There is sort of a bottleneck between CPUs and purpose-built accelerator chips in data centres. That means that rapidly growing data workloads, including artificial intelligence and machine learning, are having a hard time developing and growing at maximum pace.
Today, a group of tech giants joined forces to create a new standard with the goal of eliminating these bottlenecks and breathing new wind into AI's sails.
Called Compute Express Link (CXL), the standard will provide an ultra-fast interconnect between data centre CPUs and accelerator chips.
The consortium includes Intel, Microsoft, Alibaba, Cisco, Dell EMC, Facebook, Google, Hewlett Packard Enterprise, and Huawei.
“CXL is an important milestone for data-centric computing and will be a foundational standard for an open, dynamic accelerator ecosystem,” said Jim Pappas, director of technology initiatives at Intel, in a statement.
“Like USB and PCI Express, which Intel also cofounded, we can look forward to a new wave of industry innovation and customer value delivered through the CXL standard.”
For starters, the CXL Specification 1.0 has been ratified. It will improve communication between CPUs and devices like secondary processors or accelerators. It will also create an improved interconnect, boosting memory coherency.
The accelerators include GPUs and field-programmable gate arrays, all of which was built on the PCI Express infrastructure.
More details about the standard can be found on this link.
Image Credit: Welcomia / Shutterstock