The industry association working to establish an open standard for GPU interconnectivity has released its initial specification, which seeks to compete with Nvidia's proprietary technology while greatly enhancing AI computing performance.

The Ultra Accelerator Link (UALink) Consortium, established last summer, unites major players in semiconductors and data centers, such as AMD, Intel, and Microsoft, to create standards aimed at improving the connectivity of multiple GPUs.

The organization has now formally approved the UALink 200G 1.0 Specification, enabling up to 1,024 AI accelerators to interact with one another at high speeds within computing clusters.

The Consortium aims to improve the efficiency of connected chips executing AI tasks while decreasing dependence on NVLink, Nvidia's exclusive interconnect technology.

“As the demand for AI compute grows, we are delighted to deliver an essential, open industry standard technology that enables next-generation AI/ML applications to the market,” said Kurtis Bowman, board chair of the UALink Consortium and director for architecture and strategy at AMD.

“The groundbreaking performance made possible with the UALink 200G 1.0 Specification will revolutionise how Cloud Service Providers, System OEMs, and IP/Silicon Providers approach AI workloads.”

The UALink standard establishes a switch ecosystem for accelerators, facilitating inter-hardware communication among system nodes to develop multi-node systems for AI applications.

The specification aims to significantly lower latency in computing clusters while offering an uncomplicated load/store protocol that matches Ethernet's raw speed yet maintains the latency level of PCIe switches.