I’ve got a Radeon 9700Pro card and when I use linear filtering on textures and clamping with tex. coordinates below 0.0f or above 1.0f, the part of the texture that is not in the 0.0f-1.0f dimension (where the texels are “stretched”) is half the brightness it should be. I tried it on other cards and it works fine (GeForce 4 Ti, GeForce 2 MX…), only on this version of Radeon it does not work.
If you’re using GL_CLAMP wrap mode, is sounds like the Radeon has the correct behaviour and not the GF. Texture coordinate 0 and 1 is the edge between the texture data and the texture border, and with a black border color, the resulting color is half as bright as the texture data on the edge. Sounds like you want GL_CLAMP_TO_EDGE instead, which clamps the texture coordinates to the center of the edge texels. Check the spec for more information on different wrap modes.
[This message has been edited by Bob (edited 09-25-2003).]
Bob is right, this is a nvidia bug. They know about it but leave it as it is since so many apps rely on the wrong behaviour. If you have newer nvidia drivers there should be a checkbox you can tick in the control panel to enable conformant GL_CLAMP behaviour.
Thank you very much for your reply, I’ll try it.