Back to Glossary
GPU Cloud
RDMA
RDMA (Remote Direct Memory Access) allows direct memory access between GPUs without CPU overhead.
Definition
Remote Direct Memory Access (RDMA) is a networking technology that allows direct memory-to-memory data transfers between computers without involving the operating system or CPU. In GPU clusters, RDMA enables GPUDirect RDMA — direct GPU-to-GPU data transfers across nodes, critical for distributed training. Harch Corp's GPU clusters use RDMA over InfiniBand (RDMA-IB) for maximum performance, reducing communication overhead by 10x compared to traditional TCP/IP networking.
Related Keywords
rdmagpudirect rdmardma over infinibandremote direct memory access
Related Terms
InfiniBand
InfiniBand is a high-speed networking standard used in HPC and AI GPU clusters.
GPU Cluster
A GPU cluster is a network of GPUs working together for distributed AI training and HPC workloads.
Distributed Training
Distributed training splits AI model training across multiple GPUs or nodes for faster convergence.