GL_ARB_texture_cube_map with RADEON 9800 PRO


I bought a RADEON 9800 PRO video card the other day, and I tested the dynamic cubemap demo (GL_ARB_texture_cube_map) that I wote before.
Then I was disappointed…FPS in the demo marked only 20.
It marked higher than 100 with GeForce4 Ti 4200, but with RADEON 9800 PRO, the performance extremely falls.

I guess the CATALYST driver doesn’t be tuned up with glCopyTexSubImage2D() or GL_ARB_texture_cube_map, but I don’t understand how to solve this performance issue. Please help…

It’s hard to say without more details; what i can just say is that i’m doing real-time dynamic cube mapping on my Radeon 9700, and it’s very fast (> 100 fps), so you must be doing something wrong. And yes, i’m also using glCopyTexSubImage2D.


What’s your cube map texture internal format? You should be using a sized format (ie GL_RGBA8 instead of GL_RGBA), and it should match your framebuffer format.

To find out the bottleneck:

Take out a single thing. Like, remove the call to CopyTexSubImage for example (comment it out). Measure frame rate. Repeat with other changes; i e bind a 2D texture instead of cube map, etc. Just comment out single lines or very short snippets, letting most of the program run as usual.

This ought to give you a good insight into where the time is going. Then focus on fixing that part.

Thanks for your helps. I solved the issue at last.
I created the cubemap texture with internal format “RGBA”…

The answer is;
glTexImage2D(g_Face_Target[face], 0, 1, CUBE_MAP_SIZE, CUBE_MAP_SIZE, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);

Huh? The third parameter is the internalFormat. You sent 1 which is neither GL_RGBA nor GL_RGBA8, but a one component texture.