MQM9790-NS2F NVIDIA Quantum-2 Based NDR InfiniBand Switch, 64-Ports NDR, 32 OSFP Ports, Unmanaged, P2C Airflow (Forward)
NVIDIA MQM9790-NS2F Quantum-2 64-Port NDR InfiniBand Switch – Unmanaged, Forward Airflow
High-Speed 64-Port Unmanaged InfiniBand Switch with 400Gb/s, 32 OSFP & P2C Cooling
The NVIDIA MQM9790-NS2F is a high-density, unmanaged 64-port NDR (400Gb/s) InfiniBand switch powered by the Quantum‑2 ASIC. Designed for plug-and-play simplicity, it supports 32 OSFP ports, delivering exceptional 51.2Tb/s throughput and ultra-low latency—ideal for data centers, HPC clusters, and AI fabrics. Its Power-to-Connector (P2C) airflow aligns with standard front-to-back rack cooling.
🔧 Product Specifications: MQM9790-NS2F NVIDIA
More Information: User Manual
Feature | Details |
---|---|
Model Number | MQM9790‑NS2F |
Switch Type | Unmanaged InfiniBand |
Ports | 32× OSFP (64× 400Gb/s NDR) |
Switching Capacity | 51.2 Tb/s |
Packet Rate | 66.5 Bpps (Billion Packets/sec) |
Latency | ~130 ns |
Airflow Direction | P2C (Power-to-Connector) – Front-to‑Back cooling |
Form Factor | 1U Rack-mountable |
Power Supplies | Dual redundant, hot-swappable AC |
Fans | 6+1 hot-swappable fan units |
Management Interfaces | None (unmanaged configuration) |
Environmental Range | 0–35 °C operating, up to 3050 m altitude |
Dimensions (W×H×D) | 17″ × 1.7″ × 26″ / 433 mm × 43.6 mm × 660 mm |
Weight | ~31.97 lbs (14.5 kg) |
Typical Power Consumption | 640 W (up to 1.61 kW active) |
Certifications | RoHS, ENERGY STAR, 80 Plus Gold, CE, FCC, VCCI, ICES, RCM |
❓ Frequently Asked Questions (FAQs)
Q1: What does ‘unmanaged’ mean for this switch?
A: The MQM9790‑NS2F is plug-and-play with no CLI or SNMP interface; ideal for fabrics managed externally (e.g., via NVIDIA UFM).
Q2: What cooling direction does it use, and why does it matter?
A: Uses P2C forward airflow, pushing air from rear to front—compatible with front-to-back rack cooling setups.
Q3: How much bandwidth does each port support?
A: Each OSFP port provides 400 Gb/s; 32 ports yield up to 51.2 Tb/s total throughput.
Q4: Can it operate in multi-node AI and HPC clusters?
A: Yes. Supports RDMA, adaptive routing, SHARP, and topologies like Fat‑Tree and DragonFly+.
Q5: How heavy is the switch and how much power does it draw?
A: Weighs about 31.97 lbs (14.5 kg) and draws ~640 W under typical loads, with peak up to 1.61 kW.