Has anyone tried the test with a very low resolution to find possible bottlenecks outside pixel shading?
Yes, actually (all in DX11-Mode, on my C2D E8500@3,8 GHz, Win7 x64 and HD 5870 with Cat 10.3 preview drivers)
Tessellation off, Shaders: High:
640x400 149,7
1280x800 104,9
2560x1600 40,4
Tessellation on, Shaders: High:
640x400 68,2
1280x800 52,8
2560x1600 27,2 (about two thirds the perf w/o tessellation)
So, there's some pixel work going on besides Tessellation. Let's reduce that a bit:
Tessellation off, Shaders: Low
640x400 149,0 (here, it's obviously my CPU bottlenecking everything else)
1280x800 119,4
2560x1600 51,6
Now, that's some 25is percent more performance with reduced pixel load. Let's try some more geometry.
Tessellation on, Shaders: Low
640x400 68,7 (wow, we've gained 0,5 fps!!
)
1280x800 57,6
2560x1600 32,2
Now, we switch shadows to low and turn off the other options in Heaven (Refraction, Reflektion and Volumetric)
Tessellation off, Shaders/Shadows: Low, Refr./Refl./Vol.: Off
640x400 150,7 (huge gains to be had here
)
1280x800 127,3
2560x1600 59,8 (well, we're touching the magical 60-fps-barrier)
Tessellation on, Shaders/Shadows: Low, Refr./Refl./Vol.: Off
640x400 72,1
1280x800 62,3
2560x1600 36,4
So, after we've escaped my systems limits at 2560x1600, almost regardless of pixel load, tessellation costs between 39 (lowest pixel load) and 33 percent of performance (highest pixel load) in this particular benchmark. When we're not pixel-bound at 1280x800, it's more pronounced at around 50 percent perf-cost. (-50/-51/-52% performance, lowest to highest pixel load).
With AA enabled I've only numbers for 1920x1080 and 2560x1600:
1920:
29,7 vs. 47,5 fps for 4xAA,
25,3 vs. 38,8 fps for 8xAA,
37,4 vs. 63,1 fps without AA.
2560:
20,9 vs. 29,7 fps for 4xAA,
5,3 vs. 6,8 fps for 8xAA,
27,0 vs. 39,7 fps without AA.
AA numbers are older and done with Catalyst 9.12.