For all you on a 68xx graphics card, I ran some games with both driver versions on a 6850 to see whether the improvements seen with the 6950 would be similar with one of these cards. Overall, I think that these drivers do improve performance somewhat, though not as muc as AMD claim (there was a larger performance increase for the 6900 cards). However, even a bit of performance is always better than nothing.
The universal resolution for all the games tested was 1920x1080 and the settings for these games were all set to the highest, with maximum AA and AF and everything else. Advanced DoF was not used in Metro 2033, and APEX PhysX was not used in Mafia II. I ran the built in/associated benchmark tool for all the games with the exceptions of the following - FPS measured using FRAPS:
Bad Company 2 (first 5 minutes of Upriver)
Metro 2033 (first 5 minutes of Prologue)
For Crysis I didn't run the default benchmark, Instead I ran through Assault_Harbor because it better simulates a real gameplay situation. For Just Cause 2 I ran through Dark Tower, and for Far Cry 2 I ran a custom time demo which was about 3 and a half minutes of gameplay in the desert with some explosions and fire. I ran test b for Lost Planet 2 and London for Dirt 2. The rest only have 1 test for the benchmark.
The CPU (i7-920) was running at 4Ghz with HT enabled
and the graphics card was running at 1000/1150 core/memory
And if you don't like looking at and comparing numbers , I did a table that shows the basic improvements in percentages. + shows an increase over Catalyst 11.2 and - shows a decrease. Next time i'll colour-code these, the numbers may get confusing.
Edited by Razi3l - 3/18/11 at 2:35pm