NVIDIA 920-9B210-00FN-0D0 | Quantum-2 QM9790 64-Port 400G NDR InfiniBand Switch, 1U

Q: What is the difference between managed and unmanaged QM9790 models?
A: The unmanaged QM9790 (920-9B210-00FN-0D0) is externally managed via NVIDIA Unified Fabric Manager (UFM) or OpenSM, making it suitable for large-scale deployments beyond 2,000 nodes. Managed versions have internal subnet management suitable for smaller clusters up to 2,000 nodes.

Q: Can the QM9790 split ports to lower speeds?
A: Yes, each OSFP connector supports Twin-port 400G technology, allowing the 64 x 400Gb/s ports to be split into 128 x 200Gb/s ports. The switch also supports backward compatibility with HDR 200G equipment by downshifting port speeds.

Q: What network topologies does the QM9790 support?
A: The QM9790 supports Fat Tree, SlimFly, DragonFly+, and multi-dimensional Torus topologies. It's designed for maximum flexibility in AI training clusters, HPC fabrics, and hyperscale data center deployments requiring ultra-low latency InfiniBand connectivity.

Q: What cable types are compatible with the OSFP ports?
A: The 32 OSFP connectors support passive copper DAC, active copper cables, active optical cables (AOC), and optical modules. For interconnection with HDR equipment, OSFP-2xQSFP56 breakout cables are available.

Q: What is the difference between P2C and C2P airflow?
A: P2C (port-to-connector) airflow on the 920-9B210-00FN-0D0 moves air from the front ports to the rear power connectors (forward airflow). This is optimized for hot aisle/cold aisle data center designs where cold air enters from the front and exhausts to the rear hot aisle.