NVIDIA Quantum MQM9790-NS2F 400G InfiniBand Switch
Key Advantages
- 64 ports of 400Gb/s InfiniBand with 25.6Tb/s aggregate throughput
- Enhanced NVIDIA SHARP v3 technology for advanced in-network computing
- Ultra-low latency under 100 nanoseconds for critical workloads
- Port splitting capability for 128*200Gb/s configurations
- Externally managed with NVIDIA UFM platform integration
1. Product Overview
The NVIDIA Quantum MQM9790-NS2F represents the pinnacle of high-performance InfiniBand switching technology, designed to meet the most demanding data center requirements. This next-generation network switch delivers sixty-four 400Gb/s ports with completely non-blocking architecture, providing unprecedented bandwidth for AI training, scientific research, and hyperscale computing applications. As an externally managed solution, this advanced network switch offers unparalleled control and visibility for large-scale fabric deployments.
2. Key Characteristics
- Maximum radix of 64*400Gb/s ports or 128*200Gb/s through port splitting
- 25.6Tb/s aggregate switching capacity with non-blocking performance
- Enhanced NVIDIA SHARP v3 technology with expanded collective operations
- Advanced congestion control and quality of service mechanisms
- Quad hot-swappable, 80 Plus Platinum certified power supplies
- NS2F airflow configuration for high-density data center environments
3. Core Technologies
The MQM9790-NS2F incorporates NVIDIA's third-generation Scalable Hierarchical Aggregation and Reduction Protocol (SHARP v3) technology, representing a significant evolution in network card and switch integration. This advanced in-network computing architecture embeds sophisticated computational capabilities directly within the switch ASIC, enabling dramatic acceleration of collective operations for distributed AI training and HPC simulations. The technology represents a substantial advancement over previous MQM9700 implementations, delivering enhanced programmability and expanded operation support.
4. Operational Principles
The Quantum MQM9790-NS2F operates using advanced cut-through switching with dynamic adaptive routing capabilities. When paired with modern network card solutions like NVIDIA ConnectX-7 adapters, the switch establishes ultra-low latency pathways optimized for distributed computing patterns. The enhanced SHARP v3 engine processes complex collective communication patterns directly within the switch fabric, eliminating the traditional overhead of host-based reduction operations. This network switch architecture enables unprecedented scalability while maintaining consistent performance across thousands of nodes.
5. Application Scenarios
This cutting-edge network switch is engineered for the most demanding deployment environments:
- Large-scale AI training clusters with distributed model parallelism
- Exascale computing facilities for scientific research and discovery
- Hyperscale cloud infrastructures requiring maximum fabric utilization
- High-frequency trading platforms demanding nanosecond latency
- Enterprise data centers running business-critical analytical workloads
6. Technical Specifications
| Parameter | Specification |
| Product Model | MQM9790-NS2F |
| Port Configuration | 64*QSFP-DD (400Gb/s each) |
| Aggregate Throughput | 25.6Tb/s non-blocking |
| Switch Latency | <100 nanoseconds |
| Airflow Configuration | NS2F (Network Side to Fabric Side) |
| Power Supply | 2+2 redundant, 200-240V AC |
| Management | Externally managed with NVIDIA UFM platform |
| Physical Dimensions | 1.7" H * 17" W * 30.5" D (1U) |
| Operating Temperature | 5°C to 35°C |
| Compatible Network Card | NVIDIA ConnectX-7, BlueField-3, and earlier generations |
7. Competitive Advantages
The MQM9790-NS2F establishes new benchmarks for high-performance networking, delivering substantial advantages over previous-generation switches including the MQM9700 series. The enhanced SHARP v3 technology supports more complex collective operations with higher precision, while the 400Gb/s port speed represents a 2x improvement over previous generations. The externally managed architecture enables centralized control of large-scale fabrics through NVIDIA's UFM platform, providing comprehensive visibility and proactive management capabilities. This network switch solution demonstrates exceptional power efficiency despite the significant performance increase, with advanced power management features that optimize consumption based on actual utilization.
8. Service & Support
- 3-year manufacturer warranty with optional extensions to 5 years
- 24/7 dedicated technical support with 4-hour response SLA
- Immediate availability from our 10-million component inventory
- Advanced replacement program for critical failure scenarios
- Custom integration services for heterogeneous network environments
- On-site deployment assistance and fabric design consultation
9. Frequently Asked Questions
Q: What management capabilities differentiate the MQM9790-NS2F?
A: As an externally managed switch, the MQM9790-NS2F requires the NVIDIA UFM platform for full functionality, providing centralized management, advanced analytics, and proactive fault prevention for entire fabrics rather than individual switches.
Q: How does this network switch compare to the previous MQM9700 generation?
A: The MQM9790-NS2F offers double the port speed (400Gb/s vs 200Gb/s), enhanced SHARP v3 technology, lower latency, and improved power efficiency compared to the MQM9700 series, representing a significant generational advancement.
Q: What type of network card is recommended for optimal performance?
A: We recommend NVIDIA ConnectX-7 or BlueField-3 network card solutions to fully leverage the 400Gb/s capabilities and advanced features of the MQM9790-NS2F switch.
Q: Can this switch operate in mixed-speed environments?
A: Yes, the MQM9790-NS2F supports automatic speed negotiation and can operate in mixed 400Gb/s, 200Gb/s, and 100Gb/s configurations, providing flexibility for gradual infrastructure upgrades.
10. Important Considerations
- NS2F airflow configuration requires specific data center cooling alignment
- Externally managed architecture necessitates NVIDIA UFM platform deployment
- Verify compatible network card and cable infrastructure for 400Gb/s operation
- Plan for adequate power capacity and circuit redundancy
- Consider fabric design implications for optimal topology utilization
- Review physical space requirements for deeper chassis dimensions
11. Company Introduction
With over a decade of industry expertise, our organization maintains extensive technical capabilities and manufacturing partnerships. We have cultivated a strong global reputation through consistent delivery of premium networking solutions. Our substantial inventory encompasses over 10 million components available for immediate deployment, ensuring timely fulfillment for projects of all scales and complexities.
We specialize in providing authentic, high-performance networking equipment, including advanced switches, network card solutions, wireless systems, and comprehensive connectivity components from industry leaders such as NVIDIA Networking, Aruba, Ruckus, and Extreme Networks. Our dedicated technical team delivers round-the-clock consultation, implementation support, and optimization services, ensuring peak performance of all deployed solutions including the advanced MQM9790 series network switches.
