Mellanox OFED GPUDirect
NVIDIA GPUDirect technology provides the capability for direct communication between the NVIDIA GPU and the Mellanox ConnectX VPI adapter family through the host memory with zero copy and low latency RDMA (Remote Direct Memory Access).
GPUDirect with Mellanox InfiniBand adapters accelerates local GPU to remote GPU communication by 30%, resulting in higher application performance.
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 will soon be releasing a version of MLNX_OFED software stack with support for this feature. Please contact your local Mellanox Sales Representative for details of this release.