Super-Sampling AA

VOLT!!!!! :D

cheer.gif
cheer.gif
dance2.gif
cheer.gif
cheer.gif


So where ya been? How ya doing?
 
Demirug said:
Why only 75%? Every current chip is able to suppport SSAA.

R3XX: MS vs SS

Yeah i know i mean from a driver support pov

Even that's not true as radeon Mac driver have SS for all r3xx,r4x0 products

I'm just saying how can you ask if it(when) ever will be back as it was never gone in the first place
 
Probably Rev means dedicated hardware support for it, i.e. sparse grid supersampling/hybrid sampling. And not being limited to 1024 res or less.
 
Xmas said:
Probably Rev means dedicated hardware support for it, i.e. sparse grid supersampling/hybrid sampling. And not being limited to 1024 res or less.

I can understand sparse or rotated grid SSAA, but we've already hybrid Super-/Multisampling modes since NV2x. Albeit any portion of SS will have a bigger fillrate impact, I don't think that the hybrid modes limit to just 1024, depending on application.

Also not exactly a very good point when it comes to performance, yet S3 Deltachrome supports in fact RGSS.

Based on the above, does anyone have a clue how Quantum3D is or might be reaching 16xRGSS with their Clarity algorithm?

Oh and Rev could you be just a bit more precise?
 
Demirug said:
Why only 75%? Every current chip is able to suppport SSAA.

R3XX: MS vs SS

4xOGSS for the supersampled shot? What's the performance reading on that one? 20-something? It's not that readable, hence the question.
 
I'm glad to see S3 supports 4x RGSS :) (I don't know any other card that does that besides my voodoo5). Screenshots of Day of Defeat with RGSS vs RGMS (on ATI or NV) would be nice, that's a game where I'd want anti-aliased alpha blending and you could waste power doing SSAA. (on the S3, not on my V5 :()

Xmas said:
Based on the above, does anyone have a clue how Quantum3D is or might be reaching 16xRGSS with their Clarity algorithm?

Oh and Rev could you be just a bit more precise?

You can do 2x RGSS with a voodoo4 (single VSA/100), 4x on voodoo5 5500 (and T-buffer effects), 8x on voodoo5 6000 and 16x on a bad ass Aalchemy with 8 VSA/100.
For example I think that for 4x with a voodoo5 5500, each GPU renders an 2x AAed image with different sample positions and the resulting image is a combination of the two (same to get 2x)
With Quantum3D mercury you have four cards, each being Voodoo2 SLI, and the output is combined to do 4x RGSS 8) (I wish I had this monster to play CS :oops: )

Interestingly enough voodoo4/5 are capable of TemporalAA (ATI meaning, not some motion blur) and it has been implemented by 3dfx fans.


/edit : just found out that "Clarity" is for nvidia based quantum3D stuff, not VSA/100? :)
It may still be the same way of doing it. There was the quad-R300 monster from somebody else too, with "24x" anti-aliasing, maybe the combination of four 6x MSAAed images.
 
Blazkowicz_ said:
I'm glad to see S3 supports 4x RGSS :) (I don't know any other card that does that besides my voodoo5). Screenshots of Day of Defeat with RGSS vs RGMS (on ATI or NV) would be nice, that's a game where I'd want anti-aliased alpha blending and you could waste power doing SSAA. (on the S3, not on my V5 :()

I'd say that you're better off using 2xRGSS on the DC in today's mainstream resolutions and even there performance is way too low in many occassions.

Interestingly enough voodoo4/5 are capable of TemporalAA (ATI meaning, not some motion blur) and it has been implemented by 3dfx fans.

Motion blur aka temporal antialiasing was part of the T-buffer feature-set all along and was not "implemented" by any fans.


/edit : just found out that "Clarity" is for nvidia based quantum3D stuff, not VSA/100? :)
It may still be the same way of doing it. There was the quad-R300 monster from somebody else too, with "24x" anti-aliasing, maybe the combination of four 6x MSAAed images.

I believe that's how the quad chip boards at E&S reach 24x sparsed MSAA, albeit I think they're not scaling geometry in that case (might be wrong though).

Clarity is a patented algorithm from Quantum3D that they developed for their Independence systems containing multi-GPU configs (NVIDIA Quadros since they started cooperating with Q3d).

I was merely trying to make a hypothetical connection of multi-board configs transitioning from the professional simulation market into PC desktop sollutions. 16xOGSS is already possible (albeit very slow) on NV40; question might be if there's a way to get to RGSS (even with a lot less samples) through the drivers in a multi-board config if Q3d can achieve it.
 
Ailuros said:
Interestingly enough voodoo4/5 are capable of TemporalAA (ATI meaning, not some motion blur) and it has been implemented by 3dfx fans.

Motion blur aka temporal antialiasing was part of the T-buffer feature-set all along and was not "implemented" by any fans.

He said: ATI meaning, not motion blur.
 
NocturnDragon said:
Ailuros said:
Interestingly enough voodoo4/5 are capable of TemporalAA (ATI meaning, not some motion blur) and it has been implemented by 3dfx fans.

Motion blur aka temporal antialiasing was part of the T-buffer feature-set all along and was not "implemented" by any fans.

He said: ATI meaning, not motion blur.

yes :p
(samples pattern was already programmable, and modifiable with custom drivers exposing the option in the 3dfx tool, I think they hacked the glide2x.dll and glide3x.dll to change it each other frame.)
 
NocturnDragon said:
Ailuros said:
Interestingly enough voodoo4/5 are capable of TemporalAA (ATI meaning, not some motion blur) and it has been implemented by 3dfx fans.

Motion blur aka temporal antialiasing was part of the T-buffer feature-set all along and was not "implemented" by any fans.

He said: ATI meaning, not motion blur.

It's bad enough that ATI is abusing the term and yes I overlooked that one.

(samples pattern was already programmable, and modifiable with custom drivers exposing the option in the 3dfx tool, I think they hacked the glide2x.dll and glide3x.dll to change it each other frame.)

While I have the highest respect for ATI's (programmable gamma corrected) sparsed sampled MSAA, I couldn't find anything noteworthy in their so called "tempAA". I'd rather grab a higher resolution with as much MSAA samples as possible instead.

If ATI has in the meantime implemented a more sophisticated incarnation of the method (with a tad of hardware support), it might actually turn out to be interesting.
 
Back
Top