UALink 200G 1.0 Specification Published For Connecting Up To 1,024 Accelerators
UALink as the open standard alternative to NVIDIA NVLink that is backed by Intel, AMD, and other vendors has published the UALink 200G 1.0 specification.
The Ultra Accelerator Link 200G 1.0 specification was published today for this scale-up interconnect focused on next-gen AI workloads. The UALink 200G 1.0 specification is designed for 200G per lane connectivity that can scale up to a possible 1,024 accelerator/GPU limit within an AI computing pool. UALink 200G 1.0 is an open standard unlike NVIDIA's NVLink.
UALink is focused on being a low-latency, high performance interconnect while being low-power and cost efficient.
The UALink 200G 1.0 specification is publicly available for download from UALinkConsortium.org. More details on today's 1.0 specification release via the press release.
The Ultra Accelerator Link 200G 1.0 specification was published today for this scale-up interconnect focused on next-gen AI workloads. The UALink 200G 1.0 specification is designed for 200G per lane connectivity that can scale up to a possible 1,024 accelerator/GPU limit within an AI computing pool. UALink 200G 1.0 is an open standard unlike NVIDIA's NVLink.
UALink is focused on being a low-latency, high performance interconnect while being low-power and cost efficient.
The UALink 200G 1.0 specification is publicly available for download from UALinkConsortium.org. More details on today's 1.0 specification release via the press release.