My simple but speedy reduction code (runs 106.4GB/s on GTX 295) 106.4/111.9=95.1% to the peak bandwi

Accelerated Computing CUDA CUDA Programming and Performance
19

999 (MHz memory clock) * 2 (DDR multiplier) * 448 (bit width/pin count of the memory interface) / 8 (bits per byte) = 111888 Mb/s