Skip to main content
Asterfusion CX764QO-N 64x 400GE OSFP data center switch front panel.
Asterfusion CX764QO-N 64x 400GE OSFP data center switch.
Asterfusion CX764QO-N 64x 400GE OSFP data center switch front panel interface diagram.
Asterfusion CX764QO-N 64x 400GE OSFP 25.6t data center switch rear panel interface diagram.

CX764QO-N

64×400G Ethernet Switch, 25.6Tbps, Marvell Teralynx 10, For AI/ML/Cloud Data Center/HPC

Highlights
  • Advanced AI features: WCMP, INT-driven Routing,Packet Spray and more
  • Deliver port-to-port latency of around 560ns.
  • Max TDP of 2150W with 64x400G OSFP
  • Large on chip buffer of 200+MB for better RoCE performance.
  • 10ns PTP and SyncE performance supports.
  • Open AsterNOS based on SONiC with the best SAI support.
  • Compatible with heterogeneous GPUs & SmartNICs from various vendors.
  • Line-rate programmability to support evolving UEC standards.
  • ZR/ZR+ module supported
Please login to request a quote

64 x 400G OSFP, 25.6Tbps, Enterprise SONiC, AI/ML/Cloud Data Center/HPC/Storage

Designed to power the most demanding AI training & inference fabrics, cloud data centers & DCI, high-performance computing (HPC), and distributed storage workloads, the Asterfusion CX764QO-N is a game-changing, high-density 400G 25.6T data center switch. Powered by a single-chip Teralynx 10 and featuring 64 x 400G OSFP ports, it scales effortlessly to support ultra-large networks with line-rate, non-blocking performance. With an industry-leading ultra-low port-to-port latency of ~560ns, the switch guarantees deterministic, high-throughput connectivity critical for latency-sensitive applications. Furthermore, its integrated support for 400G Coherent ZR/ZR+ optics ensures high-capacity, cost-effective interconnectivity for modern long-reach cloud DCI architectures, eliminating the need for expensive external transponder systems

560ns

Ultra
Low Latency

Outperform IB

27.5% Higher TGR
20.4% Lower Latency

200+ MB

Packet
Buffer

KEY POINTS

Key Facts

560ns

Ultra Low Latency

Outperform IB

27.5% Higher TGR
20.4% Lower Latency

200+ MB

Packet Buffer

Optimized Software
for Advanced Workloads
– AsterNOS

Feature Lists ↓
Pre-installed with Enterprise SONiC, the CX764QO-N packs cutting-edge features like RoCEv2 and EVPN multihoming, delivering lightning-fast forwarding and massive packet buffers for unmatched performance. Fully UEC-compliant, it offers rich open APIs to seamlessly integrate into diverse data center and HPC ecosystems.
As a vendor-neutral platform, it effortlessly supports heterogeneous GPUs and network cards from multiple manufacturers. By combining industry-leading low latency with expansive packet buffering, the CX764QO-N sets the gold standard for next-generation AI-era data centers.

Optimized Software for Advanced Workloads – AsterNOS

Fully compliant with UEC standards, it offers extensive open APIs for smooth integration into diverse data center and HPC environments. As a vendor-neutral platform, it supports heterogeneous GPUs and network cards from multiple manufacturers. Combining cutting-edge low latency with large packet buffers, the CX764QO-N sets the benchmark as the go-to switch for next-generation AI-era data centers.
Feature Lists ↓

RoCEv2

Easy RoCE
PFC
ECN
DCBX

Cloud
Virtualization

DCI
VXLAN
BGP EVPN

In-Band Network
Telemetry

INT Based Routing
Flowlet
WCMP
Packet Spray

High
Availability

ECMP
MC-LAG
EVPN Multihoming

Cloud Virtualization

DCI
VXLAN
BGP EVPN

RoCEv2

 PFC/ECN/DCBX
Easy RoCE

High Availability

ECMP
MC-LAG
EVPN Mulithoming

Ops Automation

REST API/NETCONF/gNMI
In-band Telemetry

Specifications

Ports64x 400G OSFP, 2x 10G SFP+Switch ChipMARVELL TERALYNX 10
Switching Capacity25.6TbpsPort-to-port latency560ns
Packet forwarding rate 28800MppsPacket Buffer200+ MB
CPU Intel Xeon 4/Marvell OCTEON10 CN102
USB1 x USB2.0
RAM16GB/32GB DDR4, extensible to 96GBConsole1 x Console RJ45
SSD256GB M.2 SATAMGMT1 x MGMT GE RJ45
Hot-swappable Fans4 (3+1Redundancy)Hot-swappable Power Supplies
2 (1+1Redundancy)
Height2UDimensions
(WxHxD mm) 		440x87x650
Input voltage200-240V AC
200V~320V DC		Operating temperature0 to 40℃(32 to 104 °F)
Maximum power consumption2150WRelative humidity5% - 90% (non-condensing)
PTP (Optional)Class C Level

Quality Certifications

ISO-9001-SGS
ISO-14001-SGS
ISO-45001-SGS
ISO-IEC-27001-SGS
fcc-icon
ce-icon

Connectivity Solutions

400G Spine + Leaf Connectivity Solutions

Features

Marvell Teralynx 10 Inside: Unleash Industry-Leading Performance & Ultra-Low Latency

With Marvell Teralynx 10 at its core, Asterfusion 400G switches
achieve blazing-fast sub-560ns latency—perfect for AI/ML, HPC, and NVMe where speed drives results.
Marvell Teralynx 10 chip powering Asterfusion 800G switches with ultra-low latency
25.6Tbps

