How to get more Gflops ? :)

Accelerated Computing CUDA CUDA Programming and Performance
21

I’ve used my own custom scripts as well to generate CUDA code, it works fine, but I’m sure the compiler will be smarter itself in the future, seeing NVidia did make Cg very optimizing.

22

My stock 8800GTS 512MB gives 412 GFlops with this benchmark!
This is with CUDA 1.1 and Windows XP.

Quite impressive. :)

Thanks Simon.

Fernando