AI inference glossary
Hardware

NVL72

Also known as GB200 NVL72, GB300 NVL72, rack-scale system

In plain English

NVL72 is a rack where 72 accelerators share one high-speed fabric, so they behave more like a single large machine than a cluster.

Technical definition

NVL72 is a rack-scale NVIDIA system that places 72 accelerators in a single NVLink scale-up domain rather than in separate eight-chip nodes.

Engineering details

The dashboard specs record NVLink 5.0 at 900 GB/s per chip unidirectional across a scale-up world size of 72, switched through NVSwitch. A conventional node keeps that bandwidth among eight chips and falls back to slower scale-out networking beyond them, so the difference is not raw speed but how many chips are reachable before the fabric changes character.

Why it matters

Techniques whose cost is dominated by collectives change economics inside a large domain. Wide expert parallelism spreads experts across many chips and pays all-to-all traffic for every token, which is tolerable at scale-up bandwidth and often is not across a scale-out fabric.

How to read it in InferenceX

A rack-scale advantage is not automatic. Higher cost per chip has to be earned back, and on agentic traffic the orchestration layer can become the bottleneck before the fabric does, so NVL72 configurations sometimes trail eight-chip nodes on TCO-normalized throughput for models that do not exercise wide parallelism.