Question: free MSAA? edit: and also free AF?

[Xmas]

But yes, I suppose you could do some tests and examine what the average degree is, and optimize for that...

I think it's currently around 6x; just don't quote me on that one

More like 3:1, I would guess. And that's only the average degree of anisotropy, with all the optimizations the average number of samples (not considering trilinear) should be less than 3.

 

[Ailuros]

Two brainfarts in one thread is a bit too much LOL. In any case yes the 6x average is way too high. I don't know what exactly I was thinking while typing, when I actually had 6x samples maximum in mind. Gee I even tried to elaborate on my own stuff, actually repeating the same mistake *shoots himself*

And that's only the average degree of anisotropy, with all the optimizations the average number of samples (not considering trilinear) should be less than 3.

That was one of the spots I was actually targeting at. With let's say just optimized trilinear, the angle dependency and MIP optimisations, AF should be already in relative terms fillrate "free".

http://www.3dcenter.de/artikel/forceware_61.77_66.00/index4.php

The article needs a translation and there are obviously the no-AF scores missing. Time to run a couple of tests myself I guess.

 

[Ailuros]

I used UMark's Suntemple flyby mode in 1600*1200*32:

CP "high quality"

1xAF = 145.33
4xAF = 73.89
8xAF = 63.55

CP "quality"

1xAF = 156.21
4xAF = 119.11
8xAF = 111.40

I've no idea how zeckensack's Archmark exactly handles textured fill-rates, but here is what it looks like (those don't have to be necessarily accurate):

Trilinear/quad texturing

CP "high quality"

1xAF = 498 MPixels/s
4xAF = 242 MPixels/s

CP "quality"

1xAF = 579 MPixels/s
4xAF = 570 MPixels/s