Google, AMD, Meta, Microsoft, and other leading technology vendors have introduced a new industry standard aimed at enhancing AI connectivity in data centers.
The Ultra Accelerator Link (UALink) is designed to boost performance and deployment flexibility in AI computing clusters within data centers. This new standard specifically targets accelerators found on GPUs, facilitating more efficient interconnections for hardware powering AI training and inference workloads.
Version 1.0 of UALink will allow data center operators to connect up to 1,024 accelerators within a single computing pod, with formal adoption expected later this year. Other major tech companies, including AMD, Broadcom, Cisco, Intel, and HPE, have also joined the initiative to establish this open industry standard.
The UALink standard aims to enable data centers to scale computing resources within a single instance, thereby increasing capacity on demand without interrupting ongoing workloads.
.@Google, @AMD, @Meta and @Microsoft, along with other technology vendors, have launched a new industry standard for AI connectivity in data centers | @ultraethernet #AI #HPC https://t.co/ZcQ7SG8B1D
— AI Business (@Business_AI) June 7, 2024
Martin Lund, executive vice president of Cisco’s common hardware group, emphasized the growing importance of ultra-high performance interconnects as AI workloads expand. He stated, “Together, we are committed to developing UALink as a scalable and open solution to overcome the challenges of building AI supercomputers.”
Forrest Norrod, general manager for AMD’s data center solutions business group, highlighted the critical nature of creating an open, high-performance, and scalable accelerator fabric for the future of AI. He noted, “Together, we bring extensive experience in creating large-scale AI and high-performance computing solutions based on open standards, efficiency, and robust ecosystem support.”
The companies backing the UALink standard are part of the Ultra Ethernet Consortium (UEC), an industry group that supports collaboration around Ethernet-based networking. J Metz, UEC’s chair, remarked on the rapid industry response to AI and high-performance computing challenges, stating, “Interconnecting accelerators like GPUs requires a holistic perspective to improve efficiencies and performance. At UEC, we believe UALink’s scale-up approach complements our own scale-out protocol, and we look forward to collaborating on creating an open, ecosystem-friendly, industry-wide solution.”
Notably, Nvidia has not joined the UALink initiative, opting instead to continue using its proprietary NVLink technology for GPU interconnections.
Start Growing with Cloudways Today.
Our Clients Love us because we never compromise on these
Abdul Rehman
Abdul is a tech-savvy, coffee-fueled, and creatively driven marketer who loves keeping up with the latest software updates and tech gadgets. He's also a skilled technical writer who can explain complex concepts simply for a broad audience. Abdul enjoys sharing his knowledge of the Cloud industry through user manuals, documentation, and blog posts.
