OpenGL FSAA Fix
Not long after the X1000 series was announced ATI issued a patch that reconfigures the memory controller in order to optimise the access patterns under OpenGL when FSAA is enabled. We'll take a look at the result of this path to see what differences it makes under Doom 3.
| Doom 3 - Normal (FPS) | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT - Unpatched | 137.9 | 141.7 | 140.2 | 119.3 | 95.3 |
| XT - Patched | 139.8 | 142.1 | 139.9 | 118.9 | 93.7 |
| XL - Unpatched | 138.5 | 134.9 | 119.8 | 89.9 | 69.0 |
| XL - Patched | 139.7 | 134.8 | 119.4 | 90.0 | 69.3 |
| % Patched performance Increase | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT | 1.4% | 0.3% | -0.2% | -0.3% | -1.7% |
| XL | 0.9% | -0.1% | -0.3% | 0.1% | 0.4% |
The performances of both the X1800's under normal rendering are virtually identical, or at least within the margin of error for a benchmark run, but this is to be expected given the patch was billed as a improvement for FSAA.
| Doom 3 - 2x FSAA (FPS) | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT - Unpatched | 141.0 | 141.0 | 135.8 | 117.9 | 84.2 |
| XT - Patched | 141.2 | 139.9 | 134.7 | 110.0 | 84.6 |
| XL - Unpatched | 137.8 | 130.8 | 111.6 | 81.9 | 62.5 |
| XL - Patched | 137.7 | 129.0 | 110.3 | 80.6 | 60.8 |
| % Patched Performance Increase | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT | 0.1% | -0.8% | -0.8% | -6.7% | 0.5% |
| XL | -0.1% | -1.4% | -1.2% | -1.6% | -2.7% |
Applying 2x FSAA again shows very little difference between the pre and post patch performances; if anything, on balances, it appears to have a slightly detrimental effect.
| Doom 3 - 4x FSAA (FPS) | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT - Unpatched | 139.1 | 127.7 | 102.9 | 72.5 | 52.4 |
| XT - Patched | 140.6 | 136.9 | 121.3 | 92.2 | 68.2 |
| XL - Unpatched | 127.5 | 108.7 | 83.3 | 58.0 | 42.7 |
| XL - Patched | 134.1 | 118.4 | 92.2 | 66.5 | 49.8 |
| % Patched Performance Increase | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT | 1.1% | 7.2% | 17.9% | 27.2% | 30.2% |
| XL | 5.2% | 8.9% | 10.7% | 14.7% | 16.6% |
With 4x FSAA applied we can see some remarkable performance differences between the pre and post patch execution, up to 30% speed improvement for the XT and 17% for the XL, bring the post patched XL scores close the pre-patched XT performances.
| Doom 3 - 6x FSAA (FPS) | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT - Unpatched | 139.1 | 128.3 | 105.2 | 75.0 | 55.0 |
| XT - Patched | 138.9 | 128.3 | 105.2 | 75.0 | 54.9 |
| XL - Unpatched | 123.4 | 102.0 | 77.8 | 54.4 | 35.5 |
| XL - Patched | 123.4 | 102.0 | 77.7 | 54.4 | 35.5 |
| % Patched Performance Increase | 640x480 | 800x600 | 1024x768 | 1280x1024 | 1600x1200 |
| XT | -0.1% | 0.0% | 0.0% | 0.0% | -0.2% |
| XL | 0.0% | 0.0% | -0.1% | 0.0% | 0.0% |
As with Normal and 2x FSAA rendering, 6x FSAA also sees no performance changed pre and post patch.
When taking all of these performances into account we can see that 4x FSAA certainly has improved with these new memory adjustments, but what sticks out is that 6x FSAA on the XT is actually faster than 4x was in the first place, which logically shouldn't be the case as 6x FSAA will always be requiring more samples to be calculated and more bandwidth used - it would seem to be the case that there was actually something wrong with the OpenGL memory optimisation in the first place, and that the patch actually puts it in the region of where it should have been had there not been an issue.
