NVIDIA MQM9700-NS2R
Mellanox Quantum-2 64-Port 400Gb/s NDR InfiniBand Switch– Reverse Airflow (C2P)
The NVIDIA Mellanox MQM9700-NS2R is a next-generation 64-port 400Gb/s NDR InfiniBand Switch built on the Quantum-2 architecture. Designed for large-scale AI, ML, and HPC workloads, this 2U switch provides ultra-low latency, high throughput, and unmatched scalability with reverse airflow (C2P) for optimized cooling in data center environments.
Features
Equipped with 64 OSFP ports, each delivering 400Gb/s, the MQM9700-NS2R provides an aggregate switching capacity of 51.2Tb/s—ideal for high-performance compute fabrics and dense server interconnects. The Quantum-2 ASIC enables accelerated compute-to-compute communication through in-network computing, congestion management, and adaptive routing.
Advanced Options
Engineered for data centers requiring connector-to-power airflow, this model ensures thermal compliance in rack-level deployments where airflow direction matters. The system also supports telemetry-driven management, NVIDIA SHARPv3, and full compatibility with NVIDIA HPC SDK and Magnum IO software stack.
Product Key Features
-
-
64 × 400Gb/s NDR OSFP InfiniBand ports
-
Unprecedented 51.2Tb/s total switching bandwidth
-
Connector-to-Power (C2P) reverse airflow design
-
Built on NVIDIA Quantum-2 InfiniBand architecture
-
Advanced telemetry and real-time congestion control
-
Ideal for AI clusters, HPC, and hyperscale data centers
-
Reliability & Performance
Quantum-2 delivers sub-microsecond latency and line-rate throughput across all ports, providing high reliability and fabric scalability. Combined with NVIDIA full-stack networking software, the MQM9700-NS2R offers intelligent congestion control, workload-aware scheduling, and granular visibility to support AI-driven workloads and performance-critical applications.
Key Features :
-
-
64 × OSFP ports supporting 400Gb/s (NDR) each
-
Total switching bandwidth of 51.2Tb/s
-
Quantum-2 ASIC for in-network computing
-
Connector-to-Power (C2P) reverse airflow configuration
-
Integrated system management and diagnostics
-
Compatible with NVIDIA SHARPv3 and Magnum IO
-
Optimized for AI, deep learning, and HPC fabrics
-
2U rackmount design with hot-swappable components
-
Supports GPU-to-GPU communication at scale
-
Redundant power and cooling components
-
Write a Review