Search

Mellanox Spectrum-4 SN5610 64 OSFP Port 800GbE Ethernet Switch

SKU: 920-9N42F-00RI-3C1
NVIDIA Spectrum-4 SN5000 Series SN5610 - Switch - 64 x 800 Gigabit OSFP + 1 x 25 Gigabit SFP28 - front to back airflow - rack-mountable More details...
Price: $68,000.00
Save to my Wish List
 Mellanox Spectrum-4 SN5610 64 OSFP Port 800GbE Ethernet Switch
 
Product Details

NVIDIA Spectrum-4 SN5610 — 51.2 Tbps Data Center Switch

High-density, AI-ready Ethernet switch built to accelerate hyperscale and enterprise data center performance.


Unmatched Speed and Flexibility

The NVIDIA Spectrum-4 SN5610 delivers up to 51.2 Tbps of non-blocking bandwidth with 64 OSFP ports for 800 GbE connectivity — all in a compact 1U design. This switch enables the next generation of cloud, AI, and HPC data centers with breakthrough efficiency, ultra-low latency, and seamless scalability.

51.2 Tbps
Switching Capacity
64
OSFP Ports (800 GbE)
128 / 256
Breakout 400 GbE / 200 GbE
1U
Rack Height

Technical Specifications

Switch ASICNVIDIA Spectrum-4 (SN5000 Series)
Throughput51.2 Tbps bidirectional
Port Configuration64 × OSFP (10/25/50/100/200/400/800 GbE) + 2 × SFP28 (1/10/25 GbE) | Supports breakout up to 256 × 100 GbE
CPUAMD EPYC 3251 (8 cores)
Memory / Storage32 GB DDR4 RAM, 80 GB NVMe SSD
TelemetryAdvanced real-time visibility, fully shared packet buffer, and cut-through switching for deterministic performance
Power200–240 VAC | 900 W (typical) | 2.08 kW (max with optics)
Cooling & Fans5 hot-swappable fans | Front-to-back or back-to-front airflow options
Power Supplies4 redundant PSUs (AC/DC options)
Management Interfaces1 × RJ45 management, 1 × RJ45 console, 1 × USB-A port
Operating Temp0 °C to 40 °C | 10–85% RH (non-condensing)
Dimensions / Weight87 × 428 × 788 mm | 26.9 kg
ComplianceCB, cTUVus, CE, FCC, VCCI, ICES, RCM, BSMI, KCC, CCC | RoHS

Ideal Use Cases

  • AI / Machine Learning Clusters — High-bandwidth interconnects for GPU-based compute fabrics.
  • Cloud and Hyperscale Data Centers — Scalable, efficient 800 GbE switching for top-of-rack or spine deployments.
  • Enterprise Core Networks — Predictable latency and telemetry for mission-critical workloads.
  • High-Performance Computing (HPC) — Non-blocking architecture for compute-intensive applications.
  • AI Fabric Backbone — Optimized for RoCEv2 and end-to-end NVIDIA Spectrum networking stacks.
Previous Item   Next Item

Customer Account:



Talk to Us for all your questions / RFQs

Recently Viewed Products