Top
Author
Published
10 Jul 2026Form Number
LP2461PDF size
12 pages, 974 KBSubscribed to LP2461.
Thank you for your feedback.
Abstract
Omni-Path Architecture (OPA) is a high-performance fabric designed for artificial intelligence (AI), high-performance computing (HPC), and large-scale data center environments. Built to deliver low-latency, high-bandwidth communication at scale, OPA enables efficient scaling from small clusters to fabrics supporting more than 100,000 nodes while maximizing compute utilization and reducing total cost of ownership. The Cornelis CN5000 Omni-Path Switch is Lenovo's current OPA switch offering, delivering 48 ports of 400Gbps connectivity in air-cooled and liquid-cooled configurations.
This document introduces the Lenovo Omni-Path switch portfolio and provides the associated ordering information. It is intended for technical specialists, sales specialists, sales engineers, IT architects, and other IT professionals who want to learn more about Lenovo-supported Omni-Path networking solutions and their use in IT environments.
Introduction
Cornelis Omni-Path switches are high-performance interconnect solutions designed to meet the demand of modern HPC and AI workloads. They provide low-latency, high-bandwidth communication across compute nodes, enabling efficient scaling and workload distribution. With features like adaptive routing, congestion control, and advanced monitoring, Omni-Path technology supports reliable and cost-effective infrastructure for data-intensive environments.
The Cornelis CN5000 Switch is a 400 Gbps Omni‑Path scale‑out fabric designed for modern AI and HPC data centers, delivering lossless, congestion‑free throughput, ultra‑low latency, and exceptional scalability to more than 100K nodes while maximizing compute efficiency and lowering TCO.
Through features such as credit‑based flow control, link‑level retransmissions, fine‑grained adaptive routing, and advanced real‑time telemetry, the CN5000 ensures low tail latency, eliminates packet loss, and provides precise network visibility, making it an ideal interconnect for performance‑critical, parallelized workloads
Did you know?
Omni-Path Architecture eliminates packet loss and minimizes network congestion through built-in credit-based flow control and adaptive routing. Unlike Ethernet-based AI fabrics that often require extensive tuning to achieve lossless operation, CN5000 Omni-Path fabrics are designed to deliver predictable low-latency performance at scale, helping AI and HPC clusters achieve higher utilization and faster time-to-results.
Ordering information
The following table lists the Cornelis Omni-Path switches orderable from Lenovo.
* PSE = Port Side Exhaust; oPSE = Opposite Port Side Exhaust
Each switch comes standard with two 2m C14 power cables. Other cable lengths are also available as listed in the Power cables section.
Lenovo EveryScale
Lenovo EveryScale is a framework for designing, manufacturing, integrating, and delivering data center solutions, with a focus on High Performance Computing (HPC), Technical Computing, and Artificial Intelligence (AI) environments.
With EveryScale, Lenovo pre-tests and validates a wide selection of Lenovo and third-party components, including Cornelis Omni-Path switches, to create a "Best Recipe" of components and firmware levels that work seamlessly together as a solution. From this testing, customers can be confident that these best-practice solutions run optimally for their workloads and are tailored to their specific needs.
Lenovo EveryScale hardware is pre-integrated, pre-cabled, and pre-loaded, optionally with an OS image, and is tested at the rack level during manufacturing. This process ensures reliable delivery and minimizes installation time in the customer data center.
With EveryScale, customers can focus their efforts on maximizing business value instead of consuming valuable resources to design, optimize, install, and support the infrastructure required to meet business demands.
Learn more by reading the Lenovo EveryScale product guide.
