CN5000 Omni-Path®

The World's First Lossless, Congestion Free Scale-Out Network

  • Request a Design Consult Arrow - Right
  • Maximize AI and HPC Performance

    New CN5000 Omni-Path networking solutions help you get faster model training completion time and lower inference latency vs comparable solutions in market today.

    A 3D render of a female elf's face with half shown as the polygonal mesh of lines and vertices.
  • Optimizations for AI Arrow - Right
  • A bird's eye view from space of a hurricane
  • Optimizations for HPC Arrow - Right
  • Built for Scale

    Training models are rocketing past a Trillion+ parameters as real-time inference data explodes and HPC infrastructure scales beyond traditional limits.

    • Scale seamlessly and linearly from single-rack pods to multi cluster supercomputers
    • Switch-integrated support for scale-up environments
    • Optimized communication paths for xPU-to-xPU traffic
    • Universal interoperability with standard CPUs, GPUs, storage
    Scales fluidly to >500k endpoints

    Industry Leading Efficiency at Scale

    Efficiency Gains Improve Your Application Performance, CAPEX, and OPEX

    • Icon of a central circle with four cubes arranged in a square around it, with all five items linked together with lines. Interoperable across infrastructure types
    • Icon of a gauge with the needle almost at maximum value. Increased CPU/GPU Utilization
    • Icon of a cloud with three vertical line-with-circle-endpoint-nodes below it. Congestion free and credit-based flow control
    • Icon of three servers stacked vertically with a down arrow in the lower right. Reduces rack space and expensive optical cabling costs
    • Icon of a clock in a circle with an arrow pointing down. Reduce admin time and training needs with streamlined management tools

    Simplified Cluster Deployment

    • Plug & Play integration with standards-based hardware

    • Compatible with all platforms and frameworks

    • Open-source software stack

    • In-box drivers for quick, easy deployment

    • Advance real-time telemetry and diagnostics

    Universal interoperability with:

    • nVidia CUDA logo
    • AMD ROCm logo
    • PyTorch logo
    • TensorFlow logo
    • Canonical Ubuntu logo
    • Red Hat logo
    • Open SUSE logo
    • Logo for the OpenFabrics Alliance
    • AMD Epyc logo
    • AMD Instinct logo
    • Custom ASICs logo
    • Intel Xeon logo
    • nVidia logo

    All trademarks are the property of their respective owners. Use does not imply affiliation or endorsement.

    CN5000 Omni-Path Stands Apart

    • Universal Interoperability: InfiniBand-level performance without vendor lock-in

    • Ultra-efficient congestion control vs. deep-buffer pause frames

    • Tuned for AI workloads true lossless, low-latency, predictable, scalable

    • OpenFabrics Interface optimized RoCE and Verbs compatible

    Explore the CN5000 Omni-Path Product Family

    Get smart on the latest and greatest in HPC and AI networking

  • Networking Courses Arrow - Right
  • Ready to try

    CN5000?

  • How to buy Arrow - Right