PDA

View Full Version : COD2 and AA


bagman
02-14-06, 08:34 PM
Hi there
I noticed in Call of Duty 2\main\players\[Profilename] directory there is a .cfg file command " r_aaAlpha " which, I understand, has something to do with various levels of alpha blending used in conjuction with AA.
I'm wondering if this has the same effect as "Transparency Antialiasing" in the nVidia Control Panel ??

Can anyone confirm this ??

If it turns out that it is the same as "Transparency Antialiasing" and you wish to tweak the command " r_aaAlpha ", then I assume you would turn "Transparency Antialiasing" off in the nVidia control panel ??

cheers

Wolfhound
02-15-06, 01:29 PM
r_aaAlpha [0,1,2] - This setting determines the type of alpha blending used for antialiasing (if AA is enabled). 0 is off, which gives the best performance, 1 is dither mode which is the default, and 2 is Supersampling which provides better image quality but reduces performance.


http://www.tweakguides.com/COD2_9.html

bagman
02-15-06, 04:00 PM
So I notice that there is a Supersampling mode under " r_aaAlpha "
Therefore should should you turn off "Transparency Antialiasing" in the nVidia control panel ??

Wolfhound
02-15-06, 04:05 PM
Conventional multi-sampling AA only performs anti-aliasing on pixels detected to be at the edge of a given polygon, incurring less of a performance hit but only applying the smoothing effect of AA to the edges of 3D objects. This is fine for a solid 3D object such as a wall or tree, but what about grass? Or leaves? Or a chain link fence like the thousands found in the world's most action packed fence simulator (Half-Life 2)?

The polygons making up 3D rendered foliage and fences use transparency (alpha) effects to allow realistic sight lines through them, but this means that the pixels making up the visible edge of a blade of grass or leaf are not always on the edge of their polygon, but rather somewhere inside it next to other pixels which have been made transparent. Multi-sampling AA will not smooth out these pixels, meaning that textures that use transparencies will always appear to be jaggy, even in an otherwise glass-smooth 8xAA sampled scene.

The other common method of Anti-Aliasing, SuperSampling, can solve this problem in a way. SuperSampling renders each entire frame at a higher resolution then blends the colour and shade of areas where there is a large amount of difference between pixels, then re-renders the image at the display resolution. Supersampling does not care whether it is blending pixels at the edges of polygons or in the middle of them. The disadvantage though, is that supersampling incurs a huge performance hit, making it difficult to justify.

nVidia's TRSS and TRMS AA technologies provide a more elegant solution. Transparency Adaptive Multi-sampling performs the same AA operations as conventional multi-sampling does, blending pixels on the edges of polygons. It then goes one step further by examining each polygon for pixels with alpha (transparency) information. If it finds them, it applies AA to the pixels adjacent to the transparent areas too.

As far as we can tell, Transparency adaptive SuperSampling performs simple multi-sampling on most of the image as normal, but applies full supersampling to polygons with alpha pixels in them. This results in higher image quality and a larger performance hit than TRMS.

In all, TRSS and TRMS seem like a sensible upgrade to current AA technologies, enabling better all-around image quality without the performance hit associated with full supersampled AA.


Hope this resolvs all your questions ;)