{{define "nvlink"}} NVLink domains · llmgw

NVLink domains

{{len .Domains}} multi-node domain{{if ne (len .Domains) 1}}s{{end}} — racks whose GPUs reach each other over NVLink across nodes

{{if .OnAdminListener}}{{else}}{{template "nav" .Nav}}{{end}}
{{if not .Domains}}

No GPU reports a multi-node NVLink domain. An eight-GPU machine's NVSwitch ends at its chassis and is shown on its node's page; a domain appears here once the fabric manager has joined the GPUs of an NVL72 rack, or an NVLink Switch system, to one.

{{end}}
{{range .Domains}}

{{.Short}}

{{.ClusterUUID}} · clique {{.CliqueID}}

{{.Usable}}/{{len .Members}}trays usable
{{.Joined}}/{{.GPUs}}GPUs joined
NVLink Switch
{{range .Members}}
{{.Node}}
{{range .GPUs}}{{.Index}}{{end}}
IMEX
{{end}}
{{if .Problems}}
    {{range .Problems}}
  • {{.Level}} {{.Message}}
    • {{end}}
{{end}}
{{end}}

A replica that spans nodes is placed inside one of these first: tensor parallelism between its nodes then runs over NVLink rather than InfiniBand. Only trays with every GPU joined and IMEX up are used — a tray short of either would leave the replica a card, or its memory, short.

{{end}}