A consortium of technology companies has formed to create a new standard, called Compute Express Link, to provide an extra-fast interconnect between datacenter central processing units (CPUs) and accelerator chips.

The consortium includes Intel, Microsoft, Alibaba, Cisco, Dell EMC, Facebook, Google, Hewlett Packard Enterprise, and Huawei. The aim is to deliver breakthrough datacenter performance that can help computers keep up with the explosion of data in the age of the internet of things (IoT).

Navin Shenoy, executive vice president and general manager of the Data Center Group at Intel, said in a blog post that CXL will remove the bottlenecks between CPUs and purpose-built accelerator chips in the datacenter. It’s an example of how Intel sees itself not just as a CPU company, but also as a shepherd for the entire PC and server business. As the company said at its recent Architecture Day event, its focus is on processing, architecture, memory, software, security, and interconnect.

The new Compute Express Link (CXL) will accelerate rapidly growing data workloads, such as artificial intelligence and machine learning, rich media services, high-performance computing, and cloud applications.

“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.”

The group has ratified the CXL Specification 1.0, which will improve the communication between CPUs and other devices, such as secondary processors or accelerators. It will provide improved interconnect and better memory coherency for higher performance of data-intensive applications.

And the new open standard will help foster an open accelerator ecosystem for high performance, heterogeneous computing. Interested member companies and institutions are encouraged to join. It’s worth noting that Intel’s rival, Advanced Micro Devices, is not yet part of the consortium. The CXL is an open industry standard aimed at supporting an open ecosystem for datacenter accelerators and other high-speed enhancements. The 1.0 spec is available now.

“Google supports the open Compute Express Link collaboration,” said Rob Sprinkle, technical lead for platforms infrastructure at Google, in a statement. “Our customers will benefit from the rich ecosystem that CXL will enable for accelerators, memory, and storage technologies.”

Intel's Navin Shenoy said Xeon AI revenues were $1 billion in 2017.

Above: Intel’s Navin Shenoy

Image Credit: Dean Takahashi

The accelerators include graphics processing units (GPUs), field-programmable gate arrays (FPGAs) and other purpose-built accelerator solutions. The technology is built upon the well-established PCIExpress (PCIe) infrastructure, leveraging the PCIe 5.0 physical and electrical interface to provide advanced protocol in three key areas: input/output protocol, memory protocol, and coherency interface. These allow for memory sharing across different computing resources, reducing the software stack complexity and overall system cost. This permits users to focus on target workloads as opposed to the redundant memory management hardware in their accelerators.

“Microsoft is joining the CXL consortium to drive the development of new industry bus standards to enable future generations of cloud servers,” said Leendert van Doorn, distinguished engineer for Azure at Microsoft, in a statement. “Microsoft strongly believes in industry collaboration to drive breakthrough innovation. We look forward to combining efforts of the consortium with our own accelerated hardware achievements to advance emerging workloads, from deep learning to high performance computing, for the benefit of our customers.”

CXL was designed to be an industry open standard interface for high-speed communications, as accelerators are increasingly used to complement CPUs in support of emerging applications, such as artificial intelligence and machine learning. The specification is available to companies that join the CXL Consortium.

“Facebook is excited to join CXL as a founding member to enable and foster a standards-based open accelerator ecosystem for efficient and advanced next-generation systems,” said Vijay Rao, director of technology and strategy at Facebook, in a statement.

“The CXL is a direct competitor to CCIX and IBM’s OpenCAPI and Nvidea’s NVLink,” said Kevin Krewell, analyst at Tirias Research. “This is Intel’s alternative to CCIX. Both CXL and CCIX use PCIe as the underlying electrical connection. It’s unfortunate that there’s two directly competing standards trying to do the same thing — connecting CPU and accelerators (like neural net chips, GPUs, and FPGAs) in a memory coherent protocol (PCIe is not a coherent interface).”

He added, “CXL was going to be … proprietary, but instead, at the request of Intel’s customers, the company has opened it as a consortium.  Still, this will complicate building heterogeneous systems by having two opposing groups. I hope that at some time in the future the two groups can find a middle ground to merge the specifications.”