NVIDIA Mellanox MQM9700-NS2F NDR 400Gb/s InfiniBand Switch (64-Port QSFP56)
Product Details:
| Brand Name: | Mellanox |
| Model Number: | MQM9700-NS2F(920-9B210-00FN-0M0) |
| Document: | MQM9700 series.pdf |
Payment & Shipping Terms:
| Minimum Order Quantity: | 1pcs |
|---|---|
| Price: | Negotiate |
| Packaging Details: | outer box |
| Delivery Time: | Based on inventory |
| Payment Terms: | T/T |
| Supply Ability: | Supply by project/batch |
|
Detail Information |
|||
| Model Number: | MQM9700-NS2F(920-9B210-00FN-0M0) | Brand: | Mellanox |
|---|---|---|---|
| Uplink Connectivity: | 400 Gbps | Name: | InfiniBand MQM9700-NS2F Mellanox Network Switch 1U NDR 400Gb/S For Server |
| Keyword: | Mellanox Network Switch | Port Configuration: | 64x 400G, 32x OSFP |
Product Description
64-port Quantum-2 managed switch for next-generation HPC and AI infrastructures
The NVIDIA Mellanox MQM9700-NS2F delivers 64 ports of 400Gb/s full bidirectional bandwidth in a compact 1U form factor. With 51.2Tb/s aggregate throughput, ultra-low latency, and SHARP in-network computing, it is purpose-built for the most demanding high-performance computing and AI workloads.
The MQM9700-NS2F is a managed member of the NVIDIA Quantum-2 series, built around the next-generation NDR (Next Data Rate) ASIC. It provides 64 QSFP56 ports running at 400Gb/s per port. This switch is designed for top-of-rack leaf and spine connectivity, scaling from medium clusters to massive HPC and enterprise data center fabrics. Its integrated x86 dual-core CPU runs MLNX-OS, enabling comprehensive on-box management including integrated subnet manager for large-scale fabrics.
By offloading collective communication operations from the CPU to the switch network via SHARP technology, the MQM9700-NS2F dramatically reduces MPI operation time and frees valuable CPU resources for computation. This results in an order-of-magnitude improvement in application performance for parallel workloads.
The NVIDIA SHARP v3 technology embedded in the Quantum-2 ASIC enables the MQM9700-NS2F to perform collective operations such as MPI reductions directly within the network fabric. This eliminates multiple data traversals between endpoints, drastically reducing MPI operation time and freeing CPU cores for application computation. For fully populated fat-tree networks, partial reduction operations can complete in under three microseconds. This co-design approach accelerates communication frameworks and maximizes effective fabric bandwidth.
- Top-of-Rack Leaf Connectivity – Ideal for spine-leaf architectures in HPC and AI clusters.
- Medium to Extremely Large Clusters – Scales from a few hundred nodes to tens of thousands of nodes with integrated subnet management.
- Enterprise Data Centers – Supports high-performance storage, database, and compute workloads requiring low latency and high throughput.
- AI Training and Inference – SHARP v3 acceleration boosts NCCL and MPI-based distributed training performance.
- Cloud and Bare-Metal Infrastructures – Ideal for multi-tenant environments requiring high-bandwidth, low-latency network fabrics.
The MQM9700-NS2F is fully interoperable with NVIDIA Mellanox ConnectX adapters (NDR, HDR, EDR, FDR), LinkX cables, and transceivers. It is compliant with IBTA Specification 1.4 and 1.3, and supports backward compatibility to HDR (200Gb/s) and EDR (100Gb/s) technologies. The switch integrates with NVIDIA UFM (Unified Fabric Management) for advanced telemetry, AI-powered analytics, and proactive troubleshooting across the entire InfiniBand fabric.
