In a rare instance of collaboration, AMD and Intel have joined forces with Broadcom to create the Ultra Accelerator Link (UALink). This new open standard aims to challenge NVIDIA’s proprietary NVLink interface. The UALink is designed to enhance communication between AI accelerators, potentially revolutionizing the way data centers operate.
The UALink consortium, which includes AMD, Intel, and Broadcom, is working towards developing an industry standard that will enable AI accelerators to communicate more effectively. By creating an interconnect based on open standards, UALink aims to provide system OEMs, IT professionals, and system integrators with a pathway for easier integration, greater flexibility, and scalability of their AI-connected data centers.
Despite the promising prospects of UALink, it’s unlikely to yield any cross-vendor or heterogeneous accelerator interoperability. However, the consortium is hopeful that the new standard will help establish an open and high-performance environment for AI workloads, providing the highest performance possible.
The UALink consortium is expected to be incorporated in Q3 of 2024, and the 1.0 specification is anticipated to be available in the same quarter. This specification will define a high-speed, low-latency interconnect for scale-up communications between accelerators and switches in AI computing pods. It will enable the connection of up to 1,024 accelerators within an AI computing pod and allow for direct loads and stores between the memory attached to accelerators, such as GPUs, in the pod.
Read more: www.phoronix.com