Peer-to-Peer access GPUDirect

Accelerated Computing CUDA CUDA Programming and Performance
1

I am looking for GPU peer-to-peer technology .

  1. My doubt is that , If I’ve CUDA 4.0 and CUDA driver 270. is it mean that I’ve GPUDirect2.0
    for peer-to-peer memory access and no additional hardware is required.
    Like in case of GPUDirectv1.0 we require Mellanox ConnectX-2 InfiniBand adapters.

  2. What is the difference between UVA and Peer-to-Peer memory access approach?

Please help me to clarify my doubts