Switching Capacity

560ns

End-to-end Latency

200+ MB

Buffer Size

51.2 Tbps

Switching Capacity

560ns

Ultra Low Latency

200+ MB

Buffer
Size

Full RoCEv2 Support for Ultra-Low Latency and Lossless Networking

Asterfusion 400G switches deliver full RoCEv2 support with ultra-low latency and
near-zero CPU overhead. Combined with PFC and ECN for advanced congestion control, it enables a true lossless network.
Diagram comparing TCP/IP stack vs RDMA stack, highlighting Asterfusion 800G switches' full RoCEv2 support for ultra-low latency and lossless networking with near-zero CPU overhead
Optimized Features for AI/ML/HPC
Asterfusion Enterprise SONiC is optimized for AI/ML/HPC scenarios, supporting dynamic intelligent traffic distribution (Packet Spray);
Efficient and resilient path selection via WCMP (Weighted Cost Multi-Path);
Real-time precise telemetry routing based on INT Driven Routing;
as well as technologies like Flowlet and Auto Load Balancing to enhance traffic management.
Asterfusion Enterprise SONiC features for AI/ML/HPC networking illustration

Asterfusion Network Monitoring & Visualization

AsterNOS supports Node Exporter to send CPU, traffic, packet loss, latency, and RoCE congestion metrics to Prometheus.
Paired with Grafana, it enables real-time, visual insight into network performance.
AsterNOS network monitoring with Prometheus and Grafana dashboard

O&M in Minutes, Not Days

Automate with Easy RoCE, ZTP, Python and Ansible,  SPAN / ERSPAN Monitoring & more—cutting config time, errors, and costs.
Asterfusion network O&M automation tools diagram

Asterfusion 400G Switch Outperforms InfiniBand in AI Inference with Higher TGR and Lower Latency

In AI inference networks, the Asterfusion 400G switch delivers higher TGR (Token Generation Rate) and lower P90ITL (90th Percentile Inter-Token Latency) compared to InfiniBand, demonstrating faster inference speed and greater overall throughput performance.
Bar chart comparing Token Generation Rate (TGR) between Asterfusion 800G RoCE switch and InfiniBand switch, showing up to 27.5% higher AI inference performance
Bar chart comparing P90ITL (90th percentile inter-token latency) between Asterfusion 800G RoCE switch and InfiniBand switch, showing up to 20.4% lower AI inference latency
↑27.5%

Token Generation Rate

↓20.4%

Inference Latency

↑27.5%

Token Generation Rate

↓20.4%

Inference Latency

Related Products

CX764QD-NCX764QO-NCX764QH-N
Port Speeds64x400G QSFP-DD, 2x10G SFP+64x400G OSFP, 2x10G SFP+64x400G QSFP112, 1x25G SFP28
Hot-swappable PSUs and FANs1+1 PSUs, 3+1 FANs1+1 PSUs, 3+1 FANs1+1 PSUs, 3+1 FANs
MC-LAG√√√
BGP/MP-BGP√√√
EVPN√√√
BGP EVPN-Multihoming√√√
RoCEv2√√√
PFC √√√
ECN√√√
In-band Telemetry√√√
Packet Spray√√√
Flowlet√√√
WCMP√√√
INT-based Routing√√√
Auto Load Balancing√√×
SRv6××√

Panel

Asterfusion CX764QO-N 64x 400GE OSFP data center switch front panel interface diagram.
Asterfusion CX764QO-N 64x 400GE OSFP data center switch rear panel interface diagram.

What’s in box

800G switch with accessories

Warranty & Support

2-year hardware warranty

1-year free software upgrades & technical support

FAQs

Traditional ECMP often causes link imbalances under AI traffic. How does the CX764QO-N fix this?

The CX764QO-N addresses ECMP hash collisions by using hardware-driven Flowlet-Based Adaptive Routing and Packet Spraying. It dynamically splits massive elephant flows into granular flowlets and routes them based on real-time path congestion. This ensures full fabric utilization, eliminates bottlenecks, and slashes tail latency for RoCEv2 AI workloads.

Does it support Inband Telemetry / INT? Can it be used for traffic visualization and bottleneck analysis?

CX764QO-N supports INT. It can report metrics of BDC (Buffer Drop Collector) and HDC (High Delay Collector) to Promethus, bottleneck analysis can be realized via Grafana.

Compared to Broadcom Tomahawk 5, what are Asterfusion’s advantages?

→ Latency: 560ns vs ~800ns
→ Packet Buffer: 200MB vs 165MB

What 400G optical module standards are supported?

All OSFP-form 400G modules are supported, including ZR/ZR+ modules.

What switching silicon powers this platform, and what is its latency performance?

Yes. The entire CX764Q-N series, including the CX764QH-N, natively supports 400G Coherent ZR/ZR+ modules. This allows network operators to implement cost-effective, high-capacity Data Center Interconnect (DCI) directly over long distances without needing external transponder equipment.

How does the CX764QO-N integrate with modern cloud-native monitoring tools?

Built on a containerized architecture, AsterNOS includes a native Prometheus Exporter running right inside the OS. It streams real-time telemetry metrics directly to your Prometheus server, allowing you to instantly visualize microbursts and fabric health on standard Grafana dashboards without heavy SNMP polling overhead.

Related Blogs

32 ports 400G switches

How to Choose the Perfect 32 ports 400G Switch for AI Clusters, HPC, and NVMe-oF Storage

March 23, 2026
prometheus-exporter-layout

Prometheus Exporter for SONiC: Efficient Network Monitoring

September 18, 2025
cover for netconf vs. netconf

What is RESTCONF? The Difference between RESTCONF and NETCONF ?

July 30, 2026