Anisotropic Quality across the board

[PSarge]

In Wavey's GFFX preview I was particularly interested in the Anisotropic filtering analysis using Xmas's tool.
This led me on to wanting to see what other cards looked like running the same tool. Going back to the 9700pro p/review didn't help because I don't think Xmas had written the tool back then.

So does anyone have a collection of pictures showing how the different boards compare? Has there been a thread on here with it all in (I couldn't find one). It would be nice to see images for anything that can do Aniso (GF4/R200/R300/Parhelia/Any others?)

 

[Dave Baumann]

Ummm, yes, I do - I just never got round to publishing them

You can put them into a few categories:

Radeon 8500/9000 & Wildcat VP - only fully Aniso's at 90 degree angles.

Radeon 9700 - fully aniso at 90 degree, partially (up to 8x AFAIK) at 22 degree angles as well.

GF2/3/4 - What you see in the GFFX review at 'Application' mode

GFFX - As you see in the review.

 

[PSarge]

Any chance I could take a look? (a zipfile of pics would do)

 

[McElvis]

Any chance you could put them up in a short article ?

Only if you have the time though. I'm sure there are more important things you are working on at the moment...

 

[Hyp-X]

Ummm, yes, I do - I just never got round to publishing them

You can put them into a few categories:

Radeon 8500/9000 & Wildcat VP - only fully Aniso's at 90 degree angles.

And dropping down to no anisotropy at 45°.
Btw, I've never seen the WVP, I'd expected a high-end card to do better.

Radeon 9700 - fully aniso at 90 degree, partially (up to 8x AFAIK) at 22 degree angles as well.

Make that:
Radeon 9700 - fully aniso at 90 degree, partially (up to 8x AFAIK) at 45 degree angles as well.
And dropping down to approx. 2x anisotropy at 22°.

GF2/3/4 - What you see in the GFFX review at 'Application' mode

+ the Parhelia (with some strange precisity problems).


The interesting thing is that (unlike popular belief) it makes a difference at horizontal surfaces as well.
When the direction of anisotropy is computed correctly the mipmap boundaries are starting to bend in the opposite direction as with no anisotropy.
Cards that only support a few directions for anisotropy the mipmap boundaries are usually flat which means slightly reduced detail at the sides of the picture.

 

[Xmas]

+ the Parhelia (with some strange precisity problems).


The interesting thing is that (unlike popular belief) it makes a difference at horizontal surfaces as well.
When the direction of anisotropy is computed correctly the mipmap boundaries are starting to bend in the opposite direction as with no anisotropy.
Cards that only support a few directions for anisotropy the mipmap boundaries are usually flat which means slightly reduced detail at the sides of the picture.

Yes, Parhelia really does some weird things.

The bending effect can be seen on GeForce cards. Hey, that's the reason my tool has a plane mode

 

[Bambers]

Make that:
Radeon 9700 - fully aniso at 90 degree, partially (up to 8x AFAIK) at 45 degree angles as well.
And dropping down to approx. 2x anisotropy at 22°.

I think its full 16x aniso at 45 and 90 degree angles dropping down to ~2x at 22.5