What do I need to get accelerated convolution filtering using the glConvolutionFilter() calls? I've tried the xxx2D() version while writing 600x400 pixmaps with glDrawPixels(), and bandwidth reduced to almost nothing.
Atm I'm using the GeForce 6150 GPU, not the most impressive hardware.

Is there a technique that allows me to get hardware support? Or do the better NVidia GPUs support it?
Thanks for any help,
Robert