Anisotropic dysfunction

A certain mysterious gerbil has been battling back and forth with Forge in the comments for several weeks now. While Forge is hardly green when it comes to the matter of gerbil warfare, it seems in this case he may well be outnumbered; the gerbil called for reinforcements.

In particular, Wumpus of Game Basement, the man who brought to light the GeForce DXT1 problem, is on the bug hunt again. This time he has found either another hardware issue, limitation, or a very peculiar driver choice on the ATI Radeon—an issue that has made him a very unhappy camper.

In the Beyond3d forums Wumpus has made two posts calling for recognition and understanding that the Radeon does have issues that nearly every reviewer has ignored (or failed to notice). He even has a bevy of pictures to prove his point. All just a click away here and here

The quick lowdown:

  1. Anisotropic filtering and mipmap blending are mutually exclusive. That means bilinear-style mipmap bands whenever aniso is enabled in any form.

  2. Mip-mapping appears to be done per-scanline instead of per-pixel. So the mipmap bands are obvious straight lines running parallel to the viewpoint.

  3. Enabling anisotropic filtering triggers a dramatic pushback in LOD bias.

Who was saying something around here about actually wanting an ATI workstation card in his work machine? It was Two something. Well, anyway, he's crazy. :)

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.