Mellanox (NVIDIA Mellanox) 920-9B110-00FH-0D0 Debuts: Next-Gen InfiniBand Switch Accelerates HPC and AI Data Center
April 14, 2026
As AI clusters and high-performance computing push beyond the limits of traditional networking, the need for ultra-low latency, lossless, and scalable interconnect has never been more critical. Targeting this exact challenge, Mellanox (NVIDIA Mellanox) 920-9B110-00FH-0D0 arrives as a purpose-built InfiniBand switch designed to eliminate bottlenecks in demanding data center environments. The 920-9B110-00FH-0D0 bridges the gap between compute density and network efficiency, setting a new baseline for HDR 200Gb/s fabrics.
Modern AI training and simulation workloads suffer from traditional Ethernet's packet drops and variable latency. Network engineers and architects require deterministic performance at scale. The 920-9B110-00FH-0D0 InfiniBand switch OPN addresses this by delivering cut-through forwarding and credit-based flow control, ensuring no packet loss even under full-mesh traffic. Moreover, as GPU clusters expand, the 920-9B110-00FH-0D0 MQM8790-HS2F 200Gb/s HDR model provides the port density and bandwidth required to keep hundreds of GPUs efficiently synchronized.
- High Bandwidth & Density: Based on the Quantum HDR platform, the 920-9B110-00FH-0D0 delivers up to 200Gb/s per port, ideal for fat-tree and torus topologies.
- Low Latency: Hardware-based congestion control and dynamic routing minimize latency for collective communication operations (all-reduce, all-to-all).
- Management & Compatibility: Referencing the 920-9B110-00FH-0D0 datasheet and 920-9B110-00FH-0D0 specifications, the switch integrates seamlessly with NVIDIA UFM (Unified Fabric Manager), enabling real-time telemetry and predictive maintenance.
- OPN Flexibility: As a 920-9B110-00FH-0D0 InfiniBand switch OPN solution, it offers flexible ordering and configuration options for large-scale deployments.
For architects evaluating vendor lock-in, the 920-9B110-00FH-0D0 compatible ecosystem includes all major HDR cables (copper/fiber) and transceivers, alongside standard mounting rails for 19" racks. When comparing total cost of ownership, 920-9B110-00FH-0D0 price remains competitive given its feature set, while availability (920-9B110-00FH-0D0 for sale) is already ramping through authorized NVIDIA partners.
| Parameter | Details |
|---|---|
| Product Model | 920-9B110-00FH-0D0 (MQM8790-HS2F) |
| Port Speed | 200Gb/s HDR (also backward compatible with HDR100, EDR) |
| Form Factor | 1U rackmount, up to 40 ports |
| Management | OpenSM, UFM, REST API, SNMP |
For network engineers seeking deterministic fabric performance, the NVIDIA Mellanox 920-9B110-00FH-0D0 eliminates the guesswork. Whether you are designing a new AI cluster or upgrading an existing HDR infrastructure, this switch provides the reliability, bandwidth, and compatibility demanded by modern workloads. To explore the complete 920-9B110-00FH-0D0 datasheet and validate 920-9B110-00FH-0D0 compatible cabling options, refer to official NVIDIA documentation.

