And? Apart from the test application shots, comparable screenshots I've seen didn't suggest that Matrox did a better job on poly edges with 16xFAA than ATI's sparse 6x MSAA (and yes FAA was fully working in those cases). While gamma correction will improve things to a certain degree I severely doubt it entirely accounts exclusively for all improvements or makes the biggest difference.
EER was a term first introduced by Matrox and if it's so "unfair" then they've shot themselves in their own foot marketingwise. Apart from that take 4xS and compare it to 8xS on a NV4x and then tell me how irrelevant the 2*4 vs 4*8 EER difference in real time actually is. Alternatively 2*2 vs 4*4 OGSS.
No-one ever said that a sparsed 6x AA method will equal to 36x OG due to it's sparsed sample distribution; the effectivity though on poly edges in the first case is times higher than with a 6x ordered grid pattern.
Sorry, this image was achieved by blending even and odd frame with 6x AA + temporal enabled I don't see any (easy) way how to emulate 10x/14 AA because of SS/MS combinations.
Sorry, this image was achieved by blending even and odd frame with 6x AA + temporal enabled I don't see any (easy) way how to emulate 10x/14 AA because of SS/MS combinations.
The actual blend operation needs to be performed in linear space, each time you blend AA samples. You've blended in gamma 2.2 space which means it's not going to be the same as ATI's 12x AA.
Unless ATI isn't doing gamma-corrected blending in SuperAA. Which would be fairly surprising.
ok, we'll see I think ATI's output will be very similar (no major differences). I'm looking forward to compare it with CS screenshot. btw here are both TAA outputs, if somebody wants to blend it personaly.
Oops, some misinformation posted by me in that thread I see. It's 3 years old though. MSAA works fine with stencil and properly antialiases stencil shadow edges etc, unlike what I claimed in that thread.