NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches: Special Offer
The NVIDIA Quantum-2 QM9701 offered here is a 1U enterprise InfiniBand switch from the NVIDIA QM97XX switch family, configured with 64× NDR 400Gb/s ports and 32× OSFP cages. It is a managed switch powered from a 48V DC busbar instead of AC power supplies. This is a fixed, tested, deployment-ready configuration available from stock.
NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches
Limited stock at special pricing
Configuration Overview: NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches
This NVIDIA Quantum-2 QM9701 configuration is built around a 64-port NDR 400Gb/s InfiniBand platform, designed for AI clusters, GPU server networking, HPC environments, and high-bandwidth data center fabrics.
It is designed for:
AI cluster networking
GPU-to-GPU and node-to-node communication
HPC and high-performance data center infrastructure
The switch offered is based on a fixed configuration selected for stability, compatibility, and real-world deployment. Our overview of NVIDIA QM9700, QM9790 and Quantum-X800 InfiniBand switches explains the wider NVIDIA switch family.
Full Configuration: NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches
Base system: NVIDIA Quantum-2 QM9701 1U switches
Component | Details | Why it matters |
Switch Platform | NVIDIA Quantum-2 QM9701 | Purpose-built high-performance InfiniBand switch platform |
Part Number | MQM9701-NS2R, as commonly listed by distributors | Confirm the exact part number and airflow direction before ordering |
Switch Family | Same Quantum-2 NDR switch family as QM9700 and QM9790 | |
Form Factor | 1U rackmount | High port density in a compact data center format |
Power Type | 48V DC busbar | Needs a rack with DC busbar power |
Architecture | NVIDIA Quantum-2 | Designed for low-latency, high-bandwidth InfiniBand fabrics |
The NVIDIA Quantum-2 QM9701 is suitable for buyers who need reliable high-speed networking for AI clusters, GPU servers, HPC systems, storage infrastructure, and large data center workloads.
Networking: 64× NDR 400Gb/s InfiniBand ports
Component | Details | Why it matters |
InfiniBand Ports | High-speed connectivity for GPU clusters and HPC systems | |
Physical Connectors | 32× twin-port OSFP cages | Dense connectivity with support for high-speed cabling options |
Port Splitting | Up to 128× 200Gb/s ports | Flexible deployment for different cluster designs |
Aggregate Bidirectional Throughput | 51.2Tb/s | Supports large-scale east-west data movement |
Packet Rate | More than 66.5 billion packets per second | Handles heavy AI and HPC traffic |
The 64-port NDR design allows the QM9701 to support dense GPU cluster connectivity while reducing the number of switches, cables, and rack space needed compared to lower-speed networking generations.
Each OSFP cage carries two 400Gb/s NDR ports, so an "800G" cable on the QM9701 carries 2×400G NDR, not one native 800G XDR link. Read more in our comparison of NVIDIA NDR 400G vs XDR 800G InfiniBand.
Management
Component | Details | Why it matters |
Switch Management | Managed switch configuration (internally managed) | Enables easier configuration and monitoring |
Operating Software | NVIDIA MLNX-OS | Standard software for managed Quantum-2 switches |
Management Access | CLI, WebUI, SNMP, and JSON interfaces | Useful for administration, automation, and monitoring |
Subnet Manager | Supports simpler fabric bring-up and management | |
Deployment Type | Enterprise data center switch | Suitable for AI, HPC, and GPU cluster environments |
The QM9701 is managed in the same way as the NVIDIA QM9700. The main difference is power: the QM9701 is DC powered using a busbar, while the QM9700 and the externally managed NVIDIA QM9790 use AC power supplies. Check the exact part number, airflow direction and rack power before deployment.
Cabling and connectivity
Component | Details | Why it matters |
Connector Type | Twin-port OSFP, finned-top | Modern high-speed connector format for NDR InfiniBand |
Supported Speeds | 400Gb/s NDR and 200Gb/s via port splitting | Flexible cluster design options |
Cable Options | DAC, linear active copper (LACC), and optical transceivers with fiber | Supports different rack layouts and distance requirements |
Topology Support | Fat Tree, SlimFly, DragonFly+, multi-dimensional Torus, and more | Useful for small, medium, and large-scale AI or HPC fabrics |
Correct cabling is important with NDR InfiniBand. Cable length, airflow direction, connector type, and switch port planning should be checked before deployment. NVIDIA does not offer AOCs for NDR twin-port OSFP links, except 1:2 splitters for HDR backward compatibility. See our guides to NVIDIA 800G OSFP to OSFP cables and NVIDIA GPU cluster cables.
