AI Network Designer Design a network →

NVIDIA QM9790 (Quantum-2, externally managed)

64 × 400G in 1U. That is 25.6 Tb/s of switching through 32 OSFP cages.

64 logical ports through 32 physical cages. Each cage carries a module terminating 2 links, so optic counts do not equal port counts. See optics compatibility.

Breakout

A port faster than the endpoint it serves can be broken out when the speeds divide cleanly. That raises the usable port count. Two-tier capacity goes as the square of the port count, so it moves the tier boundary a long way.

Endpoint speedBreakoutUsable ports
400G none 64
200G 2:1 128
100G 4:1 256

Roles and optics

Coding matters more than it looks. A switch may refuse a transceiver that is not coded for it, and that is one of six axes that have to agree before a link comes up. The others are form factor, reach, fibre type, fibre connector and ferrule polish. See optics compatibility.

Notes

Externally managed variant of the QM9700. Usual choice for spine and core tiers, where the subnet manager is not needed on every box. Verify against the current vendor datasheet before quoting.

Source: NVIDIA QM97x0 hardware user manual (networking-docs.nvidia.com/qm97x0hw/Specifications), verified 2026-08. Confirmed 2026-08-06

Related