| Parameter | Specification |
|---|---|
| Model | MQM9700-NS2F |
| Series | NVIDIA Quantum-2 |
| Ports | 64 x QSFP56 (NDR 400Gb/s) |
| Aggregate Throughput | 51.2Tb/s (non-blocking) |
| Latency | Under 90ns (port-to-port) |
| Switch ASIC | NVIDIA Quantum-2 NDR |
| CPU | Intel x86 dual-core |
| System Memory | 8GB RAM, 32GB SSD |
| Management | MLNX-OS (on-box), CLI, SNMP, UFM |
| Power Supplies | 1+1 redundant, hot-swappable, 100-240VAC, 80 Gold+ |
| Typical Power (ATIS) | 180W – 200W (fully populated) |
| Cooling | Front-to-rear (P2C) or rear-to-front (C2P), hot-swappable fans |
| Form Factor | 1U rackmount; 43.6mm (H) x 433.2mm (W) x 590.6mm (D) |
| Weight | 12.5 kg (approx.) |
| Operating Temperature | 0°C to 45°C |
| Certifications | CE, FCC, VCCI, ICES, RCM, RoHS, CB, cTUVus, CU |
| Ordering Part Number (OPN) | Description | Airflow | Management |
|---|---|---|---|
| MQM9700-NS2F | 400Gb/s InfiniBand 1U switch, 64 QSFP56, x86 dual-core, standard depth | P2C (front-to-rear) | Managed |
| MQM9700-NS2R | 400Gb/s InfiniBand 1U switch, 64 QSFP56, x86 dual-core, standard depth | C2P (rear-to-front) | Managed |
| MQM9790-NS2F | 400Gb/s InfiniBand 1U router, 64 QSFP56, x86 dual-core, standard depth | P2C (front-to-rear) | Managed |
| MQM9790-NS2R | 400Gb/s InfiniBand 1U router, 64 QSFP56, x86 dual-core, standard depth | C2P (rear-to-front) | Managed |
Note: Verify airflow direction (P2C or C2P) matches your data center hot/cold aisle configuration before ordering.
- Industry-leading performance – 51.2Tb/s throughput and ultra-low latency in a compact 1U chassis.
- Reduced IT costs – High port density (64 ports in 1U) reduces infrastructure footprint and operational expenses.
- Self-healing networking – InfiniBand fabric recovers from link failures 5,000× faster than software-based solutions.
- Flexible deployment – Backward compatible with HDR and EDR; supports modern topologies like Dragonfly+, SlimFly, and 6D torus.
- Enhanced telemetry – With UFM, gain real-time network insights and AI-driven analytics for proactive maintenance.
Hong Kong Starsurge Group Co., Limited provides global delivery and multilingual support for the MQM9700-NS2F. Our services include:
- Pre-sales consultation on topology design and compatibility.
- Firmware updates and technical documentation.
- Warranty support and RMA handling.
- Custom integration and system deployment assistance.
The MQM9700 is a switch, while the MQM9790 is a router with additional routing capabilities. Both share the same hardware platform but differ in software functionality.
QSFP56 connectors support passive copper DAC, active optical cables (AOC), and optical transceivers. For specific SKUs, refer to the NVIDIA LinkX cable family.
Yes, the Quantum-2 series is backward compatible with HDR (200Gb/s) and EDR (100Gb/s) InfiniBand technologies.
The onboard subnet manager can manage fabrics with up to 2,000 nodes.
Choose P2C (front-to-rear) if your rack has cold air intake at the front; choose C2P (rear-to-front) if your rack exhausts hot air through the front. Always match the airflow direction of adjacent equipment.
- Ensure input voltage range (100-127VAC or 200-240VAC) matches your facility power before connecting.
- Allow sufficient clearance for front-to-rear or rear-to-front airflow; do not block fan exhaust.
- Operating temperature must be maintained between 0°C and 45°C; relative humidity 10%–85% non-condensing.
- Use only approved QSFP56 cables and optical modules to ensure signal integrity and compliance.
Hong Kong Starsurge Group Co., Limited is a technology-driven provider of network hardware, IT services, and system integration solutions. Founded in 2008, the company serves customers worldwide with products including network switches, NICs, wireless access points, controllers, cables, and related networking equipment. Backed by an experienced sales and technical team, Starsurge supports industries such as government, healthcare, manufacturing, education, finance, and enterprise. The company also offers IoT solutions, network management systems, custom software development, multilingual support, and global delivery. With a customer-first approach, Starsurge focuses on reliable quality, responsive service, and tailored solutions that help clients build efficient, scalable, and dependable network infrastructure.
64 x NDR 400Gb/s
51.2Tb/s
Under 90ns
On-box (MLNX-OS)
180W – 200W
1U, 27" depth







