Even on a high end machine, frames rarely hop over 100 fps (as compared to Demo1 which routinely tops 400+ fps). This test hammers both the CPU and the GPU.

     Trilinear filtering was enabled, as well as 32 bit textures, and high quality rendering.

     While ATI has been hounded as of late about their OpenGL drivers, there really isn’t anything wrong with them. ATI provides very good driver support, but it can be argued that NVIDIA’s OpenGL support is slightly better. At the lowest resolution the 6800 barely pulls ahead of the 9800XT, but once resolution increases, then the differences become very significant. When things finally get to 1600x1200, the 6800 is almost 20 fps ahead of the 9800XT. This is nearly a 30% increase in speed over the 9800XT.

UT2004

     This latest iteration of the Unreal series offers the best UT graphics to date, and does feature DX 9.0 functionality (though it is still mainly comprised of DX7 and DX8 effects). Each botmatch was set to high quality rendering and 12 bots. Dom-Suntemple was used as the arena.

     Something very strange is happening at 1280x1024 for the 6800, but each time I retested it came out to be the same. The other resolutions did act correctly, so pay no attention to 1280x1024. UT has always been a firm stronghold of NVIDIA, and the 6800 does pull ahead in each test. In Direct3D, ATI really does a superb job of keeping up, even though it is much more limited in raw fillrate. When things get tougher though, the lower fillrate really does start to affect performance for the 9800XT.

Halo

     The Halo built in timedemo was used to get some decent raw numbers from these cards. As you know, Halo was ported to the PC and utilizes PS 1.1 and PS 2.0 code paths. Overall the game on the PC does look superior to that on the X-Box, but it is a bit of a hardware hog. I did find out something interesting though, and I am waiting for confirmation. When I look up the benchmark results of the 6800, Halo has listed that the code path used was PS 3.0. Of course, the code path for the 9800XT was PS 2.0. I will update this when I find out more. High quality settings were enabled as well as 8x AF. No AA was used.

     Since Halo is chock full of PS 2.0 effects, it is more math dependent than fillrate dependent. As such, the 6800 with its extra quad outpaces the 9800XT. Also, if PS 3.0 plays an effect here, this could also constitute an increase in performance. All resolutions are very playable on the 6800.

X2 Benchmark

     This little number features some really good effects, as well as a lot of nifty spaceships and texturing. High quality settings were used, but no AA or shadows were enabled. AF was set to 8X for both cards.

     NVIDIA again takes the prize here as the 6800 is significantly faster across the board. Image quality between the two were indistinguishable. Both rendered the scenes as they should, and I saw no anomalies between runs. A whopping 20 fps difference between the two at 1600x1200 is no mean feat. People who bought the 9800XT at $450 are probably crying that only 9 months later their card is being slapped around by something that costs $150 less.

Next: Even More Results


 

Copyright 1999-2004 PenStar Systems, LLC.