Front and rear views
The following figure shows the front of the CN5000 Switch:
Figure 2. Cornelis CN5000 Switch front view
The following figure shows the rear of the CN5000 Switch with air cooling:
Figure 3. Cornelis CN5000 Switch rear view air-cooled
The following figure shows the rear of the CN5000 Switch with liquid cooling:
Features
The CN5000 Omni-Path Switch has the following key features:
- 48 ports of 400Gb connectivity
- 38.4Tb full duplex bandwidth
- Throughput greater than 800 million packets/second
- Less than 1µs MPI latency
- Supports more than 100K node clusters
- Optimization for message rate and latency
- Virtual lanes (VLs): Configurable from one to four VLs plus one management VL
- Low-Latency bit error recovery and optional correction
- Security: Secure key EEPROM, Secure Boot
- Hot swappable/replaceable fans & power supplies
Built for Scale and Efficiency
The CN5000 Switch delivers lossless, congestion-free networking at scale, supporting up to 500,000 endpoints through key architectural features: credit-based flow control, link-level retransmissions, fabric-level congestion management, and multipath routing without in-order delivery constraints. These innovations ensure low tail latency, eliminate packet loss, and drive higher compute utilization and lower TCO.
Optimized for AI and HPC Workloads
The CN5000 Switch delivers 400Gb of bandwidth per port and is purpose-built for AI training, inference, and large-scale HPC performance demands. It provides ultra-low latency and congestion-free throughput, accelerating time-to-results for generative and agentic AI workloads. With exceptional efficiency and architectural robustness, CN5000 switches are the ideal fabric for parallelized compute environments and mission-critical HPC applications.
Reliability, Adaptability, and Efficiency
Consistent, reliable network performance is critical for AI and HPC applications. The CN5000 Switch enhances system-wide efficiency with advanced error correction and dynamic workload-aware traffic management, ensuring adaptability to changing communication patterns and sustained performance under load.
Advanced Telemetry for Intelligent Network Management
Equipped with cutting-edge telemetry capabilities, the CN5000 Switch provides real-time network insights and fine-grained visibility to optimize workload performance. By continuously analyzing traffic patterns, identifying congestion points, and dynamically adjusting network flows, it ensures smooth, high-efficiency operation, even in the most complex compute environments.
With industry-leading precision network intelligence, the CN5000 Switch enables greater control, efficiency, and performance across any scale. In contrast to RoCEv2 and legacy InfiniBand implementations, the CN5000 switch delivers lossless, congestion-free networking that scales with the demands of AI and HPC clusters.
Omni-Path Architecture
CN5000 Architecture delivers lossless and congestion-free data transmission with credit-based flow control and dynamic fine-grained adaptive routing. It is designed for maximum performance, reliability, scalability, and data integrity with dynamic lane scaling and optimized link protection. The CN5000 Switch requires the use of CN5000 SuperNIC adapters to fully enable its congestion-free, lossless operation at scale.
Air-cooled or Liquid-cooled Configurations
Available switch configurations:
- Air-cooled model with Port-side exhaust (PSE) fans - air flow travels from the power side to the port side of the switch
- Air-cooled model with Opposite-port-side exhaust (oPSE) fans - air flow travels from the port side to the power side of the switch
- Water-cooled model with 94% of heat removed using liquid cooling (PSE orientation for power supply fans)
Advanced Congestion Management
- Fine-Grained Adaptive Routing (FGAR)
- Static Dispersive Routing (SDR)
- Lossless, congestion-free networking through fabric-wide adaptive routing and incast-aware flow control
Management Features
- Integrated OpenBMC-based management
- Redfish protocol and data model/schema support
- In-band and out-of-band management options
- Command line interface through 10/100/1000 BASE-T Ethernet
Specifications
The following table shows the specifications of each switch. There are two variants for each switch:
- PSE (Port Side Exhaust): Airflow is Back-to-front - Air enters from power side and exits at port side
- oPSE (Opposite / reversed airflow): Airflow is Front-to-back - Air enters from port side and exits at power side
Networking cables
The following tables lists the supported cables.
In this section:
Cable length restrictions: The following cable length restrictions apply when configuring the CN5000 switch:
| Connection | DAC Cables | ACC Cables | AOC Cables |
|---|---|---|---|
| Switch-to-SuperNIC | up to 2.5m | up to 5m | up to 100m |
| Switch-to-Switch | up to 1m | up to 4m | up to 100m |
Use of the LLR (link-level retransmission) feature is subject to additional restrictions. Contact Cornelis for details.
