The last time I saw something like this it was with a specific 3D benchmark we used to use at Compaq. They had a test case in it that, due to a bug in the video driver, used a non-standard "tweak" that resulted in the test case not crashing. When the graphics card manufacturer fixed the bug and did some optimizations on that portion of rendering, the benchmark dropped by about 20% for several builds until they got the optimizations right. Then when the benchmark maker undid the "tweak" performance shot back up to >18% over the original performance. This is one of the problems with trusting a single benchmark as authoritative and why most of these guys benchmark multiple games/apps and driver versions.
Point being, I suspect WoW has some "tweaks" that improved performance in certain areas with NVidia drivers that have since been optimized in such a way in the video drivers that they are probably more stable (esp. SLI) but defeats the Blizzard "tweak(s)."
[edit]
Reply With Quote
Connect With Us