» Thu Jun 21, 2012 3:19 am
Here's some testing I did a while back with this:
Screenshots
http://www.imagebam.com/image/84021b163595823
http://www.imagebam.com/image/7733af163595851
http://www.imagebam.com/image/82cd55163595889
http://www.imagebam.com/image/86530c163595914
http://www.imagebam.com/image/958234163595925
http://www.imagebam.com/image/b30f9c163595935
Performance
As the game is so CPU dependent, and performance is bottle-necked by it for me most of the time, I see no point in measuring FPS, as it stays the same in most if not all situations when comparing anti-aliasing modes. So here's some GPU load percentages for you. They were measured in http://www.imagebam.com/image/1b192b163595954, which was giving me constant 80 FPS in all AA modes (evil CPU bottle-neck).
Configuration
CPU: Core 2 Duo E8200 (3.8 GHz OC)
GPU: ATI Radeon HD 6850 1GB (950/1175 MHz OC)
RAM: 4GB DDR2 800
Vsync and iPresentInterval are both turned off. (Only for testing)
GPU Load
NoAA ~70%
FXAA ~79% (+9%)
SMAA ~95% (+25%)
The traditional 4xMSAA seems to be almost on par with NoAA (maybe around 2% more GPU load), but it obviously eats loads of VRAM. For example, playing with 1920x1080 and 4xMSAA increases VRAM usage from ~650mb to ~950.
So if you have enough VRAM and need GPU power for something else, say ENB, I'd go with MSAA and vice versa for SMAA.