Driver Performance Difference

Part of the way through the testing of this board S3 released a new set of drivers, so whilst the review displays the performances from the latest set (6.14.10.1960-15.15.08c) available to us we did take some performances with the earlier set (6.14.10.1904-15.14.10e). We'll compare the performance of the old and new drivers to see if any performance gains have been made.



6.14.10.1960-15.15.08c 36.6 23.3 14.0 10.7 6.4
6.14.10.1904-15.14.10e 34.3 22.1 13.4 10.6 6.1
% Increase 6.7% 5.4% 4.7% 0.9% 4.6%
 
6.14.10.1960-15.15.08c 23.4 14.8 8.8 6.3 3.9
6.14.10.1904-15.14.10e 13.8 8.9 5.4 5.7 3.6
% Increase 69.4% 67.1% 63.8% 9.8% 8.1%
 
6.14.10.1960-15.15.08c 54.2 41.0 28.0 18.1 13.1
6.14.10.1904-15.14.10e 52.1 40.8 28.1 18.1 13.2
% Increase 4.0% 0.5% -0.4% 0.0% -0.8%
 
6.14.10.1960-15.15.08c 49.0 36.9 25.6 16.9 12.4
6.14.10.1904-15.14.10e 47.9 36.9 25.8 17.1 12.5
% Increase 2.3% 0.0% -0.8% -1.2% -0.8%

Looking at the results it looks as there there has been a small increase in performance in Half Life 2 in general, however there is a much larger performance increase for the new drivers for 2x FSAA, certianly at the lower resolutions - this could be because the switch to the ordered grid FSAA sampling pattern in the new drivers has less of a performance overhead.

Doom 3's performance has increased a little at the low resolutions, which would tend to indicate that the S3 have concentrated on lowering the drivers CPU overhead a little.