NVIDIA Mellanox 920-9B110-00FH-0D0 Announced: A Precision-Crafted 200Gb/s HDR InfiniBand Switching Option

August 25, 2026

últimas noticias de la compañía sobre NVIDIA Mellanox 920-9B110-00FH-0D0 Announced: A Precision-Crafted 200Gb/s HDR InfiniBand Switching Option

NVIDIA Mellanox 920-9B110-00FH-0D0 Announced: A Precision-Crafted 200Gb/s HDR InfiniBand Switching Option

As HPC and AI clusters continue to evolve, not every deployment requires 400Gb/s NDR bandwidth. For the vast majority of existing and planned infrastructures, 200Gb/s HDR remains the sweet spot—offering exceptional performance at a more accessible price point. Recognizing this need, NVIDIA has introduced the NVIDIA Mellanox 920-9B110-00FH-0D0, a purpose-built InfiniBand switch that brings enterprise-grade reliability and high-density connectivity to HDR-based fabrics.

Background: Why This Switch Now?

Many organizations are either upgrading from EDR (100Gb/s) or building new clusters where 200Gb/s HDR provides ample bandwidth for most workloads—from small-scale AI training to medium-sized HPC simulations. The challenge has been finding a switch that delivers the density and feature set of higher-end platforms without the associated cost and complexity of NDR infrastructure. The 920-9B110-00FH-0D0 InfiniBand switch OPN (Ordering Part Number) fills this exact gap, offering a familiar, proven architecture that integrates seamlessly into existing HDR ecosystems.

This switch is particularly relevant for organizations that have standardized on HDR for their GPU compute nodes and storage backends. By choosing the 920-9B110-00FH-0D0 MQM8790-HS2F 200Gb/s HDR platform, IT managers can avoid a forklift upgrade to NDR while still gaining the benefits of higher port density and advanced management features.

Key Technical Features

At the heart of the NVIDIA Mellanox 920-9B110-00FH-0D0 lies the MQM8790-HS2F switching silicon, which delivers 40 ports of 200Gb/s HDR in a compact 1U form factor. Key specifications include:

  • Port configuration: 40 OSFP ports, each supporting 200Gb/s HDR with optional split support for 100Gb/s (HDR100) connectivity.
  • Switching capacity: 8.0 Tb/s aggregate bandwidth, sufficient for medium to large-scale fabrics.
  • Latency: Sub-200ns cut-through latency, ensuring minimal overhead for MPI and RDMA workloads.
  • In-network computing: Integrated SHARP (Scalable Hierarchical Aggregation and Reduction Protocol) v2.0 for offloading collective operations.
  • Advanced features: Adaptive routing, congestion control, and comprehensive telemetry via NVIDIA's UFM.

For engineers evaluating the 920-9B110-00FH-0D0 datasheet, the switch supports the same NVIDIA firmware and management tools as the broader Quantum-2 family, ensuring consistency across mixed HDR and NDR fabrics. This makes the 920-9B110-00FH-0D0 compatible with existing NVIDIA cables, transceivers, and fabric management software—significantly reducing integration complexity.

Addressing Real-World Deployment Needs

For network architects designing leaf-spine fabrics for GPU clusters with up to 1,000 nodes, the 920-9B110-00FH-0D0 InfiniBand switch OPN solution offers an ideal balance of performance and cost. Its 40-port density enables efficient top-of-rack (ToR) deployment, with each switch supporting up to 40 dual-port servers in a single rack. When combined with spine-tier MQM8790-HS2F switches, the fabric scales to several thousand endpoints without oversubscription.

From an operational perspective, the switch inherits the robust management capabilities of NVIDIA's InfiniBand portfolio. Fabric administrators can leverage UFM for zero-touch provisioning, health monitoring, and automated failover. The 920-9B110-00FH-0D0 specifications also highlight dual-redundant power supplies and hot-swappable fans, ensuring 99.999% availability for mission-critical workloads.

When considering 920-9B110-00FH-0D0 price and total cost of ownership, the switch offers compelling economics for organizations that do not require NDR's 400Gb/s bandwidth. The per-port cost is significantly lower than NDR alternatives, making it an attractive option for academic research labs, enterprise AI teams, and cloud service providers building cost-optimized HDR fabrics.

Summary: The Right Tool for the Right Scale

The NVIDIA Mellanox 920-9B110-00FH-0D0 is not a compromise—it's a deliberate choice for environments where 200Gb/s HDR delivers the necessary performance without overprovisioning. By offering high port density, advanced in-network computing, and seamless compatibility with existing HDR infrastructure, this switch enables architects to build scalable, efficient fabrics at a fraction of the cost of NDR upgrades.

For organizations planning their next-generation HPC or AI network, the 920-9B110-00FH-0D0 provides a clear, cost-effective path forward. Whether you're expanding an existing HDR fabric or building a new cluster from scratch, this switch ensures that your network keeps pace with compute demands without breaking the budget. The 920-9B110-00FH-0D0 for sale is now available through NVIDIA's channel partners, with volume shipments expected this quarter.