As AI and HPC workloads continue to evolve in complexity, the demand for high-performance, scalable networking solutions has never been more critical. The Cornelis CN5000 Omni-Path adapters are purpose-built to deliver industry-leading performance and scalability, accelerating time-to-results and maximizing investment in extreme-scale environments.
The CN5000 adapter offers unmatched scalability via dynamic adapter routing, traffic flow optimization and zero-latency packet integrity protection. These features enhance the network's ability to handle high-demand workloads by reducing latency, preventing data loss, and ensuring efficient data transmission. As a result, the utilization of the GPU is improved leading to TCO optimization.
The CN5000 adapter provides up to 400Gbps bidirectional bandwidth while achieving the highest message rates and lowest latency. For AI-driven training workloads, CN5000 provides the bandwidth at scale and for inference it accelerates test-time compute and time-to-results by ensuring ultra-low-latency and eliminating tail latency. Built for highly parallelized compute, CN5000 remains the best choice for all HPC applications for its robustness, efficiency and performance.
In AI and HPC, consistent and reliable network performance is crucial for maximizing computational efficiency. CN5000 features advanced error correction, and dynamic network function placement, ensuring maximum network efficiency. These innovations deliver superior reliability and adaptability for evolving AI and HPC environments. CN5000 consumes up to 50% less power than competing adapter products1, reducing operational costs while maintaining peak efficiency.
The CN5000 adapter is equipped with advanced telemetry engines, delivering real-time insights and granular network visibility to optimize AI and HPC workloads. By continuously monitoring traffic patterns, detecting congestion points, and dynamically adjusting network flows, CN5000 ensures seamless performance, efficiency, and reliability in even the most demanding compute environments. With industry leading real-time telemetry and precision network intelligence, CN5000 ensures unparalleled visibility, efficiency and control – delivering high-performance networking across complex compute environments of any scale.
The Cornelis CN5000 product family proudly includes the Omni-Path Adapter, Edge Switch, and Director Class Switch, Copper and Optical cables, and open-sourced Host & Management Software options to offer you greater breadth and flexibility of architectural choices for your networking environment
Performance advantages increase as your cluster size grows
Industry-leading price-performance
Advanced telemetry engines deliver real-time information for optimal network traffic control
MSI-X interrupt handling supported
Enhanced congestion control
Enhanced resource scaling
KDETH, VERBS, and MAD transport using 16B or 9B packets
Enhanced send and receive side options
Single Root I/O Virtualization (SR-IOV) supported
Industry standard management interfaces (PLDM, MCTP) supported
Up to 400 Gbps
PCI Express Base Spec 5.0 compliant
MPI message rate >1BPmps
Latency as low as <1us
Congestion avoided via enhanced congestion control and fine-grained adaptive routing
Optimized for scalable MPI and SHMEM message rate and latency
Single and Dual port ADAPTER options
OpenFabrics Alliance (OFA) Software supported
| Bandwidth | 400G |
| Fabric Ports | One or two 400Gbps ports |
| Card Adapter Form Factors | Low Profile PCIe (LP bracket installed, Std bracket in package) |
| Power Consumption (Typical): - Single Port - Dual Port | 15W (w/o optics) 19W (w/o optics) |
| Cooling Options | Air-cooled Liquid-cooled: optional, semi-custom |
| Item Name | Item Number | Item Description |
|---|---|---|
| CN5HFA2D5ALS | 980074 | CN5000 Host Fabric Interface Adapter, Dual Port, QSFP-DD, x16 PCIe Gen 5, Air-Cooled, Low Profile, Single |
| CN5HFA1Q5ALS | 980075 | CN5000 Host Fabric Interface Adapter, Single Port, QSFP, x16 PCIe Gen 5, Air-Cooled, Low Profile, Single |
US/Canada | TUV NRTL 62368-1, CSA 22.2.No. 62368-1 |
Europe | TUV EN62368-1 |
International | CB Scheme: IEC 62368-1 |
Temperature | Operating: 0° to 55° C (derated 1C/175m above 900m) Storage: -40° to 70° C |
Humidity | Operating: 10% to 85% non-condensing Storage: 5% to 95% non-condensing |
Altitude | Operating: 0-3,000m Non-operating: 0-12,000m |
US/Canada | FCC Part 15, Subpart B, Class A, |
Europe/International | EN55032 (CISPR32), |
Japan | VCCI, Class A |
New Zealand/ Australia | AS/NZ CISPR 32, Class A |
Korea | Emissions: KS C 9832, Class A |
Taiwan | BSMI (CNS 15936), Class A |
RoHS | RoHS II Directive 2011/65/EU2, RoHS Directive 2015/863 |
REACH | (EC) No 1907/2006 |
1: Comparison utilizing competing product found at https://docs.nvidia.com/networking/display/connectx7vpi/specifications performing at 24.9W and 25.9W to Cornelis CN5000 Adapter power consumption of 15W and 19W.
From training large language models and powering agentic AI frameworks, to real-time fraud detection in finance and computer vision in autonomous systems - CN5000 Omni-Path AI solutions deliver the high-performance, lossless networking foundation that next-generation AI demands.
Push the boundaries of scientific discovery and engineering innovation with CN5000 Omni-Path-based HPC solutions. From advancing climate modeling and seismic analysis, to accelerating simulation and design in aerospace and automotive, and enabling complex genomic research in life sciences, CN5000 Omni-Path solutions power breakthrough results—faster, smarter, and at scale.
