 
|
NVIDIA mellanox Quantum-2 MQM9790-NS2F InfiniBand Switch 64-Port 400G NDR Smart Switch
Datos del producto:
| Nombre de la marca: | Mellanox |
| Número de modelo: | MQM9790-NS2F (920-9B210-00FN-0D0) |
| Documento: | MQM9700 series.pdf |
Pago y Envío Términos:
| Cantidad de orden mínima: | 1 Uds. |
|---|---|
| Precio: | Negotiate |
| Detalles de empaquetado: | Caja exterior |
| Tiempo de entrega: | Basado en el inventario |
| Condiciones de pago: | T/T |
| Capacidad de la fuente: | Suministro por proyecto/lote |
|
Información detallada |
|||
| Modelo: | MQM9790-NS2F | Puertos: | 64 × 400 g |
|---|---|---|---|
| Velocidad: | NDR InfiniBand | Potencia (máxima): | 1084W (pasivo) |
| flujo de aire: | Delantero a trasero | Marca: | Mellanox |
| Resaltar: | Mellanox Quantum-2 InfiniBand Switch,400G NDR Smart Switch,64-port Mellanox network switch |
||
Descripción de producto
NVIDIA Quantum-2 MQM9790-NS2F: 64-Port 400Gb/s InfiniBand Switch
Deliver Unprecedented Data Throughput for AI, HPC, and Hyperscale Environments
The NVIDIA Quantum-2 MQM9790-NS2F is a high-performance, fixed-configuration switch that sets a new standard for network density and speed. In a compact 1U form factor, it provides 64 ports of 400Gb/s InfiniBand, enabling massive scalability and ultra-low latency for the most demanding data-driven applications.
Key Facts at a Glance
| Ports & Speed | 64 ports of 400Gb/s InfiniBand (32 OSFP connectors) |
| Aggregate Throughput | 51.2 Tb/s bidirectional |
| Packet Processing | Over 66.5 Billion Packets Per Second (BPPS) |
| Form Factor | 1U Rack Mount |
| Management | Externally Managed (Optimized for NVIDIA UFM) |
Product Overview
Built on the revolutionary NVIDIA Quantum-2 platform, the MQM9790-NS2F is engineered to accelerate research and product development by providing unparalleled data capacity. It integrates third-generation NVIDIA SHARP™ In-Network Computing technology, enabling a 32X boost in AI acceleration power compared to prior generations. This switch is the ideal foundation for building extreme-scale, future-proof data center fabrics that require the highest performance and efficiency.
Key Features
- Unmatched Port Density: 64 non-blocking 400Gb/s ports in a 1U chassis, maximizing rack space utilization.
- NDR 400Gb/s Technology: Leverages the latest NVIDIA high-speed interconnect for extremely low latency and high bandwidth.
- In-Network Computing with SHARPv3: Offloads and accelerates collective communications, dramatically boosting AI and HPC application performance.
- Advanced Network Design: Supports flexible topologies including Fat Tree, DragonFly+, and multi-dimensional Torus for optimized network architecture.
- Adaptive Routing & Congestion Control: Ensures the highest effective throughput by dynamically managing network traffic.
- Port Split Technology: Enables operation as 128 ports of 200Gb/s for cost-effective network design.
- Self-Healing Network Capabilities: Enhances fabric resilience and reliability.
Advanced Technology: NVIDIA SHARPv3
The MQM9790-NS2F incorporates SHARPv3 (Scalable Hierarchical Aggregation and Reduction Protocol), a cornerstone of NVIDIA In-Network Computing. This technology allows computation to be performed within the network fabric itself as data moves between nodes. For AI training and HPC simulations, this reduces the amount of data that must be processed by endpoints, leading to significantly faster job completion times and virtually unlimited scalability for data aggregation tasks.
Typical Deployments
- AI and Machine Learning Fabrics: Ideal for large-scale AI clusters requiring massive parallel data exchange.
- High-Performance Computing (HPC): Suitable for research institutions running complex simulations and modeling.
- Hyperscale Data Centers: Used as a high-density spine or leaf switch to build efficient, scalable networks.
- Enterprise Core/Leaf: For large enterprises consolidating high-bandwidth workloads.
- GPU Cluster Networking: Directly connects to NVIDIA GPUs for optimized performance.
Compatibility & Ecosystem
The MQM9790-NS2F is fully backward compatible with previous InfiniBand generations (HDR, EDR, etc.), ensuring seamless integration into existing infrastructures. It is supported by the full NVIDIA software ecosystem, including:
- NVIDIA UFM (Unified Fabric Manager): For advanced provisioning, monitoring, and troubleshooting.
- MLNX-OS: Provides robust management capabilities via CLI, WebUI, and JSON.
- NVIDIA GPUDirect & RDMA: Enables direct data transfer between GPUs and storage, bypassing the CPU.