Optical cables
Copper cables
RJ45 cables
Rack mount kit
The Cornelis CN5000 OPA switches include a rack mount kit (rail kit). The rails support an adjustable depth from 26 inches to 32 inches.
Power cables
The Cornelis switches come standard with two 2m power cables (feature CJTM, as listed in the table below). Additional power cables can be ordered if needed.
* Two 2m power cables are included with the switches
Operating environment
The switches have the following operating characteristics:
- Normal operating temperature range:
- 10 to 35°C (up to 900 m above sea level)
- Temperature is derated 1°C per 175 m above 900 m)
- Storage temperature range: -40 to 70 °C
- Normal operating humidity: 5% to 85%, non-condensing
- Non-operating humidity: 5% to 95%, non-condensing
- Altitude:
- Operating: 0 to 3,000 m
- Storage: 0 to 12,000 m
- Acoustics (noise) - Air-cooled switch: 72.50 dB @ 25 °C
Water requirements (liquid-cooled switch):
- Up to 94% heat capture to liquid depending on cable configuration
- ASIC, QSFP, and voltage regulators are liquid-cooled
- Quick disconnect couplings: no-drip/no-spill Staubli SCG06 with color indicators for flow
- Leak detection is indirect: Relies on the rack-level Coolant Distribution Unit (CDU) to monitor flow parameters (flow rate, pressure, temperature, etc.) to indicate a potential risk of a leak and to alert the user
- Liquid types supported:
- Deionized (DI) Water with additives
- Propylene Glycol mixture, 25% maximum (PG25)
- Ethylene Glycol mixture, 25% maximum (EG25)
- Flow rate: 1.5 to 3.0 LPM
- Max Inlet Temperature: 45 °C (113 °F)
- Temperature rise (ΔT): 3.0 - 5.0°C
- Maximum system pressure guidance: 3 bar at the system connection
Related publications and links
For more information, see these resources:
- Lenovo EveryScale product guide
https://lenovopress.lenovo.com/lp0900-lenovo-everyscale-lesi - Cornelis CN5000 product web page and product brief
https://www.cornelis.com/product/cornelis-cn5000-omni-path-switch - Cornelis CN5000 Switch User Manual
http://docs.cornelis.com/en/cn5000-product-family-descriptions/fabric-hardware-components/cn5000-switch.html
Trademarks
Lenovo and the Lenovo logo are trademarks or registered trademarks of Lenovo in the United States, other countries, or both. A current list of Lenovo trademarks is available on the Web at https://www.lenovo.com/us/en/legal/copytrade/.
The following terms are trademarks of Lenovo in the United States, other countries, or both:
Lenovo®
ThinkAgile®
The following terms are trademarks of other companies:
Intel®, the Intel logo is a trademark of Intel Corporation or its subsidiaries.
Other company, product, or service names may be trademarks or service marks of others.
Configure and Buy
Please select a locale
Full Change History
Course Detail
Employees Only Content
The content in this document with a is only visible to employees who are logged in. Logon using your Lenovo ITcode and password via Lenovo single-signon (SSO).
The author of the document has determined that this content is classified as Lenovo Internal and should not be normally be made available to people who are not employees or contractors. This includes partners, customers, and competitors. The reasons may vary and you should reach out to the authors of the document for clarification, if needed. Be cautious about sharing this content with others as it may contain sensitive information.
Any visitor to the Lenovo Press web site who is not logged on will not be able to see this employee-only content. This content is excluded from search engine indexes and will not appear in any search results.
For all users, including logged-in employees, this employee-only content does not appear in the PDF version of this document.
This functionality is cookie based. The web site will normally remember your login state between browser sessions, however, if you clear cookies at the end of a session or work in an Incognito/Private browser window, then you will need to log in each time.
If you have any questions about this feature of the Lenovo Press web, please email David Watts at dwatts@lenovo.com.