Power & cooling
Component | Details | Why it matters |
Power Input | DC busbar, 40–59.5V DC (48V DC nominal); no replaceable power supply units | The rack must provide DC busbar power |
Power Consumption | 720W typical with passive cables, up to 1,660W with active cables | Important for rack power planning |
Cooling | 7 replaceable fan modules | Designed for stable operation in rack environments |
Airflow Direction | Confirm with the exact part number | Important for hot aisle / cold aisle planning |
Form Factor | 1U chassis, 854 mm deep (deeper than the 660 mm QM9700 and QM9790) | Check rack depth before installation |
High-speed InfiniBand switches require proper rack airflow and power planning, especially when deployed together with dense GPU servers and AI cluster nodes.
Platform Overview: NVIDIA Quantum-2 QM9701 Switches
The NVIDIA Quantum-2 QM9701 is designed for organizations that need:
High-speed InfiniBand networking
NDR 400Gb/s connectivity
AI cluster infrastructure
GPU server networking
HPC cluster connectivity
Storage and data center networking
Low-latency server communication
Practical network hardware for demanding workloads
This switch platform is a good fit for IT teams looking to expand or refresh their infrastructure with NVIDIA InfiniBand networking for AI, HPC, and GPU-based workloads, especially in racks that already distribute 48V DC power over a busbar.
Use Cases: NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches
AI cluster networking
Multi-node GPU training
HPC simulation workloads
Research and scientific computing
High-performance storage networking
Data center fabric refresh projects
Large-scale InfiniBand infrastructure
The NVIDIA Quantum-2 QM9701 configuration makes this switch useful for organizations that need high-speed and reliable InfiniBand networking for demanding infrastructure workloads, such as clusters of NVIDIA DGX B200 or other 8-GPU servers.
FAQ: NVIDIA Quantum-2 QM9701 Switches
What is the NVIDIA Quantum-2 QM9701 used for?
The NVIDIA Quantum-2 QM9701 is used for AI clusters, GPU server networking, HPC environments, storage networking, data center fabrics, and high-performance server communication. It is an NVIDIA InfiniBand switch designed for businesses that need reliable high-speed infrastructure hardware.
Is the NVIDIA Quantum-2 QM9701 suitable for AI clusters?
Yes. This NVIDIA Quantum-2 QM9701 configuration with 64× NDR 400Gb/s InfiniBand ports and 32× OSFP cages is suitable for AI clusters, GPU server networks, HPC workloads, and large data center infrastructure.
What ports are included in this NVIDIA Quantum-2 QM9701 switch?
This stock configuration includes 64× NDR 400Gb/s InfiniBand ports through 32× OSFP cages. It is suitable for high-speed AI, HPC, GPU cluster, and data center networking workloads.
What is the difference between QM9700, QM9790, and QM9701?
QM9700, QM9790, and QM9701 are part of the NVIDIA QM97XX Quantum-2 NDR InfiniBand switch family, with the same 64 NDR ports and 51.2Tb/s throughput. The QM9700 is managed and AC powered. The QM9790 is externally managed (for example with NVIDIA UFM) and AC powered. The QM9701 is managed like the QM9700 but DC powered using a busbar. Before purchase, check the part number for airflow direction and included accessories.
Does the NVIDIA QM9701 need a 48V DC busbar?
Yes. The QM9701 is DC powered using a busbar, with an input range of 40–59.5V DC, and it has no AC power supplies. It needs a rack that provides DC busbar power.
How much bandwidth does the NVIDIA Quantum-2 QM9701 support?
This NVIDIA Quantum-2 QM9701 configuration supports NDR 400Gb/s InfiniBand connectivity per port and up to 51.2Tb/s aggregate bidirectional throughput. This provides a strong networking foundation for AI clusters, GPU servers, HPC systems, and large infrastructure deployments.
Where can I buy NVIDIA Quantum-2 QM9701 switches in Europe?
You can request a quote from server-parts.eu for NVIDIA Quantum-2 QM9701 switches, NVIDIA InfiniBand switches, AI cluster networking hardware, GPU server networking equipment, and related data center parts.
Testing, Condition, and Warranty: NVIDIA Quantum-2 QM9701 Switches
Condition
Stock switch configuration
Hardware details based on available configuration information
Refurbished, tested, deployment-ready hardware
Testing includes
Switch power-on validation
Port and link testing
Management access verification
Fan module inspection
DC power input verification
Basic switch hardware validation
Warranty
Warranty information available on request
Additional support options may be available depending on stock and configuration
Final warranty terms confirmed during quotation
Request NVIDIA QM9701 Pricing and Availability
To get a quote, send us the switch model, part number, quantity, airflow direction, rack power setup and delivery country through our request a quote form. We can also quote matching NVIDIA 800G OSFP cables and transceivers.
Sources: NVIDIA QM9701 Documentation
NVIDIA Quantum-2 QM9701 NDR InfiniBand Switches
Limited stock at special pricing






Comments