CORNELIS® CN6000 800G NETWORKING SOLUTIONS
The first 800G multi-protocol SuperNIC, delivering industry-leading AI and HPC performance with Omni-Path, RoCEv2, and Ultra Ethernet on PCIe 6.0.
Multi-Protocol Performance Without Compromise
Ready to Accelerate AI and HPC at Scale?
Built for AI Factories at Scale
As AI factories grow toward gigawatt-scale deployments and HPC environments expand to hundreds of thousands of nodes, networking infrastructure must deliver predictable performance, low latency, and seamless interoperability.
- Support Omni-Path, RoCEv2, and Ultra Ethernet on a single SuperNIC
- Scale efficiently from enterprise AI clusters to exascale HPC environments
- Deliver industry-leading message rates with sub-microsecond latency
- Maximize GPU utilization with optimized collective communications and RDMA acceleration
- Integrate seamlessly with standard CPUs, GPUs, storage, and Ethernet switch ecosystems
Industry-Leading Performance at Scale
Accelerate AI and HPC workloads with industry-leading message rates, sub-microsecond latency, and multi-protocol flexibility that maximize infrastructure utilization while reducing operational complexity.
-
Concurrent Omni-Path, RoCEv2, and Ultra Ethernet support on a single SuperNIC -
Increase CPU and GPU utilization with deterministic, low-latency communications -
Advanced congestion management and adaptive routing for predictable performance at scale -
Interoperate seamlessly with existing Ethernet switch ecosystems and CN5000 infrastructure -
Secure deployments with Post-Quantum Cryptography (PQC)-enabled Root of Trust -
Simplify deployment and management with industry-standard telemetry, monitoring, and firmware management tools
CN6000 Omni-Path Stands Apart
One SuperNIC. Three Protocols. Concurrent Omni-Path, RoCEv2, and Ultra Ethernet support on a single silicon platform
Highest Message Rate in the Industry: 1.6 billion messages per second with sub-1 µs latency
Built for Gigawatt-Scale AI: Advanced congestion management and adaptive routing keep GPUs fully utilized
Open Networking Without Lock-In: Interoperates with leading Ethernet ecosystems and existing CN5000 infrastructure
Future-Ready Security: Post-Quantum Cryptography (PQC)-enabled Root of Trust protects every deployment fully Verbs-compatible
Simplified Cluster Deployment
Deploy high-performance AI and HPC infrastructure faster with open standards, multi-protocol flexibility, and seamless interoperability across existing Ethernet and Omni-Path environments.
Concurrent Omni-Path, RoCEv2, and Ultra Ethernet support on a single SuperNIC
Seamless interoperability with standards-based Ethernet switch ecosystems
Backward compatible with existing CN5000 infrastructure
Industry-standard telemetry, monitoring, and firmware management interfaces
Open, standards-based architecture with support for leading AI and HPC frameworks
Secure deployment with PQC-enabled Root of Trust and SPDM 1.4
Universal Interoperability with:
Standard CPUs, GPUs, and storage platforms
NVIDIA and AMD accelerators via GPU Direct technologies
PyTorch, NCCL, RCCL, UCX, MPI, SHMEM, and OpenFabrics
Arista, Cisco, Juniper, NVIDIA Spectrum, and other Ethernet-based networks
Existing CN5000 Omni-Path deployments
Explore Next-Generation AI and HPC Networking
Get the latest insights on Omni-Path, RoCEv2, Ultra Ethernet, and the technologies powering the future of large-scale AI and HPC.