Using the ChameleonMark utility from NVIDIA, we can gauge the pixel shader performance for a variety of shaders. In each case, the Atlantis Pro exhibits excellent scores and indicates the focus placed upon this area by the manufacturer. In this case, the card’s programmable shaders and 96-bit floating point precision certainly give it an advantage in this benchmark.

As game engine become more complex and realistic, developers turn to shaders to provide them with a means to create effects which will mimic the real world. Any features which can exploit the flexibility and power of these shaders will surely result in a more immersive gaming experience.

Easily one of my favorite new games this year, No One Lives Forever 2 proved to be an absolute monster of a benchmark for this review. Although the results seem to indicate the same processor bottleneck seen in previous benchmarks, a quick look at the scores reveals something more. As high quality image settings are introduced, the average frame-rate is an unplayable 6.41fps while minimum frame-rate drops to an obscene 1fps!

In order to give the reader a greater understanding of the benchmark results, a visual recording of the gameplay was taken. Note that Quicktime is required to view the movie clips. Click here to view the 3.1MB default quality (no antialiasing, no anisotropic filtering) movie clip and here to view the 2.9MB high quality (6X antialiasing, 16X anisotropic filtering) movie clip.

With the image quality settings set to default values, the game plays exceptionally well at 1600x1200. However, once we enable high quality image settings as seen in the second movie, the game becomes choppy and borders on being unplayable.

Perhaps the most anticipated game of the year, Unreal Tournament 2003, has already become a standard benchmark of choice. Although the characters still lack enough detail in my opinion, the environments are stunning and provide a challenge to even the best systems.

With high detail settings disabled, the demo proves to run extremely well at every resolution. Looking at the results, we do find a strange scenario with the performance. Specifically looking at the results for 1280x1024, we find that there is an enormous range of frame rates throughout each benchmark. In the case of Antalus, the average managed to reach an impressive 110fps. However, the minimum frame rate is dramatically lower at 13fps. At this time, I am looking into this scenario and will post my findings in a future review.

As with NOLF2, I've provided two video clips that illustrate gameplay performance. Click here to view the 4.7MB high quality (6X antialiasing, 16X anisotropic filtering) 1024x768 move clip and here to view the 4.7MB default quality (no antialiasing, no anisotropic filtering) 1600x1200 movie clip.

Viewing each file, we see that this system exhibits higher performance when using a maximum resolution without any image quality enhancements. After playing the demo for a number of weeks, I certainly prefer to run at high resolutions than to drop the resolution and enable anisotropic filtering and antialiasing. In my opinion, the benefits do not outweigh the consequences in that case and the frame rate is too low for my preferences.

Next Page - Issues / Conclusion / Acknowledgements