NVIDIA is the target of a lot of backlash the past week, since releasing updated PhysX drivers that optimize higher-end GPUs to handle the workload. The drivers themselves aren’t the problem, but rather inflated 3DMark Vantage scores. Because the benchmark utilizes PhysX libraries in a few tests, adding in acceleration can increase the overall score by a fair margin, putting NVIDIA cards right into the leaderboards.
In a normal system, the algorithms are sent straight to the processor, and as you are probably aware, the processor isn’t so good for specific calculations like this, whereas the GPU excels, and that’s evident by running this test. Many are calling NVIDIA cheaters, but I think they are far from it. The physics calculations have just shifted from the CPU to the GPU, where they calculate faster.
Why people care so much about high 3DMark scores is beyond me, but what many are overlooking is the fact that this driver not only inflates what is apparently the most important number in computing, but it also opens up PhysX capabilities in any of the games that support it, including Unreal Tournament III. I gave this a quick test last night and was impressed. Finally, cool physics done from a GPU in a game people actually play.
Right now it’s limited to higher-end GPUs, and from quick tests, I believe it might be due to the amount of processing power the physics actually need. I’ll be testing it more, and report further when I have a few findings. If you have a 9800 GTX or GTX 260 / 280, feel free to give them a go.
During the benchmark install, a runtime library is updated to allow the test to run on the GPU and then during the test, it addresses the benchmark DLLs to the GPU instead of the PPU or CPU. Nothing within the benchmark is changed at all. No software libraries or even a line of code changes in the benchmark whatsoever. The only thing that changes is that installer, nothing else.