Games Benchmarks - UT 2004 (DirectX)
The first game benchmark we'll look at is UT2004. Although UT2004 is relatively recently released, to our knowledge it is still largely based on a similar build of engine as UT2003, which means its feature set utilisation still marks it primarily as a DirectX7/8 engined title. We are using a custom Firing Squad benchmark demo.

UT2004 (FPS) | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
6200 | 81.6 | 68.0 | 49.0 | 35.7 | 25.3 |
6600 GT | 84.3 | 86.6 | 86.1 | 84.9 | 70.7 |
5200 Ultra | 63.4 | 49.9 | 36.1 | 26.2 | 18.0 |
6200 % Faster Than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
6600 GT | -3.3% | -21.5% | -43.1% | -57.9% | -64.2% |
5200 Ultra | 28.7% | 36.3% | 35.9% | 36.2% | 40.5% |
Although in the high end boards UT2004 is fairly CPU performance limited, as largely displayed by the 6600 GT here, the lower end boards are significantly fill-rate limited, even at low resolutions, when coupled with a high performance CPU such as the FX-55 in this test system. At the highest resolution we see that the performance difference from the 6200 to the 6600 GT is approaching the theoretical Texture / Pixel Shader rate difference. In comparison to the 5200 Ultra, though, the 6200 has as much as as a 41% performance advantage

UT2004, 2x FSAA + 4x AF (FPS) | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
6200 | 60.4 | 45.5 | 32.5 | 23.3 | 16.2 |
6600 GT | 85.0 | 84.8 | 81.9 | 66.6 | 47.7 |
5200 Ultra | 41.9 | 31.0 | 22.9 | 16.4 | 11.1 |
6200 % Faster Than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
6600 GT | -29.0% | -46.4% | -60.3% | -65.0% | -66.1% |
5200 Ultra | 44.1% | 46.5% | 42.1% | 42.0% | 45.4% |
With 2x FSAA and 4x AF enabled we again see that the 6600 GT has a large performance advantage over the 6200, although it is a little less than the theoretical texturing / Pixel Shader performance differences. In comparison to the 5200 Ultra the performance gap opens up a little more, now with the 6200 having as high as a 45% performance advantage, which is likely attributed to the different Anisotropic Filtering algorithm, which doesn't sample equally across all angles, and possibly a slightly more optimised FSAA compression techniques along with its Pixel Shader performance advantages.

6200 (FPS) | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
Normal | 81.6 | 68.0 | 49.0 | 35.7 | 25.3 |
4x AF | 66.4 | 50.6 | 36.1 | 26.4 | 18.4 |
2x FSAA | 76.4 | 60.7 | 42.8 | 30.6 | 20.9 |
2x FSAA + 4x AF | 60.4 | 45.5 | 32.5 | 23.3 | 16.2 |
% Diff from Normal | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
4x AF | -18.6% | -25.7% | -26.4% | -26.1% | -27.1% |
2x FSAA | -6.3% | -10.7% | -12.7% | -14.2% | -17.3% |
2x FSAA + 4x AF | -25.9% | -33.1% | -33.6% | -34.7% | -36.0% |
Looking at the 6200 performances for each of the individual settins we see that the performance has dropped under 60 FPS at 1024x768. In this instance we see that 2x FSAA has a smaller performance penalty than 4x AF.