Click on the posts !

• Five basic performance advice for porting kernels to the GPU
• The cost of communications
• Some architectural differences between AMD and Nvidia recent server GPUs