Mellanox OFED GPUDirect RDMA
The next major advancement in GPU-GPU communications is GPUDirect RDMA. This new technology provides a direct P2P (Peer-to-Peer) data path between the GPU Memory directly to/from the Mellanox HCA devices. This provides a significant decrease in GPU-GPU communication latency and completely offloads the CPU, removing it from all GPU-GPU communications across the network.
Mellanox is currently providing an alpha release version of MLNX_OFED software stack with support for this feature. For additional details and to request access to this release for evaluation, please contact firstname.lastname@example.org, or visit the Mellanox Community Site for additional support and updates.