UT2003

Our next DirectX game is UT2003, and in this instance we'll use an in-house recorded demo.

UT Flyby 640x480 800x600 1024x768 1280x960 1600x1200
5800 Ultra 85.5 83.9 81.9 78.6 70.6
5900 Ultra 86.0 85.0 81.9 79.6 72.6
Diff from 5800 640x480 800x600 1024x768 1280x960 1600x1200
FPS 0.5 1.1 0.0 1.0 2.0
% 1% 1% 0% 1% 3%

On both boards the performance with normal rendering under this custom demo in UT2003 is largely system limited, with only a slight fill-rate performance drop at 1600x1200. The performance at the highest resolution here is still well above 60 FPS on average.

5900 Ultra 640x480 800x600 1024x768 1280x960 1600x1200
Normal 86.0 85.0 81.9 79.6 72.6
8x AF 84.9 82.7 75.1 64.7 50.1
4XFSAA 80.0 77.8 73.1 66.7 53.9
8x AF + 4X FSAA 78.7 75.1 65.5 52.5 39.1
% Difference from Normal 640x480 800x600 1024x768 1280x960 1600x1200
8x AF -1% -3% -8% -19% -31%
4XFSAA -7% -8% -11% -16% -26%
8x AF + 4X FSAA -9% -12% -20% -34% -46%

UT2003 has proven itself to be quite a texture intensive title, and hence enabling 8x AF often incurs quite a performance penalty. In this case we see that the demo is getting more fill-rate limited and there is a maximum performance penalty of 31% at 1600x1200. The performance is above 60 FPS at 1280x960, but falls below 60 FPS at 1600x1200.

Using 4X FSAA takes less of a performance hit than 8x AF here, with the maximum performance penalty of 26%, but again the performance is above 60 FPS at 1280x960 and below it at 1600x1200. Utilising both 4X FSAA in conjunction with 8x AF has an even greater performance hit, at at maximum of 46%, and is still above 60 FPS at 1024x768 and over 30 FPS at 1600x1200.

While the earlier board comparison here is accurate, in that the same levels of IQ are being represented, we have discovered that under the 44.03 driver the GeForce FX is not rendering the maximum quality that it should be doing. This affects both normal rendering and when AF is enabled. We'll take a closer look at what these issues are and explain in a little more detail what is happening.