In the review of NVIDIA his new GeForce FX 5950 and 5700 Tom's Hardware Guide already expected that they found a few optimizations in ATi his drivers. THG investigated these thing after the NVIDIA Editor day were NVIDIA claimed ATi his Radeon cards did not render everything correctly. You can read that here.
THG found a few irregularities in Aquamark 3 (Explosions not fading out like it ought to be, and smoke being less dense on the Radeon 9800XT). Other irregularities were found in UT2003. THG is still investigating this, and for now they don't want to comment on the optimizations found in ATi's drivers.
The "Massive Overdraw" test in AquaMark 3 aims to test the limits of a graphics card's fillrate. In the explosion pictured below, numerous layers of textures are added onto one another (Particle Effect) and then slowly faded out. It seems that ATi's driver doesn't continue to fade out the textures until the blending ends, instead simply cutting out certain textures when a certain degree of fade-out is reached. This obviously saves memory bandwidth, leading to a higher framerate. In the direct comparison with the image rendered by NVIDIA's FX 5950 Ultra, the result is a darker and less detailed explosion (Frame 5100).
After seeing this Elite bastards examined the found irregularities/optimizations in AquaMark 3, Unreal Tournament 2003 and Halo. For some tests they used the Detonator 44.03 drivers because the newly released ForceWare 52.16 drivers are still not rendering all light effects correctly.
From the images here, it is clear that there are differences between the way ATi and nVidia cards are rendering images in certain scenarios. However, without a reference image it is unfair to label one image as 'right' and one as 'wrong'. One of the AquaMark images seems to be simply a case of one card having a brighter or higher gamma image than the other, which is in no way going to improve performance. This may also be the case in the second AquaMark image, but from the evaluation here it looks as if there may indeed be more to it than a simple gamma issue.
As for the Unreal Tournament issue, this does indeed appear to be a problem with ATi's drivers. To say it is a deliberate ploy to inflate performance seems very harsh, as the issue appears to affect so few textures, but it is something which has been present in ATi drivers for some time, and is hopefully something that ATi will look into with renewed vigour now it has been called into question. Finally, the judges are still very much out on any potential issues with Halo until more information is presented.