Games Benchmarks - DirectX
UT2004
The first game benchmark we'll look at is UT2004. To our knowledge UT2004 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 |
X1800 XT | 83.5 | 82.4 | 81.7 | 80.8 | 78.8 |
X1800 XL | 82.0 | 82.7 | 80.7 | 79.3 | 76.9 |
X850 XT PE | 82.2 | 81.9 | 79.9 | 77.7 | 74.0 |
X800 XT | 80.5 | 81.7 | 78.8 | 76.3 | 72.5 |
X1800 XT % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X1800 XL | 1.8% | -0.3% | 1.2% | 1.8% | 2.5% |
X850 XT PE | 1.6% | 0.7% | 2.2% | 3.9% | 6.5% |
X800 XT | 3.6% | 0.9% | 3.6% | 5.9% | 8.7% |
X1800 XL % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X850 XT PE | -0.3% | 1.0% | 1.0% | 2.0% | 3.9% |
X800 XT | 1.8% | 1.2% | 2.4% | 4.0% | 6.0% |
As we've become accustomed to, UT2004 is displaying a very high CPU dependence, so there is very little performance variation between these high end boards. Only at 1600x1200 do we see a slight performance difference with the X1800 XT eking out a 7% performance difference to the X850 XT PE and the X1800 XL performing 6% higher than the X800 XT.

UT2004, 4x FSAA & 8x AF (FPS) | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X1800 XT | 81.5 | 81.2 | 80.1 | 77.5 | 75.7 |
X1800 XL | 80.7 | 78.7 | 78.0 | 75.5 | 71.4 |
X850 XT PE | 80.7 | 78.8 | 76.9 | 73.6 | 68.7 |
X800 XT | 79.1 | 77.3 | 75.2 | 71.8 | 64.7 |
X1800 XT % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X1800 XL | 1.1% | 3.1% | 2.7% | 2.7% | 6.0% |
X850 XT PE | 1.1% | 3.0% | 4.2% | 5.3% | 10.1% |
X800 XT | 3.1% | 5.0% | 6.5% | 8.0% | 16.9% |
X1800 XL % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X850 XT PE | 0.0% | -0.1% | 1.4% | 2.5% | 3.9% |
X800 XT | 2.0% | 1.8% | 3.7% | 5.1% | 10.3% |
Applying 4x FSAA and 8x Anisotropic Filtering (AF) increases the graphics workload, but in this instance still doesn't move this test from being significantly CPU bound. However, at 1600x1200 the boards are showing small performance differences with the X1800 XT opening a 10% performance gap over the X850 XT PE. Interestingly we see that the X1800 XL also has a 10% performance advantage over the X800 XT despite them featuring the same fill-rate and bandwidth specifications; this highlights the efficiency improvements R520 has in relation to ATI's previous generation and is such that it allows the X1800 XL to outperform the X850 XT PE.

UT2004, 6x FSAA & 16x AF (FPS) | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X1800 XT | 79.6 | 80.9 | 79.3 | 77.1 | 74.7 |
X1800 XL | 78.2 | 78.4 | 77.0 | 73.8 | 68.6 |
X850 XT PE | 80.2 | 77.9 | 75.8 | 71.7 | 63.7 |
X800 XT | 78.6 | 76.8 | 74.6 | 68.2 | 58.1 |
X1800 XT % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X1800 XL | 1.8% | 3.3% | 3.1% | 4.4% | 8.9% |
X850 XT PE | -0.8% | 3.8% | 4.6% | 7.4% | 17.3% |
X800 XT | 1.3% | 5.3% | 6.3% | 13.0% | 28.6% |
X1800 XL % Faster than: | 640x480 | 800x600 | 1024x768 | 1280x960 | 1600x1200 |
X850 XT PE | -2.5% | 0.5% | 1.5% | 2.8% | 7.7% |
X800 XT | -0.5% | 2.0% | 3.2% | 8.2% | 18.1% |
With 6xFSAA and 16x AF there is further onus placed on the rendering and bandwidth capabilities of the boards, but while the other boards are showing some slight graphics limitation at 1600x1200, for the most part the X1800 XT is still CPU bound; despite this the X1800 XT is showing performance gain of up to 17% over the X850 XT PR at 1600x1200. Looking at the X1800 XL next to the X800 XT we see that the performance differences between these two stretch to as much as 18%, further highlighting the underlying architectural changes despite their ostensive similarities; again these architectural improvements allow the X1800 XL to outperform the X850 XT PE despite the X850 having apparent higher pixel calculation rates and more bandwidth.