NVIDIA/Mellanox MSN4700-WS2FC Mellanox Spectrum-3 Based 400GbE 1U Open Ethernet Switch with Cumulus Linux, 32 QSFP-DD Ports, 2 Power Supplies (AC),x86 CPU, Standard Depth, P2C Airflow, Rail Kit

P/N: MSN4700-WS2FC SKU: 10405

US$36,999.00

In Stock: 6 pcs
Quantity:
  • 30-day return and refund service icon 30-day return and refund
  • 3-year warranty service icon 3-year warranty
  • available for delivery within 2 days service icon Available for delivery within 2 days
  • free shipping for qualifying orders service icon Free shipping for orders over

Products Highlight

  • 32 × 400GbE QSFP-DD Ports in 1U
  • 12.8Tb/s Switching with 8.4Bpps Forwarding
  • Flexible 1–400GbE Multi-Rate Connectivity
  • Compact Spine and Super-Spine Deployment
  • RoCE-Ready Layer 2/3 Ethernet Fabric
  • Single-Pass VXLAN and IPv6 Segment Routing

Specifications

Part Number MSN4700-WS2FC Switch ASIC NVIDIA Spectrum-3
Form Factor 1U Switching Capacity 12.8Tbps
Ports 32 x QSFP-DD Port Speed 400GE
Software NVIDIA Cumulus Linux Airflow Power-to-Connector (P2C)
Cooling Mechanism Air Cooled Input Voltage 100VAC, 15A / 110VAC, 15A / 120VAC, 12A / 200–240VAC, 10A, 50/60Hz
Dimensions (H × W × D) 44 × 428 × 568.5mm Warranty 1-Year / 3-Year / 5-Year Support Options Available

Applications

Certification

Features

Switch Series & Supported Models

Browse common switch series and models supported for each brand.

Compare Items

Resources

Find answers to common questions about our products.

Which 400GbE interconnect should I choose: DAC, AOC, DR4, or FR4?

The choice depends on reach and future expansion plans. DAC is typically the most practical option for very short in-rack links. AOC is suitable for short point-to-point optical connections. DR4 is a strong option for parallel single-mode fiber deployments, especially when a future 4×100GbE breakout is required. FR4 or LR4 is better suited to longer duplex-fiber links, but these WDM optics cannot be split into separate lower-speed connections.

Our leaf layer is still mostly 100GbE. Does it make sense to deploy this 400GbE switch now?

Yes. All 32 QSFP-DD ports on the SN4700 can operate in a 4×100GbE breakout design, providing up to 128×100GbE connections without disabling adjacent ports. This allows customers to build a 400GbE core or spine layer first, then migrate leaf, GPU, or storage uplinks to 200GbE or 400GbE as the environment evolves.

Why does a 4×100GbE breakout link fail to come up even when the optics and port speeds appear correct?

In addition to checking the media type, FEC, and far-end speed, verify the Cumulus Linux lane configuration. In certain Cumulus Linux releases, a four-way breakout between an SN4700 and a Spectrum-based peer can fail with the default lanes-per-port setting. Follow the release-specific NVIDIA guidance for breakout parameters and validate the end-to-end link before large-scale deployment.

Does an SN4700 used as an EVPN/VXLAN spine need local VXLAN tunnel interfaces?

Not always. When the SN4700 acts only as an underlay router, EVPN route-exchange device, or route reflector—and does not terminate tenant traffic as a VTEP—it typically does not require local VXLAN tunnel interfaces. VXLAN configuration is needed when the switch must terminate VNIs, map local VLANs to VNIs, or provide gateway functions.

Does dual-PSU redundancy guarantee zero traffic impact during a live PSU replacement?

It should not be assumed. Dual AC PSUs provide power redundancy, but NVIDIA release notes identify potential transient frame loss on the SN4700 during PSU insertion or removal in certain Cumulus Linux releases. Treat PSU replacement as a controlled maintenance activity: confirm that the remaining power feed has sufficient capacity, ensure workloads have resilient paths, and review the release notes for the installed software version.

Have a question? Our technical support team is here to help you find the right solution.
Contact Us

Questions & Answers Title

Ask a question

No reviews yet. Be the first to share your experience with this product.