Intel, AMD, Microsoft & Others Team Up To Develop The “UALink”, Direct Competitor To NVIDIA NVLink

May 30, 2024 at 10:50pm EDT
Intel, AMD, Microsoft & Others Team Up For Develop The "UALink", Direct Competitor To NVIDIA NVLink 1

NVIDIA's dominance of the NVLink interface faces major competition as Intel, AMD, Microsoft & others have teamed up to create UALink.

It Seems Like The Tech Industry Looks To Break NVIDIA's Monopoly In The Interconnect Market With UALink, Microsoft, AMD, Intel & Others Onboard

Well, this isn't the first time we have seen big tech giants teaming up against NVIDIA to break their AI monopoly since the last instance we saw was when Intel said that they and others plan to overthrow CUDA dominance but look where we are.

Related Story AMD Is Trying to Pull Off a ‘Ryzen Moment’ With Radeon, But Admits the Perfect Gaming GPU Platform Is Still Generations Away

Moving from this, major tech firms such as Intel, AMD, Broadcom, Cisco, Google, HPE, Meta, and Microsoft have decided to form an Ultra Accelerator Link Promoter Group focused on developing an AI-focused interconnect system targeted specifically toward data centers. The group has announced intentions to develop the UALink (Ultra Accelerator Link), which aims to connect accelerators through an open protocol.

Well, the idea behind UALink is simple. The goal here is not only to provide an alternative to NVIDIA's NVLink but also to provide an option that is more powerful and accessible across a network of accelerators.

We all know that NVIDIA is confined to its ecosystem and doesn't allow support across multi-company products, which is one of the reasons why Team Green has managed to receive the AI hype it currently has. Still, with UALink, the collaboration focuses on providing a more scalable and future-proof network, courtesy of the wide range of companies involved in the process.

Diving a bit into specifics, the UALink will be based upon memory semantic fabric, which will contribute massively to the performance and scalability of the ecosystem. Moreover, the UALink Promoter Group plans to release the initial 1.0 specification by Q3 2024. The firm says that the platform will be able to scale up to 1,024 AI accelerators, and an update to increase the bandwidth is scheduled for Q4 2024. It will serve as a direct load, store, and atomic operations interface between AI GPUs. The role of Ultra Ethernet for deployment will also be crucial here for scalability and connectivity.

With that said, it's interesting to see what sort of competition the anticipated UALink poses to NVIDIA's dominance in the markets, and because big names are involved in this consortium, it will be something to set sight upon.

About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.

Follow Wccftech on Google to get more of our news coverage in your feeds.