NVIDIA Quantum-2 QM9790 Series NDR 400G InfiniBand Switch

1. MQM9790-NS2F Product Overview

The NVIDIA Quantum-2 QM9790 Series is a high-performance, ultra-low-latency InfiniBand switch designed for next-generation AI and HPC data centers. With 64 NDR 400Gb/s ports and a non-blocking architecture, it delivers an aggregate bidirectional bandwidth of 51.2 Tb/s and supports advanced features like SHARPv3, RDMA, and dynamic routing. Ideal for large-scale deployments, it offers unmatched scalability and efficiency.

2. MQM9790-NS2F Key Features

• 64 ports of NDR 400Gb/s InfiniBand connectivity
• 51.2 Tb/s aggregate bidirectional bandwidth
• Support for SHARPv3 in-network computing
• Port splitting for 128×200Gb/s (NDR200) operation
• Built-in or external management options
• Backward compatibility with previous InfiniBand generations
• Advanced QoS, congestion control, and self-healing capabilities

3. MQM9790-NS2F Core Technologies

The QM9790 leverages NVIDIA’s third-generation SHARPv3 technology, enabling scalable in-network aggregation for AI workloads. It supports RDMA, RoCE, and dynamic routing protocols. The switch also features NVIDIA MLNX-OS for comprehensive management and UFM integration for advanced orchestration and monitoring.

4.MQM9790-NS2F Working Principle

The NVIDIA Quantum-2 switch operates as a non-blocking crossbar, routing data with minimal latency. It uses adaptive routing and congestion control to optimize traffic flow. SHARPv3 offloads collective operations from CPUs to the network, significantly reducing data movement and accelerating distributed AI training and H simulations.

5. MQM9790-NS2F Applications

• AI and machine learning training clusters
• High-performance computing (HPC) environments
• Large-scale data analytics and cloud infrastructure
• Research institutions and supercomputing centers
• Financial modeling and scientific simulations

6. Specifications & Selection Guide

7. Advantages & Selling Points

• 32x higher AI acceleration with SHARPv3 vs. previous generations
• High port density in a 1U form factor
• Support for multiple network topologies (Fat Tree, DragonFly+, etc.)
• Low latency and high throughput for demanding workloads
• Comprehensive management via MLNX-OS and UFM platforms

8. Service & Support

We offer 1-year warranty, 24/7 technical support, and global logistics. Our team provides integration services, configuration assistance, and preventive maintenance planning. Bulk order discounts and customized solutions are available.

9. Frequently Asked Questions

Q: What is the difference between QM9790 and QM9790?
A: The QM9790 includes built-in management, while the QM9790 is designed for external management with NVIDIA UFM.

Q: Is the NVIDIA Quantum-2 switch backward compatible?
A: Yes, it supports previous InfiniBand generations and interoperates with existing infrastructure.

Q: What types of cables are supported?
A: It supports passive/active copper cables, active optical cables, and optical transceivers.

10. Precautions

• Operating temperature: 0°C to 40°C
• Humidity: 10% to 85% non-condensing
• Ensure proper airflow direction (P2C or C2P) based on model
• Use certified cables and transceivers for optimal performance
• Complies with CE, FCC, RoHS, and other international standards

11. Company Introduction

With over 10 years of industry experience, we operate a large factory and maintain a strong technical team. We are a trusted partner for leading brands including Mellanox, Ruckus, Aruba, and Extreme. Our product range includes network switches, network cards, wireless access points, controllers, and cables. We hold over $10M in inventory, ensuring reliable supply and competitive pricing. Our professional sales and technical support teams are available 24/7 to assist with product selection, deployment, and troubleshooting.