Specifications
Vendor
Asterfusion Data Technologies
Model
CX864E-N
Port Configuration
64x OSFP
Port Speed
800GbE per port
Switching Capacity
51.2 Tbps
Network OS
Enterprise SONiC ready
Fabric Role
Spine or leaf in AI/ML cluster fabrics
Architecture
Ultra Ethernet — RoCE-oriented AI fabric
Target Workload
AI/ML training and inference clusters
Use Case
Spine-leaf connectivity for 64-node fabrics
Oversubscription
Designed for non-blocking GPU fabric tiers
Form Factor
Fixed 1RU-class data centre switch
Management
SONiC NOS — open, scriptable, YANG/NETCONF capable
Optics
800G OSFP pluggable, breakout to 400G
Related Portfolio
CX732Q-N 32x400G, CX206Y-48GT-M PoE
Integration
Pairs with 800G OSFP transceivers and DAC/AEC cabling
Overview
The Asterfusion CX864E-N is a 64-port 800-gigabit Ethernet data-centre switch with 51.2 Tbps of switching capacity, built for AI and machine-learning cluster fabrics rather than general-purpose enterprise access. It is an open-networking platform: the unit ships SONiC-ready, so the network operating system is an open, scriptable stack rather than a vendor-proprietary NOS.
The 800GbE port density is the point. A 64-port 800G switch provides enough radix to build non-blocking spine and leaf tiers for GPU clusters without excessive cabling or power per delivered terabit, which is why 51.2T-class platforms have become the standard building block for AI fabric design.
Asterfusion positions the CX864E-N within the Ultra Ethernet ecosystem, the open specification backed by the Ultra Ethernet Consortium that targets the congestion control, determinism and observability gaps that generic Ethernet leaves open for AI workloads. In practice the switch is used with RoCE-style fabric designs where predictable tail latency across a large pod matters more than raw port count.
QS Compute supplies 800G switches, high-speed optics and AI fabric components for data-centre build-outs. Contact us for port configurations, SONiC licensing options and quote requests.
Key Benefits
51.2 Tbps in a single fixed platform gives sufficient radix for non-blocking AI fabric tiers. 64x OSFP 800GbE supports 400G breakout for mixed-speed fabrics. Enterprise SONiC ready — open NOS with no vendor lock-in and full automation hooks. Ultra Ethernet / RoCE oriented for predictable tail latency in large AI pods.
Applications
AI/ML training cluster fabrics, inference pod spine-leaf networks, GPU superpod interconnects, HPC cluster networking and hyperscale cloud data-centre fabrics.
Request a Quote — ASTERFUSION CX864E-N — 64-PORT 800GBE OPEN SONIC AI 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 →