Protocol
NVLink
NVIDIA high-bandwidth GPU interconnect for peer-to-peer and coherent-ish memory traffic inside a node or NVSwitch domain. Generations differ in lane rate and topology (mesh vs switched). NCCL and CUDA P2P performance hinge on healthy NVLink.
Tags
gpuinterconnectp2pbandwidth
Sources
- NVIDIA NVLink documentation
Related protocols
NVSwitch
Lab readySwitch fabric chip that stitches many NVLink ports into all-to-all GPU topologies in DGX/HGX-class systems. Enables full-bandwidth collec…
PCI Express
Lab readyPCI Express — packetized serial I/O with lanes, LTSSM link training, and TLP/DLLP layers. Gen5/Gen6 rates and retimers define reach; ACS …
NCCL
Lab readyNVIDIA Collective Communications Library implementing allreduce, broadcast, and related GPU collectives over NVLink, PCIe, and RDMA. Topo…
GPUDirect RDMA
Lab readyNIC DMA path that reads and writes GPU memory directly, bypassing host bounce buffers for RDMA verbs. Requires peer-to-peer BAR mapping, …