FBO vs PBuffer mipmap speeds

Hi, I’ve tried to make a small FBO application, which seems to work. However the rendering speeds I get doesn’t quite make much sense to me.

I render the (simple) scene from a different POV to a texture, then render this alternate view in a quad along with the scene from the original POV.

I can toggle between using the FBO or a PBuffer for rendering to the texture.

Now, if I don’t enable mipmapping, the FBO is about twice as fast as the PBuffer. But if I enable mipmapping the FBO is about 2/3 the speed of the PBuffer.

I use glGenerateMipmapEXT with the FBO and glCopyTexSubImage2D and GL_SGIS_generate_mipmap for the PBuffer.

Any ideas as to why this is so? I was at least expecting them to be about same in speed. Am I doing something wrong?

I’m using a 7800GT with 93.71 drivers, the FBO is rendered using a 32bit RGBA color texture, and a 24bit depth RBO.

Hm, could you try glGenerateMipmapEXT with the Pbuffer approach too? I mean, after you copy the texture instead of the SGIS flag…

Hmm didn’t think of that. However, speeds are the same, FBO slow, PBuffer fast.

Thanks for the suggestion though

As far as I know, SGIS flag actually does quite the same, except for glGenerateMipmapEXT has more control over max lod.
So the problem is not here I hope, but, maybe, somewhere in your code…

Just talked to someone else who experiences the same with his own code… He hadn’t found any solution to it either…