Unreal Tournament 2004
The diagrams we got on the Torlan level of Unreal Tournament 2004 bring no surprises as all the participating graphics cards, except the GeForce 6600 GT, get the same results, limited by the performance of the central processor of the system.
Full-screen anti-aliasing and anisotropic filtering being enabled, any differences between the participants are only apparent since 1280x1024 resolution, which indicates the bottlenecking effect of the CPU. The RADEON X800 PRO is somewhat better than the ASUS V9999 Gamer Edition, which in its turn looks preferable to the original GeForce 6800. The diagram suggests that the original card design certainly lacks fast memory.
The Metallurgy level is less complex as concerns the CPU load, so we see the graphics cards perform diversely in the highest resolution of the pure speed mode already. The ASUS is capable of challenging the RADEON X800 PRO here, but not due to its fast memory – the original GeForce 6800 with slow memory is only 6-7fps back. It’s a negligible lag, considering the absolute speeds are higher than a hundred frames per second. So why such results? Because it is the vertex processors of the graphics card that bear the main load on this game level. The ASUS and the RADEON X800 PRO have 6 such pipelines each, and the GeForce 6800 sample we’ve got seems to have 6 vertex pipelines, too, and this explain the small difference in the results (as you remember, off-the-shelf GeForce 6800 graphics cards have 5 vertex processors only).
The simplicity of the Metallurgy level helps the ASUS V9999 Gamer Edition to leave the RADEON X800 PRO farther behind than on the Torlan level with its complex textures and open environments. Note the big difference between the GeForce 6800 and the ASUS, which once again emphasizes the importance of fast memory when the card’s performing anti-aliasing and anisotropic filtering.