Why is FX 2xAA worse than before?

Randell

Senior Daddy
Veteran
ANybody here who has previewed the card? I understood that one good point of nVidia AA was that from the Gf3 on the 2xMSAA was RGMS. So whats happened with the FX - driver bugs? A performance optimised solution or just back to OGMS that has hit edge quality?
 
I'd say it's most likely a screen capture issue, just like Quincunx and 4x9 are often hard to capture on a GeForce4. We need somebody to check if the screenshots match what is seen is game.
 
well Brent is here and I'm sure he said it was very noticable whilst playing UT2003 - but it could be screen capture. Wouldnt that then indicate a filter was being applied ala quincunx rather than MSAA?
 
I've put a post about this in the other thread (hardocp review) Other modes are also displaying wrongly. Both 6xS and 8xS in the NFSHP2 screenshots are worse than they are on a gf4 using the registry tweak to force the new modes.
 
I had a theory of custom downsampling ala 3dfx, but then I realized I have no idea how that related to the performance on the Voodoo cards, so I don't know how feasible that is. If it hits performance like regular 2x AA, what would be the benefit?
 
Randell said:
well Brent is here and I'm sure he said it was very noticable whilst playing UT2003 - but it could be screen capture. Wouldnt that then indicate a filter was being applied ala quincunx rather than MSAA?
It doesn't necessarily mean anything. It's not unfeasible for the downsampling to be done at scanout (though this would require a higher-resolution back buffer), as it is for Quincunx and 4x9 on the GeForce4.

As a side note, this could actually explain very nicely the large performance hit at 1600x1200x32 w/ 4x FSAA (memory footprint too large for decent performance).
 
Brent said:
i will check it again, but thats how it looked in the game itself, in ut2k3...
What version of ut2003 were you using? Apparently, some version (have to check which one) didn't properly support 2xMSAA on GF4 either.

As for getting a screen capture of 2xMSAA, I just wrote this in Uttar's thread:
Isn't it so that, with 2xMSAA, the actual combining of the two multisample buffers happens on the ramdac and not in the framebuffer? I remember never having been able to get a proper 2xMSAA screenshot on my GF4Ti with using the print screen button, as it only grapped one of the two multisample buffers, i.e. "half" the final AA effect.

ta,
-Sascha.rb
 
Back
Top