Cornelis details AMD-based AI network architecture

Key Takeaways

• Cornelis has introduced a new reference architecture combining its CN6000 SuperNIC with AMD's 6th Gen EPYC processors and AMD Instinct MI400 series GPUs.
• The architecture is designed for disaggregated AI inference, large-scale training, and HPC simulation, providing a blueprint for building high-performance systems.
• The CN6000 SuperNIC supports multiple transport protocols, including Omni-Path, RoCEv2, and Ultra Ethernet, and offers 800 Gbps of bandwidth through a native PCIe 6.0 x16 host interface.
• The architecture targets over 1.6 billion bidirectional messages per second, ensuring low latency and high throughput.

Introduction to Cornelis' New Reference Architecture

Cornelis, a leading provider of high-performance networking solutions, has unveiled a new reference architecture that pairs its CN6000 SuperNIC with AMD's 6th Gen EPYC processors and AMD Instinct MI400 series GPUs. This architecture is designed to provide infrastructure operators with a blueprint for building high-performance systems on an integrated compute and networking platform.

Engineering High-Performance AI Networks

The new reference architecture is built to address the growing demands of AI infrastructure, which requires a network that can keep pace with the compute resources. With the ability to support multiple transport protocols, including Omni-Path, RoCEv2, and Ultra Ethernet, the CN6000 SuperNIC provides operators with the flexibility to standardize their network strategy without locking into one protocol.

Comparison of Key Components

Performance Capabilities

The Cornelis CN6000 architecture is engineered to deliver high-performance capabilities, targeting over 1.6 billion bidirectional messages per second. This ensures low latency and high throughput, making it ideal for applications such as disaggregated AI inference, large-scale training, and HPC simulation.

Conclusion and Future Plans

Cornelis plans to detail the full architecture and share OEM validation results this fall at the OCP Global Summit and the AI Infra Summit. The company's continued investment in enabling customers to deploy the latest generation of AMD compute platforms reflects its commitment to providing high-performance networking solutions.

Bottom Line

The new reference architecture from Cornelis, combining its CN6000 SuperNIC with AMD's 6th Gen EPYC processors and AMD Instinct MI400 series GPUs, provides a powerful solution for building high-performance systems on an integrated compute and networking platform. With its high-performance capabilities and flexibility, this architecture is well-suited for a range of applications, from AI inference and training to HPC simulation. As the demand for high-performance AI networks continues to grow, Cornelis' new reference architecture is poised to play a key role in enabling customers to deploy the latest generation of AMD compute platforms.