.webp)
A coalition of leading technology companies, including AMD, Broadcom, Cisco, Google, Hewlett Packard Enterprise (HPE), Intel, Meta, and Microsoft, has announced the formation of the Ultra Accelerator Link (UALink) Promoter Group. This initiative aims to develop an open industry standard intended to advance high-speed, low-latency communication for data centre AI systems.
The Ultra Accelerator Link (UALink), as the initiative is known, seeks to create a standard interface that will enhance the efficiency of AI accelerators. By adopting open standards, UALink is expected to enable system Original Equipment Manufacturers (OEMs), IT professionals, and system integrators to integrate AI-connected data centres with greater ease, flexibility, and scalability.
The companies involved in this initiative bring considerable expertise in creating large-scale AI and high-performance computing (HPC) solutions based on open standards. This collective knowledge will be instrumental in developing a robust, scale-up network capable of meeting the growing demands for AI compute power.
Forrest Norrod, Executive Vice President and General Manager of AMD's Data Center Solutions Group, emphasised the importance of creating an open, high-performance, and scalable accelerator fabric to support future AI advancements. "Together, we bring extensive experience in creating large-scale AI and high-performance computing solutions that are based on open-standards, efficiency, and robust ecosystem support," he said.
The group's primary initiative involves defining a high-speed, low-latency interconnect for scale-up communications between accelerators and switches within AI computing pods. The first specification—version 1.0—will facilitate the connection of up to 1,024 accelerators within an AI computing pod. It will also enable direct loads and stores between memory attached to accelerators such as GPUs. This specification is expected to be available in the third quarter of 2024 and will be accessible to companies that join the UALink Consortium.
Jas Tremblay, Vice President and General Manager of Broadcom's Data Center Solutions Group, highlighted Broadcom's commitment to increasing the implementation of large-scale AI technology in data centres. "It is critical to support an open ecosystem collaboration to enable scale-up networks with a variety of high-speed and low-latency solutions," he noted.
Cisco’s Martin Lund, Executive Vice President of the Common Hardware Group, underscored the urgent need for ultra-high-performance interconnects in the face of growing AI workloads. "Together, we are committed to developing the UALink, which will be a scalable and open solution available to help overcome some of the challenges with building AI supercomputers," he said.
HPE's Senior Vice President and General Manager of HPC & AI Infrastructure Solutions, Trish Damkroger, expressed optimism about contributing to HPE’s expertise in high-performance networking and systems. "As a founding member of the UALink industry consortium, we look forward to collaborating on developing a new open standard for accelerator interconnects for the next generation of supercomputing," she stated.
Sachin Katti, Senior Vice President and General Manager of Intel’s Network and Edge Group, highlighted Intel’s commitment to AI connectivity innovation. "This initiative extends Intel’s commitment to AI connectivity innovation that includes leadership roles in the Ultra Ethernet Consortium and other standards bodies," he mentioned.
J Metz, Chair of the Ultra Ethernet Consortium, pointed out the importance of a holistic perspective in improving efficiencies and performance. "We believe that UALink's scale-up approach to solving pod cluster issues complements our scale-out protocol," he added.
The formation of the UALink Promoter Group signifies a significant milestone in the technology industry’s effort to create a standardised, high-performance environment for AI workloads. With the combined expertise of AMD, Broadcom, Cisco, Google, HPE, Intel, Meta, and Microsoft, the future of AI and HPC solutions appears promising and innovation-driven.
The UALink Consortium is expected to be incorporated in the third quarter of 2024.
