I'm writing a terrain rendering library that currently uses OpenGL as it's rendering api. I'm testing it on a Radeon x700 and a Radeon HD 4200 integrated chip. When I use texture borders(which eliminates some artifacts between adjacent tiles in the map) the chips seem to drop into software rendering mode and it runs at < 1 fps. Without the border the text program runs at ~150 fps. I use 32 bit RGBA textures that are mipmaped, and linear interpolation between mip maps. I use GL_CLAMP_TO_EDGE for texture wrapping. Is it expected that both these chips don't render well with texture borders? I'm using the latest version of drivers on the x700, and I'll have to check on the 4200. If these chips don't handle the texture borders well I can make some changes to account for that, I'm just wondering if it's one of my settings that's causing problems.
Thanks for any help
are you using non power of two textures? there might be some limitations with clamp.
I tested with 512x512 textures(514x514 with the border), I'm not sure if internaly a 512x512 texture with a border counts as power of 2 or not, since the texture is really 514x514.
the 1-texel texture border is not support in the HW you have.
can you just allocate a NPOT texture 514x514 and use the wrap mode clamp to edge ?
Pierre B.
I tried a NPOT and it works but there's about a 20% drop in frame rate, on the machine I tested at least. I think I'll tile my textures at 510x510 and generate a 1 pixel border(though I won't use the border extension) to end with a 512x512 texture and modify the texture coordinates so it's mapped to my terrain in the appropriate manner so I can get the greatest rendering speed out of the card. Thanks for your suggestion.