Compatibility Matrix
| Component | Compatible Options |
|---|---|
| Cabling | Passive/Active Copper, Active Fiber (OSFP connectors) |
| Previous Gen | Fully backward compatible with HDR (200Gb/s), EDR (100Gb/s) |
| Management Software | NVIDIA UFM, MLNX-OS |
| Topologies | Fat Tree, SlimFly, DragonFly+, 2D/3D Torus |
Technical Specifications
| Category | Detail |
|---|---|
| Product Name | NVIDIA Quantum-2 MQM9790-NS2F |
| Switch Type | 400Gb/s InfiniBand Smart Switch |
| Ports | 64 ports of 400Gb/s (32 OSFP connectors) |
| Aggregate Throughput | 51.2 Tb/s |
| Packet Rate | Over 66.5 BPPS |
| Form Factor | 1U Rack Mount |
| Dimensions (HxWxD) | 1.7 x 17.0 x 26.0 in (43.6 x 438 x 660.4 mm) |
| Weight | 14.5 kg (approx. 32 lbs) |
| Power Supply | 1+1 Redundant & Hot-swappable, 200-240Vac, 80 Gold+ |
| Cooling | Front-to-Rear (P2C), Hot-swappable fans |
| Management | Externally Managed (via UFM, CLI, WebUI, JSON, SNMP) |
| CPU | x86 Coffee Lake i3 |
| System Memory | 8GB DDR4 SO-DIMM |
| Storage | 16GB M.2 SSD |
| Operating Temperature | 0°C to 40°C |
| Humidity (Operating) | 10% to 85% non-condensing |
| Altitude | Up to 3050m |
| Certifications | CE, FCC, VCCI, RoHS, CB, cTUVus, ENERGY STAR |
Selection Guide: MQM9700 vs. MQM9790
| Model | Airflow Direction | Management | Best For |
|---|---|---|---|
| MQM9790-NS2F | P2C (Ports to Connector) / Forward | Unmanaged (Ext. via UFM) | Large fabrics requiring advanced UFM telemetry |
| MQM9790-NS2R | C2P (Connector to Ports) / Reverse | Unmanaged (Ext. via UFM) | Data centers with reverse airflow requirements |
| MQM9700-NS2F | P2C / Forward | Managed (On-board Subnet Mgr) | Small to medium deployments (up to 2000 nodes) |
| MQM9700-NS2R | C2P / Reverse | Managed (On-board Subnet Mgr) | Small to medium with specific airflow |
Advantages of the MQM9790-NS2F
- Unmatched Density: 64 ports of 400G in 1U reduces data center footprint and power consumption per Gb/s.
- Future-Proof Performance: NDR technology and SHARPv3 ensure readiness for next-generation AI and HPC workloads.
- Operational Efficiency: When paired with NVIDIA UFM, it enables proactive fabric management, reducing operational expenses (opex).
- Design Flexibility: Port splitting to 128 ports of 200Gb/s offers cost-effective topology options.
Service & Support by Hong Kong Starsurge
As a technology-driven provider since 2008, Starsurge offers comprehensive support for the NVIDIA Quantum-2 series:
- Pre-sales technical consulting and topology design assistance.
- Genuine product sourcing with warranty assurance.
- Worldwide shipping and logistics support.
- After-sales technical support and RMA services.
- Custom solutions for enterprise, education, and government sectors.
Buyer Checklist
- Confirm airflow direction: P2C (Forward) or C2P (Reverse) for your data center.
- Verify power requirements (200-240Vac) and redundancy needs.
- Select compatible cables: OSFP passive/active copper or active fiber.
- Determine if an external management server with NVIDIA UFM is required.
- Check altitude and operating temperature range compliance.
Frequently Asked Questions
A: The primary difference is management. The MQM9700 series has an on-board subnet manager for up to 2000 nodes, making it suitable for smaller, standalone fabrics. The MQM9790 series (like the -NS2F) is designed to be managed externally, typically by NVIDIA UFM, for advanced monitoring and control in larger, more complex environments.
A: Yes. Using NVIDIA port-split technology, the switch can operate as 128 ports of 200Gb/s, providing flexibility and cost optimization for network designs.
A: Yes, the Quantum-2 platform is fully backward compatible with previous generations, including HDR and EDR, ensuring smooth integration into existing fabrics.
A: The standard warranty is 1 year. Extended warranty options may be available. Please confirm with your sales representative.
Precautions & Important Notes
- Ensure proper airflow clearance in the rack to maintain operating temperature between 0°C and 40°C.
- Use only specified power input range (200-240Vac) to avoid damage.
- Some advanced features require specific software licensing; verify with your NVIDIA representative or Starsurge.
- For parameters not specified in the datasheet (e.g., specific optical module reach), please confirm before ordering.
About Hong Kong Starsurge Group Co., Limited
Hong Kong Starsurge Group Co., Limited is a technology-driven provider of network hardware, IT services, and system integration solutions. Founded in 2008, the company serves customers worldwide with products including network switches, NICs, wireless access points, controllers, cables, and related networking equipment. Backed by an experienced sales and technical team, Starsurge supports industries such as government, healthcare, manufacturing, education, finance, and enterprise. The company also offers IoT solutions, network management systems, custom software development, multilingual support, and global delivery. With a customer-first approach, Starsurge focuses on reliable quality, responsive service, and tailored solutions that help clients build efficient, scalable, and dependable network infrastructure.
Related Products
- NVIDIA Quantum-2 MQM9700-NS2F (Managed Version)
- NVIDIA ConnectX-7 NDR 400Gb/s InfiniBand Adapters
- NVIDIA LinkX InfiniBand Cables (OSFP – OSFP / OSFP – QSFP)
- NVIDIA UFM Enterprise Software License