Using an fbo to render to cubemap faces

I am doing dynamic (per-frame) cube map environment mapping by the usual 90 fov, 6 renders, using a framebuffer object. I bind each cubemap face to the fbo and render 6 times, changing the camera orientation and bound face accordingly.

All of this works really fast if I don’t bind a depth render buffer to the fbo, currently about 110 fps with all my other effects, but for some reason, when I bind a depth render buffer, it slows way down to about 7 fps. Of course, without a depth buffer and without sorting it renders in a funny order, but that’s not too noticeable for an env map.

Anyway, does anyone know why I might see such a dramatic difference? Maybe it’s skipping some fast path or something if I enable depth testing? I’m using an NVidia 7950. Thanks.

Well since I got no bites, I thought I’d reply myself. I finally switched back to the 93.71 drivers, and the problem went away. Kind of unsatisfying, but it’s all I got right now for anyone else reading this. :slight_smile: