Fillrate test in 3DMark2001 with anisotropic filtering

[CosmoKramer]

No aniso: 752.3/1570.0 single/dual
8X aniso: 469.2/920.7 single/dual

So there is a hefty drop in fillrate to say the least. However, the textures in the test seem to be in a plane parallell to the viewport, so should there really be any difference?

(Test performed by a GF3 @ 230/510 using the 23.12 WinXP drivers).

 

[pcchen]

You can test it with 2X anisotropic filtering. I guess the result will be almost the same as 8X anisotropic filtering.

 

[CosmoKramer]

On 2002-02-16 12:48, pcchen wrote:
You can test it with 2X anisotropic filtering. I guess the result will be almost the same as 8X anisotropic filtering.

You are correct! So what is the conclusion?

 

[nAo]

On 2002-02-16 13:51, CosmoKramer wrote:
You are correct! So what is the conclusion?

My ipothesis:
Or those planes aren't ortoghonal to view vector and adaptive anisotropic enters in action (selecting 2:1) cutting fillrate...Or those planes are orthogonal to view vector but the adaptive anistropic filtering can't resort to act like a bilinear filtering.
Maybe it can act like a trilinear..when in this case u'd need just bilinear. Can u force it to not use anisotripic filtering and to use trilinear filtering instead?

ciao,
Marco

 

[CosmoKramer]

Can u force it to not use anisotripic filtering and to use trilinear filtering instead?

No, I'm afraid not (I'm using RivaTuner).