Statements (23)
Predicate | Object |
---|---|
gptkbp:instanceOf |
gptkb:personal_computer
|
gptkbp:connects |
multiple GPUs
|
gptkbp:enables |
high-bandwidth GPU interconnect
scalable GPU clusters |
https://www.w3.org/2000/01/rdf-schema#label |
NVIDIA NVSwitch
|
gptkbp:introducedIn |
2018
|
gptkbp:manufacturer |
gptkb:NVIDIA
|
gptkbp:maxBandwidth |
up to 900 GB/s (per switch, depending on generation)
|
gptkbp:productType |
high-speed switch
|
gptkbp:purpose |
low-latency, high-bandwidth GPU interconnect
|
gptkbp:supports |
gptkb:NVLink
multi-GPU communication peer-to-peer GPU communication |
gptkbp:technology |
gptkb:NVLink_fabric
|
gptkbp:usedIn |
gptkb:NVIDIA_DGX_systems
gptkb:NVIDIA_DGX-2 gptkb:NVIDIA_DGX_A100 gptkb:NVIDIA_DGX_H100 gptkb:NVIDIA_HGX_platform data centers AI supercomputers |
gptkbp:bfsParent |
gptkb:NVLink
|
gptkbp:bfsLayer |
6
|