NVIDIA Spectrum-3 SN4700 Ethernet Switch
Brand: NVIDIA | Category: Network Switches
SKU: NVID-MSN4700VS2FO | Part #: MSN4700-VS2FO | MPN: MSN4700-VS2FO
Contact for Pricing — Request a Quote
About the NVIDIA Spectrum-3 SN4700 Ethernet Switch
The NVIDIA Spectrum-3 SN4700 is a 32-port 400GbE Ethernet switch based on the Spectrum-3 ASIC, providing 25.6Tb/s of aggregate non-blocking switching bandwidth in a 1U form factor. Launched as NVIDIA's first generation of 400GbE Ethernet switches, the SN4700 established the foundation for modern AI-capable Ethernet fabrics and remains widely deployed in enterprises that adopted 400GbE infrastructure before Spectrum-4 became available. It supports full breakout capability, allowing each 400GbE QSFP-DD port to be split into 2x200GbE or 4x100GbE connections for high-density server and storage environments.
The SN4700 supports RoCEv2 lossless RDMA networking with hardware-based PFC and ECN, making it capable of supporting distributed AI training workloads over Ethernet. It runs NVIDIA Cumulus Linux or Onyx NOS, providing a rich set of layer 2 and layer 3 protocols including EVPN, VXLAN, BGP, OSPF, and ECMP. Advanced telemetry features including streaming gRPC/gNMI telemetry and integration with NVIDIA NetQ for network health monitoring are supported, enabling proactive operations management in production environments.
The SN4700 is commonly deployed as a spine switch in enterprises running 100GbE or 200GbE server-side connectivity, providing abundant 400GbE uplink capacity and non-blocking throughput. It also serves as a direct ToR switch in high-density GPU server racks and as an aggregation switch in storage fabrics. For organizations with existing SN4700 deployments, it continues to deliver a strong value proposition as a proven, stable platform with a full enterprise feature set.
Ideal for
- Spine switch in enterprise data centers with 100GbE/200GbE leaf-spine fabrics
- Top-of-rack switch for 400GbE-capable GPU servers in AI training environments
- High-density storage aggregation switch for NVMe-oF and parallel file system traffic
- RoCEv2 RDMA fabric switch for distributed AI/ML training over Ethernet
- Data center core switch for medium-scale enterprise data centers
- Cloud-native networking with EVPN/VXLAN multi-tenant overlay support
Technical specifications
| Manufacturer | NVIDIA |
| Part Number | MSN4700-VS2FO |
| Switch ASIC | NVIDIA Spectrum-3 |
| Ethernet Speed | 400GbE per port (native) |
| Total Ports | 32 x QSFP-DD 400GbE ports |
| Breakout Support | 32 x 400GbE, 64 x 200GbE, 128 x 100GbE |
| Aggregate Bandwidth | 25.6 Tb/s non-blocking |
| Forwarding Rate | 19.2 Bpps |
| Latency | <1 µs port-to-port (cut-through mode) |
| Buffer Architecture | Shared and dedicated buffers with Spectrum-3 architecture |
| RDMA Support | RoCEv2 with PFC, ECN, hardware congestion control |
| Layer 2 Features | VLAN, MSTP, LACP/LAG, MLAG, VxLAN, EVPN |
| Layer 3 Features | BGP, OSPF, IS-IS, ECMP, BFD, PIM, IGMP |
| Management Interface | 1 x 10/100/1000 RJ-45 OOB; USB console |
| Operating System | NVIDIA Cumulus Linux / NVIDIA Onyx NOS |
| Telemetry | Streaming telemetry (gRPC/gNMI), NVIDIA NetQ integration |
| Form Factor | 1U rack-mount |
| Dimensions | 44 x 440 x 520 mm (H x W x D) |
| Power Supply | Dual redundant hot-swap PSUs, 900W typical |
| Cooling | Hot-swap redundant fans, port-side exhaust or intake |
| Operating Temperature | 0°C to 45°C |
| Standards | IEEE 802.3bs (400GbE), 802.1Q, 802.3ad, 802.1X, 802.3x |
Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.
Technical Specifications
| Brand | NVIDIA |
| Category | Network Switches |
| SKU | NVID-MSN4700VS2FO |
| Part Number | MSN4700-VS2FO |
| Condition | New |
| Switch ASIC | NVIDIA Spectrum-3 |
| Ethernet Speed | 400GbE per port (native) |
| Total Ports | 32 x QSFP-DD 400GbE ports |
| Breakout Support | 32 x 400GbE, 64 x 200GbE, 128 x 100GbE |
| Aggregate Bandwidth | 25.6 Tb/s non-blocking |
| Forwarding Rate | 19.2 Bpps |
| Latency | <1 µs port-to-port (cut-through mode) |
| Buffer Architecture | Shared and dedicated buffers with Spectrum-3 architecture |
| RDMA Support | RoCEv2 with PFC, ECN, hardware congestion control |
| Layer 2 Features | VLAN, MSTP, LACP/LAG, MLAG, VxLAN, EVPN |
| Layer 3 Features | BGP, OSPF, IS-IS, ECMP, BFD, PIM, IGMP |
| Management Interface | 1 x 10/100/1000 RJ-45 OOB; USB console |
| Operating System | NVIDIA Cumulus Linux / NVIDIA Onyx NOS |
| Telemetry | Streaming telemetry (gRPC/gNMI), NVIDIA NetQ integration |
| Form Factor | 1U rack-mount |
| Dimensions | 44 x 440 x 520 mm (H x W x D) |
| Power Supply | Dual redundant hot-swap PSUs, 900W typical |
| Cooling | Hot-swap redundant fans, port-side exhaust or intake |
| Operating Temperature | 0°C to 45°C |
| Standards | IEEE 802.3bs (400GbE), 802.1Q, 802.3ad, 802.1X, 802.3x |
Frequently Asked Questions about NVIDIA Spectrum-3 SN4700 Ethernet Switch
What does the NVIDIA Spectrum-3 SN4700 Ethernet Switch do?
The NVIDIA Spectrum-3 SN4700 Ethernet 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.
What are the headline specs of the NVIDIA Spectrum-3 SN4700 Ethernet Switch?
Key specifications for the NVIDIA Spectrum-3 SN4700 Ethernet Switch: new condition; manufacturer NVIDIA; part number MSN4700-VS2FO; switch asic NVIDIA Spectrum-3; ethernet speed 400GbE per port (native); total ports 32 x QSFP-DD 400GbE ports; breakout support 32 x 400GbE, 64 x 200GbE, 128 x 100GbE. Manufacturer part number MSN4700-VS2FO. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
Is the NVIDIA Spectrum-3 SN4700 Ethernet Switch compatible with my infrastructure?
The NVIDIA Spectrum-3 SN4700 Ethernet 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.