Engineered for next-generation AI and high-performance computing, the NVIDIA MQM9790-NS2R Quantum-2 NDR InfiniBand switch provides massive 51.2Tb/s throughput via 64 ports of 400Gb/s OSFP connectivity. It is the ultimate networking solution for modern data centers requiring ultra-low latency and extreme bandwidth.
Item No :
9500-8iOrder(MOQ) :
1Product Origin :
Varies by product model (Specific dispatch location will be confirmed upon inquiry).Lead Time :
Standard items: 5-7 working days; Rare/Scarce items: Subject to negotiation.Warranty :
Original Brand: 1-Year Warranty; OEM/Compatible Solutions: 3-Year Warranty.Max Distance :
Up to 66.5 billion packets per second (BPPS)Compatibility :
51.2 Tb/s non-blocking bandwidthThe NVIDIA MQM9790-NS2R is a groundbreaking Quantum-2-based NDR InfiniBand switch engineered for extreme-scale AI clusters, High-Performance Computing (HPC), and hyperscale cloud data center networks.
Core Value Proposition: Built on the high-efficiency 1U chassis design, this switch delivers an incredible 64 ports of 400Gb/s connectivity through 32 high-density OSFP cages. With a massive 51.2Tb/s non-blocking bandwidth and advanced hardware acceleration technologies like RDMA and NVIDIA SHARP, it provides the ultimate networking fabric required for dense GPU acceleration.
Purpose-built for massive GPU acceleration environments, delivering ultra-low latency and scalable fabric options necessary to synchronize deep learning model training effortlessly.
Delivers robust 400Gb/s InfiniBand fabric performance vital for modern data centers requiring flexible, non-blocking switching throughput under high-concurrency cloud workloads.
Optimized natively for large-scale distributed computing workloads, utilizing advanced routing topology frameworks to sustain complex analytical tasks without packet bottlenecks.
Featuring NVIDIA’s advanced Quantum-2 InfiniBand ASIC architecture, this high-performance network device allows hyperscale networks to scale up node counts cleanly while maintaining optimal structural integrity.
Utilizes built-in NVIDIA SHARP (SHARPv3) In-Network Computing technologies to accelerate distributed processing workloads, significantly reducing overhead for data-intensive AI operations.
Equipped with 32 space-saving physical OSFP cages supporting 64 discrete NDR ports. This high-density geometry ensures smooth airflow dynamics inside standard 1U server enterprise racks.
Engineered with advanced adaptive routing and native RDMA communication protocols to establish ultra-low latency profiles across complex Fat-Tree or Dragonfly+ topologies.
| Hardware Metric | Factory Specification Details |
|---|---|
| Product Model | NVIDIA MQM9790-NS2R (Quantum-2 Platform) |
| Switch Architecture | NVIDIA Quantum-2 InfiniBand Switch ASIC |
| Port Configuration | 64 × 400Gb/s NDR InfiniBand ports |
| Physical Connector Type | 32 × OSFP cages (dual-port per cage) |
| Total Switching Capacity | 51.2 Tb/s non-blocking bandwidth |
| Packet Processing Rate | Up to 66.5 billion packets per second (BPPS) |
| Supported Protocol | NDR InfiniBand (with backward compatibility to HDR/EDR) |
| Port Breakout Capability | 128 × 200Gb/s via 1:2 splitting |
| In-Network Computing | NVIDIA SHARP (SHARPv3) acceleration support |
| Target Applications | AI clusters, HPC systems, hyperscale data centers, GPU fabrics |
Our high-performance networking components undergo intensive system-level cluster telemetry testing to guarantee absolute zero-packet-drop routing paths before large-scale site deployment.
By deploying premium Quantum-2 NDR InfiniBand hardware switch lines, cloud infrastructure operators can optimize overall computational capital expenditures without compromising system reliability or fabric throughput. Keep enterprise data center switching fabrics running at peak scale with elite network engineering.
Ready to benchmark 400Gb/s NDR InfiniBand fabric availability thresholds? Contact our hardware engineering division for detailed validation logs, volume commercial discount options, or specific trial evaluation allocations.
leave a message