by Josh Walrath

Anisotropic Filtering Performance

     To test this out, I used UT2004 and the 12 bot match on DOM-Suntemple. All settings were at high quality at 1280x1024 resolution..

     As you can see, with both trilinear and aniso optimizations on the 6800, performance is hardly affected. Though the no AF result is a bit odd, the results with AF enabled all fall into place. Visual quality between the two were fairly similar, but the 9800XT held a slight edge. Once I turned off the trilinear and aniso optimizations for the NV card, output was nearly identical.

Antialiasing Performance

     UT 2004 and the 12 bot/DOM-Suntemple map were again used. High quality settings were enabled, as well as 8x AF at 1280x1024.

     NVIDIA has improved its AA by utilizing a 4x rotated grid setup, and it is now close at that setting to ATI’s. I still slightly prefer ATI’s antialiasing quality over NVIDIA’s, but the gap has closed significantly as compared to the ordered grid sampling that the GeForce FX series featured. Quincunx was not tested, as it provides about the same performance as 2x. ATI still has what is arguably the best looking and best performing high AA setting with 6x. Also, ATI enabled temporal AA on their Radeon series of cards, and this has some significant performance gains in certain situations.

     8xs is a mixture of super-sampling and multi-sampling, and as such it provides a little bit of both worlds. The super-sampling does enable better filtering on textures, as well as enabling AA on alpha textures, while the multi-sampling portion helps speed up the overall process and provide good line AA. The only unfortunate part to this is that it is a performance hog. This is a setting best not used in first person shooters, but rather reserved for 3D strategy games and such where frames per second are not as significant. NVIDIA has not radically changed their AA unit since the introduction of the GeForce 4, and as such is not nearly as efficient or as capable as ATI’s AA unit.

NVIDIA Optimizations

     As you know, the 6800 is a very flexible chip, and NVIDIA now allows the user to force many parameters that they were not allowed to touch before. The three big ones are: forcing trilinear filtering, trilinear optimizations, and anisotropic optimizations. Throughout my tests I left the optimizations on, but did force trilinear. I thought it would be a good idea to see how much of a performance difference these optimizations make on overall gameplay.

Halo

     With the optimizations set to on, every once in a while I could see where the filtering was slightly off and showed a bit of banding, but it was only in certain areas with walls and the ground at an odd angle. Otherwise the optimizations were not terribly apparent, especially in the middle of gameplay.

     There is not a huge change in overall performance, but with the optimizations disabled I was not able to see any filtering problems in the game at all. If it were me, I would run with the optimizations disabled, as performance is not that bad at all.

UT 2004

     I did much the same for UT 2004 as I did with Halo. I again utilized the 12 bot/DOM-Suntemple setup.

     Disabling the optimizations led to around a 6% decrease in overall performance. In gameplay the optimizations did cause some filtering problems, but these were pretty minimal and the user has to really look out for them. With these disabled, I saw no filtering problems whatsoever. Again, in terms of overall performance, most people might want to leave these optimizations disabled and get the full visual effect.
 

Next: Overclocking and Game Impressions

Copyright 1999-2004 PenStar Systems, LLC.