In Game 3 test the scenes acquire much more complex geometry: the Game 3 scene consists of 280,000 polygons against 150,000 in Game 2 test. Additional polygons are used to create realistic hairs on the main character’s head. Even though 8 vertex processors of the RADEON X1800 XT work at 625MHz and this frequency should provide the newcomer with a pretty significant performance advantage, it works slower than in Game 2 test and rests at the level of GeForce 7800 GTX. When we enabled FSAA and anisotropic filtering, RADEON X1800 XT makes it to the head of the race, but the gap between the leader and GeForce 7800 GTX is again much smaller than in the second gaming test. RADEON X1800 XL yields to GeForce 7800 GT in pure speed mode, and in eye candy mode with enabled FSAA outperforms the competitor only in high resolutions.

Game 4 test is a much more complex benchmark than the Game 2, Game 3 and especially Game 1, because it uses Shader Model 2.0, and besides uses something like HDR. The total number of polygons per scene reaches 800,000. It might seem like pretty favorable conditions for the RADEON X1000 architecture. However, the results of RADEON X1800 XL do not prove this point: it works not any faster than RADEON X850 XT Platinum Edition in the pure speed mode, even though it boasts enhanced pixel processors and 8 vertex processors.

When we enable FSAA, it manages to outperform the previous generation ATI solution, although the performance advantage of the newcomer doesn’t exceed 10%. Moreover, GeForce 7800 GT with a less advanced memory controller runs faster in this mode, probably due to higher computational power. RADEON X1800 XT again stays ahead due to its high working frequencies, however, it doesn’t leave the GeForce 7800 GTX too far behind in eye candy mode. It corresponds perfectly well with the results of our theoretical benchmarks where both RADEON X1800 didn’t perform that well with shaders using complex math1ematical calculations. As for the multi-pass shaders and branching shaders, there are none of those in 3DMark03, that is why RADEON X1800 XT/XL cannot really show off here.

The results of individual tests proved the total result we discussed in the beginning of this chapter. However, we shouldn’t forget that 3DMark03 has already become somewhat outdated, so the performance picture we get for contemporary graphics accelerators from this testing suite is not very objective any more. 3DMark05 should do the job much better, so let’s move on to it now.