Brand: NVIDIA | Category: Network Switches
SKU: NVIDIAQUANTUMMQM8700HS2FINFINIBANDS | Part #: MQM8700-HS2F | MPN: MQM8700-HS2F
Contact for Pricing — Request a Quote
The NVIDIA Quantum MQM8700-HS2F delivers 16 Tb/s non-blocking aggregate bandwidth and line-rate forwarding across 40 high-speed ports, making it the fabric backbone for hyperscale data centers and AI infrastructure deployments. This single-socket InfiniBand switch employs the NVIDIA Quantum ASIC to enable HDR 200Gb/s per port connectivity with sub-140 nanosecond port-to-port cut-through latency, ensuring minimal delays in compute-intensive cluster communication. Network engineers and AI infrastructure teams standardize on this platform to eliminate fabric bottlenecks in GPU clusters, HPC systems, and distributed storage environments.
Part number MQM8700-HS2F is built on NVIDIA's proven Onyx NOS firmware and integrates SHARP v2 in-network computing offload for collective operations—all-reduce, broadcast, and barrier synchronization execute in hardware rather than consuming CPU cycles. The switch combines adaptive routing with credit-based flow control for dynamic congestion mitigation across fat-tree and dragonfly topologies. Its split-port capability delivers 80 × 100Gb/s HDR100 connections via split cables, offering deployment flexibility. Dual redundant 700W hot-swap power supplies and redundant hot-swap fans ensure carrier-grade availability, while the compact 1U form factor (44 × 430 × 520 mm) fits seamlessly into standard rack infrastructure. Out-of-band management via 10/100/1000 RJ-45 and USB console, paired with NVIDIA UFM (Unified Fabric Manager) orchestration, provides centralized visibility and provisioning across multi-switch fabrics.
Contact Omnixon Global for a detailed technical specification sheet and to submit your request for quotation on the NVIDIA Quantum MQM8700-HS2F InfiniBand switch.
| Brand | NVIDIA |
| Category | Network Switches |
| SKU | NVIDIAQUANTUMMQM8700HS2FINFINIBANDS |
| Part Number | MQM8700-HS2F |
| Condition | New |
| Form Factor | 1U |
| Switch ASIC | NVIDIA Quantum |
| InfiniBand Data Rate | HDR 200Gb/s per port |
| Total Ports | 40 x QSFP56 200Gb/s HDR ports |
| Split-Port Support | Yes — 80 x 100Gb/s (HDR100) via split cable |
| Aggregate Bandwidth | 16 Tb/s non-blocking |
| Latency | ~130 ns port-to-port cut-through |
| In-Network Computing | SHARP v2 (all-reduce, broadcast, barrier offload) |
| Congestion Control | Hardware-based adaptive routing and credit-based flow control |
| Routing | Adaptive routing, static, fat-tree, dragonfly topology support |
| Management Interface | 1 x 10/100/1000 RJ-45 out-of-band management; USB console |
| Management Software | NVIDIA UFM (Unified Fabric Manager) |
| Form Factor | 1U rack-mount |
| Dimensions | 44 x 430 x 520 mm (H x W x D) |
| Power Supply | Dual redundant hot-swap PSUs, 700W typical |
| Cooling | Hot-swap redundant fans, port-side exhaust or intake options |
| Operating Temperature | 0°C to 45°C |
| OS/Firmware | NVIDIA Onyx NOS |
| Standards Compliance | InfiniBand Architecture Specification v1.5, HDR specification |
| Telemetry | Per-port counters; UFM-integrated streaming telemetry |
The NVIDIA Quantum MQM8700-HS2F InfiniBand Switch is enterprise networking hardware for data-center, campus, and branch deployments. Common roles include core/distribution switching, server uplinks, top-of-rack fabric, and edge connectivity in mixed Cisco/Juniper/Arista environments.
Key specifications for the NVIDIA Quantum MQM8700-HS2F InfiniBand Switch: new condition; series Quantum; ports 40x HDR 200Gb/s QSFP56; switching capacity Up to 16Tb/s; form factor 1U; airflow P2C / front-to-rear airflow; management Managed data center fabric switch. Manufacturer part number MQM8700-HS2F. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
The NVIDIA Quantum MQM8700-HS2F InfiniBand Switch interoperates with standard switching protocols (LACP, MLAG/VPC, 802.1Q, OSPF/BGP). For specific cross-vendor scenarios — Cisco-to-Arista MLAG peering, fabric integration with Cumulus/SONiC, or migration from legacy Catalyst — our network engineers map a compatibility plan as part of the RFQ.