The UALink Consortium, an industry body developing an open source alternative to Nvidia’s NVLink interconnect standard for AI workloads, has just published version 2.0 of the UALink specification.
With this release, UALink introduces the UALink 200G Data Link and Physical Layers (DL/PL) Specification 2.0, which separates the DL/PL Specification from the original UALink Common Specification, thus enabling the body to move quickly as new physical layers and speeds come into play, and removing the need to make changes to other of its specifications.
The new UALink Common Specification 2.0 also includes In-Network Compute for UALink technology, which facilitates computation and communication between accelerators, as well as reducing latency and improving operating efficiencies. The new spec also supports the deployment of the technology in multi-workload environments.
Other new components include version one of a Manageability Specification that allows tools such as gNMI, Yang, SAI and Redfish to be used with the system, to enable better management, and a v1 Chiplet specification that allows integration of UALink into chiplet-based Systems on Chip (SoCs).
The new standards will help companies keep pace with the demands placed on them by the rapid deployment of AI, the consortium said.
“The advancements to UALink technology introduced in this release will enable the industry to quickly and efficiently integrate UALink solutions into their architectures. The UALink Consortium remains committed to advancing AI infrastructure through open industry standard technology that facilitates next-generation AI applications to the market,” said Kurtis Bowman, UALink Consortium board chair, in a press release.
But given there are no products currently available using UALink 1.0, UALink 2.0 might be viewed as a premature launch
Need to play catch up
David Harold, senior analyst with Jon Peddie Research, was guarded in his reaction. “While 2.0 is a significant step forward from 1.0, we need to bear in mind that even 1.0 solutions aren’t shipping yet – they aren’t due until later this year. So, Nvidia is way ahead of the open alternatives on connectivity, indeed ahead of the proprietary or Ethernet based solutions too,” he said.
What this means, he added, is that non-Nvidia alternatives are currently lagging in the market. “They need to play catch up on several fronts, not just networking. … I can’t think of a single shipping product that meaningfully has advantages over a Nvidia solution,” he said. “Ultimately UALink remains desirable since it will enable heterogeneous, multi-vendor environments but it’s quite a way behind NVLink today. ”
There are plenty of signs that organizations will find it hard to break free of the Nvidia dominance, however. A couple of months ago, RISC-V pioneer SiFive signed a deal with Nvidia to incorporate Nvidia NVLink Fusion into its data center products, a departure for RISC companies.
According to Harold, other companies could be joining it. “Custom ASIC company MediaTek is an NVLink partner, and they told me last week that they are planning to integrate it directly into next-generation custom silicon for AI applications,” he said. “This will enable a wider range of companies to use NVLink as their high-speed interconnect.”
Other options
And, Harold noted, Nvidia is already looking at other options. “Nvidia is now shifting to look at the copper limit for networking speed, with an interest in using optical connectivity instead,” said Harold.
However, this also provides another opportunity for the UALink Consortium. “UALink can ultimate support this too, and perhaps a focus on this area will let UALink close the gap. Signaling suitability for this might be part of the timing on announcing UALink 2.0 now,” he said.