NV40 Stochastic organized grid?

[AthlonXP1800]

I never heard of Stochastic organized grid, I cant find any information about it in great detail like white paper on internet, I wonder will this method much better than ATI and 3dfx's RGSS? :confused:

[gokickrocks]

Originally posted by AthlonXP1800
I never heard of Stochastic organized grid, I cant find any information about it in great detail like white paper on internet, I wonder will this method much better than ATI and 3dfx's RGSS? :confused:

its random jittered supersampling...takes a big performance hit, but if it runs at decent frames, im sure it will look real good

[silence]

Originally posted by gokickrocks
...takes a big performance hit....


any idea how big?

[Viral]

By the sounds of the NV40's speculated massive bandwidth, this is a wise idea... Nvidia need to gain back the IQ crown, and they need something to use that bandwidth at the same time.

[nobie]

why the big performance hit? surely the overhead to generate a random sample pattern for each frame can't be so great (unless it's a different pattern for each frame for each pixel)

[demonic]

I found a good reference for the different methods of AA: Supersampling, Adaptive Sampling, and Stochastic Sampling take a read here:

http://astronomy.swin.edu.au/~pbour...ering/aliasing/

Also check this out:

Without AA:

http://www3.sympatico.ca/tina_li/graphics/graphics/nonhier.png

With Stochastic Supersampling:

http://www3.sympatico.ca/tina_li/graphics/graphics/nonhier_jit.png

Link here : http://www3.sympatico.ca/tina_li/gr...tialiasing.html

[sxotty]

YOU KILLED it :) ah well. I will look another day when it lives again

[nobie]

try here:
http://www3.sympatico.ca/tina_li/graphics/graphics/antialiasing.html

16 samples. an interesting anomaly is the stars dissapeared completely. not sure what reason for that is.

[StealthHawk]

Originally posted by nobie
why the big performance hit? surely the overhead to generate a random sample pattern for each frame can't be so great (unless it's a different pattern for each frame for each pixel)

And the answer.

Originally posted by gokickrocks
its random jittered supersampling...takes a big performance hit, but if it runs at decent frames, im sure it will look real good

[Lezmaka]

Is there any reason why they couldn't do something similar for multisampling?

[tEd]

Originally posted by Lezmaka
Is there any reason why they couldn't do something similar for multisampling?

no , 3dlabs does have some kind of stochastic,adaptiv,intelligent multisampling AA up to 16 subsamples in their wildcatsIII series

http://www.beyond3d.com/articles/wildcatiii/index.php?page=page4.inc

it looks nice

[Uttar]

Yeah, frankly, I fail to see why it should be that much more expensive than traditional MSAA.
It costs more transistors, and is harder to implement. But beside that...

I think ( that means I'm not sure ) that 3DLabs' solution is not really "fully" stochaistic. What I mean by that is that IIRC, they use a grid which they offset a bit randomly - they don't really create a random grid.

Let me repeat I'm not sure though.


Uttar