AMD Radeon 6000 Series Image Quality Analysis
HQ vs Q Benchmarks
The theoretical analysis earlier was useful to establish what is happening, but now we need to look at games. Since the vast majority of published reviews out there use the default values from each vendor, the first order of business is to obtain hard numbers as to whether there’s a significant performance difference between Q and HQ. To get a nice spread of games I’ve used one from each API (DX9, DX10, DX11, and OpenGL). Here are the results:
Starting with the 5770, we see essentially zero performance difference in any of the tested games between Q and HQ.
Next up is the 6850, which shows a measurable performance increase from Q in both AvP 3 (DX11) and Crysis (DX10).
And finally we have the 470, which shows a large performance change in Wolfenstein (OpenGL). It’s possible this is caused by the LOD clamp setting if the game uses a negative LOD bias, but it’s still legitimate to point out since anyone that cares enough about image quality will be using this setting anyway.