never heard about this... but I don't need it 

4x FSAA with LOD -1.5 looks great  (even in 640x480 16bits, looks better in 800x600 and awesome in 1024x768 but it begins to run quite slow).
textures are sharp


but I'm missing some true trilinear filtering
the vsa/100 seems to support "Mipmap dithering" only  ???
i think it looks bad, so i disable it
why do we have some half assed trilinear hack, a la detonator 52.16, and not true trilinear

Not to mention that DeltaChrome will use some really important features for movies video output, and filtering. You can find informations on Deltachrome on www.s3graphics.com.

Perhaps a card that would be trully state of the art, as the Voodoo 5 was, is making the first steps... S3 Deltachrome...

I never managed to enable trilinear filtering on voodoo2 (latest official drivers, windows 98)


gl_texturemode_linear_mipmap_linear   does nothing in half-life / CS  or in quake 1/2/3 based games  ???

I don't mind about double pass or single pass,  Counterstrike is fast enough : o

How can Vertex based Anisotropic Filering be realized on a VSA-100? Here it is:

For those who don't know:
One of the main techniques to reduce texture aliasing is MIPMapping, here we got blur in a certain direction in the picture as a side effect. You see, MIPMapping works only in ONE perspective/direction. Anisotropic Filtering plays its part here because it can filter the picture in MORE than one direction (by filtering the picture in different angles) and so you won't get blurry effects anywhere in the picture. Naturally you'll get a performance hit by activating AF..let's see what a VSA-100 could realize..without a slow-down..

VSA-100 AF in OpenGL:
Instead of the GL_EXT_texture_lod_bias or GL_SGIX_texture_lod_bias (both for MIPMapping) the 3dfx_multisample extension is adressed.
You may see, to realize the filtering the same subsamples which RG-FSAA makes use of are used for AF, that would make possible up to 16xAF (!) combined with 2xFSAA without any performance hit. For that 8 texture-samples (two time-scanned for FSAA by jitter-process -> becomes 16xAF) are used (the subsample quantity corresponds here to the jitter-sample quantity -> therefore this would be called 8x2 AF).
For 4xFSAA we got 8x4 AF (because 4 subsamples have to be managed by the jitter-process), that would give us 32xFA combined with 4xFSAA but with a surely performance hit on a Voodoo5 5500. Bet the Voodoo5 6000 would do the job prompting..
So when using AF by T-Buffer you ALWAYS HAVE TO use RG-FSAA, otherwise it couldn't work (could also be a course why 3DFX didn't gave us control for AF).
Driver developers should implement a usable GL_anisotropic extension, but I found this one's already implemented in the last 3DFX drivers but can only be seen and used when the game makes use of the T-Buffer. As there's actually no OpenGL based game doing so, you can't make use of AF in OpenGL on your VSA-100 board
..maybe Daniel Borca knows more about the OpenGL case..

VSA-100 AF in Direct3D:
Direct3D can't make use AF on VSA-100 because the T-Buffer functions have never been implemented in DirectX.

VSA-100 AF in 3DFX GLide:
Well, the GLide-API wasn't developed for AF at the very end of 3DFX. (You'll see, the 3DFX driver developers maybe didn't gave you FA driver control because there wasn't a game for! At least I think so..maybe AF control was planned for inside the games..)


Remember, this is all theoratically stuff, I don't think anybody could manage a driver which gives us control for AF! Otherwise it's also possible that our well-known VSA-100 FSAA automatically uses AF as described when activated. Here I would get Blazkowicz's opinion (no need for AF because FSAA works fine enough)...but as I know/understand it you must have a native support for the T-Buffer in the game to make use of AF.
Show me your opinion about that..there goes the madness..!

Refering to Trilinear Filtering (@ Blazkowicz):
Stand-alone Voodoo cards are able to work with trilinear filering but they can't do it in only one rendering round. Therefore performance would break in by 15-30% (cards which do TF in one rendering round will only get a hit by 2-3%). Some of you may have seen this case on a Matrox G400. That's why 3DFX didn't gave us support for TF in the driver. To excuse 3DFX, they gave us MipMap-Dithering instead...doesn't works as fine as TF but better than nothing.

Well, but let's talk again about AF...really wanna know what you think about!

According to AF, i found this here, maybe it can give you a piece of mind (but I'm sure you read it already  :

"Full-scene spatial anti-aliasing, motion blur, depth of field effects, soft shadows, and reflectance blur all have one important thing in common: From a strict
technical sense, they are all variations on anti-aliasing. [...]
The T-Buffer also opens the door to more effects that we have covered here, because the T-Buffer is a tool, rather than a technique."
(Taken from 3DFX T-Buffer White Paper)

For everybody out there:
When we look at Spectre & Fear future 3dfx products we'll recognize they would have been even able to work with 128-tabs AF in 8 rendering rounds (That would mean 128xAF combined with a sort of 8xFSAA...dunno). As VSA-100 was just slipped in to compete with nVidia, most advantages of the Rampage development resource group were put into the shortterm arranged Venom development. So T-Buffer is, indeed, just a "budget variation" of M-Buffer (all T-Buffer/Supersampling enhancements & Multisampling effects), which was meant to be the one working on rampage types.
That's why we can state that it actually is possible to work with AF on VSA-100 based boards.

Quote:


yeah...that's why 3dfx didn't got it..er..? so it's a question which sample scanning method has to be used for the elliptical filter on VSA?