Colored Bands and "IQ" measurement

[RussSchultz]

Maybe a more interesting test than the colored bands would be colored grids/checkerboards? Two birds with one stone!

Well, the "tunnel" anisotropic tests already use a checkerboard texture, so aliasing can at least be detected on the base texture.

The pictures I've seen obscure it with the concentric color bands.

 

[OpenGL guy]

Maybe a more interesting test than the colored bands would be colored grids/checkerboards? Two birds with one stone!

Well, the "tunnel" anisotropic tests already use a checkerboard texture, so aliasing can at least be detected on the base texture.

The pictures I've seen obscure it with the concentric color bands.

You can set the colored bands to "full-colored" (as many do) or "tinted". The "tinted" mode allows you to see the checkerboard pattern.

 

[Mulciber]

This is the best anisotropic filtering my 8500 can achive

It bothered me EVERY TIME people would benchmark nVidias near perfect Aniso on the GF4 with this mess and say that ATI somehow has the superior method

 

[KimB]

Maybe a more interesting test than the colored bands would be colored grids/checkerboards? Two birds with one stone!

Well, the "tunnel" anisotropic tests already use a checkerboard texture, so aliasing can at least be detected on the base texture.

The pictures I've seen obscure it with the concentric color bands.

Hence the "base texture" comment. It can't detect if aliasing increases or decreases with various MIP levels.

 

[LeStoffer]

Of course, texture aliasing is the harder of the two to observe, but observations of MIP LOD simply cannot be made without observing the degree of texture aliasing.

So very true. I remember some one offering a short movie snapshot from NOLF (?) to show off texture aliasing in motion. It was pretty damn cool actually. This would also be a great idea to try out next time beyond3d is going to take a closer look at the AF implementations on R3x0 and NV3x!

 

[Reverend]

So very true. I remember some one offering a short movie snapshot from NOLF (?) to show off texture aliasing in motion. It was pretty damn cool actually. This would also be a great idea to try out next time beyond3d is going to take a closer look at the AF implementations on R3x0 and NV3x!

Yeah, I did that with NOLF (it was a download, can't remember where at B3D it's stashed at atm).

 

[Randell]

It bothered me EVERY TIME people would benchmark nVidias near perfect Aniso on the GF4 with this mess and say that ATI somehow has the superior method

I disagree, lots of articles showed teh rotation problem, several reviews called it rip-mapping.

What a lot of people said, was that the speed/quality trade off was worth it, as before the Gf4's drivers fudging LOD's and offsetting which textures were aniso'd (though were they manual settings only via rivatuner?) the Gf4's near perfect aniso was unusable at 8x for many many games due to the perfromance hit. Whereas 16 times is very usable on the 8500.

Personnally I only found one game where the angle issue showed up badly, and that was NWN. SS:SE was one part of one level only.

 

[RussSchultz]

I played around with the aniso tool last night on my 9500, and I guess I wasn't terribly impressed with the output.

I used the grid, set it to 7, removed the colors and used 16x AF. There was obvious blurring, or obvious aliasing (The checkerboard pattern should have receded into the distance into a blur, but it would 'condense' to larger squares). I don't have my GF3 installed right now, so I couldn't compare results.

I hope that's not the best that current hardware can produce.

Maybe there's something wrong with the program? It looks fine at '8' for the grid, but anything below that and it gets all wonky.

 

[Xmas]

Maybe there's something wrong with the program? It looks fine at '8' for the grid, but anything below that and it gets all wonky.

It's not wrong, it's intentional. Disable AF and play around with that slider, I'm sure you'll see how the mip maps are generated (positive and negative numbers mean different behaviour)