Processor Clock Rate or Core Clock Rate?

In “NVIDIA Fermi Compute Architecture WhitePaper”, it said “A CUDA core executes a floating point or integer instruction per clock for a thread”. We know that the Processor Clock Rate is two times as high as Core Clock Rate. Which one is the “clock” refer in White Perper?

Another sentence in WhitePaper said “Fermi’s dual warp scheduler selects two warps, and issues one instruction from each warp to a group of sixteen cores, sixteen load/store units, or four SFUs”. Note that if the “clock” means Core Clock, each of the warp scheduler will issue one warp each two Core Clock.

Thanks for your interest!

It’s one instruction per core per processor clock.

So far as I can tell, the core clock does not have any observable performance implications for CUDA. Maybe the overclocking crowd has done systematic studies of what happens when you change the ratio of core clock to shader clock…

thank you~

thank you for your reply.