Specifications
Vendor
FS
Model
N9600-64OD
Switch Silicon
Broadcom BCM78900 — Tomahawk 5
Port Configuration
64x OSFP
Port Speed
Up to 800Gbps per port
Total Throughput
51.2 Tbps
Latency
Under 1 microsecond
Network OS
PicOS with perpetual license
Fabric Role
Spine for AI/ML, HPC and cloud fabrics
Layer Support
Full L3 — comprehensive data centre routing feature set
Congestion Control
RoCEv2 / lossless Ethernet oriented
Breakout
800G to 2x400G / 4x200G / 8x100G
Deployment
Automated Day 0 / Day 2 provisioning
SDN Integration
AmpCon-DC controller support
Form Factor
Fixed 1RU-class data centre switch
Target Workload
RoCEv2 AI clusters, HPC, high-throughput storage
Overview
The FS N9600-64OD is a 64-port 800-gigabit Ethernet switch built on the Broadcom BCM78900 — the Tomahawk 5 switch ASIC. It delivers 51.2 Tbps of aggregate throughput at sub-microsecond latency, which places it in the same class as the current generation of AI-fabric spine platforms.
Port density and latency are the two numbers that matter for an AI fabric. Sixty-four OSFP cages at 800G each provide enough radix for a large non-blocking spine tier, and the breakout options down to 4x200G and 8x100G let the same chassis serve mixed-speed leaf populations without additional hardware.
The switch runs FS PicOS with a perpetual license rather than a subscription, and supports automated Day 0 / Day 2 provisioning with integration to FS's AmpCon-DC controller. For RoCEv2 AI-cluster designs, the lossless Ethernet and congestion-control behaviour is the feature that determines whether the fabric achieves predictable tail latency under all-to-all training traffic.
QS Compute supplies 800G switching, optics and cabling for AI/HPC data-centre builds. Contact us for configuration and volume pricing.
Key Benefits
51.2 Tbps on Broadcom Tomahawk 5 — current-generation AI fabric silicon. Sub-microsecond latency across all ports. Perpetual licensed NOS avoids recurring subscription cost. Flexible breakout from 800G down to 100G lets one chassis serve mixed-speed fabrics.
Applications
RoCEv2 AI/ML cluster spines, HPC fabric aggregation, high-throughput storage networks, research computing fabrics and cloud data-centre spine-leaf designs.
Request a Quote — FS N9600-64OD — 64X 800G OSFP TOMAHAWK 5 OPEN SWITCH
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →