UPDATED 16:53 EDT / MARCH 11 2019

INFRA

Intel launches new technology standard to link CPUs with accelerator chips

Intel Corp. today introduced a new interconnect technology aimed at enabling servers to make better use of graphics cards, field-programmable gate arrays and other accelerator chips optimized for specific workloads.

The Compute Express Link, or CXL, is envisioned as an industry standard that will work with products from multiple hardware makers. To that end, Intel has donated the technology to a newly formed consortium that will enable companies interested in using it to join as members. The group boasts more than a half-dozen backers on launch, including big names such as Cisco Systems Inc. and Google LLC.

CXL is designed to connect a server’s central processing unit with the secondary accelerator chips attached to the machine. Companies are rapidly adopting accelerators, particularly graphic cards, to speed up specialized workloads such as artificial intelligence software.

CXL is based on the widely used PCIe standard for connecting server components. It uses the latest, fifth iteration of the standard, which was formalized earlier this year and provides the ability to transfer a massive 128 gigabytes per second in certain configurations.

Intel has made a number of customizations to harness this speed for interchip communications. One of CXL’s key features is a mechanism that enables a CPU to share memory with accelerators, an essential part of the processing workflow. Intel said that feature replaces the dedicated hardware normally needed for the task and thus improves system efficiency. 

“CXL maintains memory coherency between the devices, allowing resource sharing for higher performance, reduced software stack complexity and lower overall system cost,” Navin Shenoy, the head of Intel’s Data Center Group, wrote in a blog post. “Emerging data-processing applications in AI, media, image and language processing, encryption and others will benefit significantly from CXL.”

The launch of the interconnect comes as Intel steps up its efforts to capture market share in the accelerator market. In the past month alone, the company introduced a field-programmable gate array for processing network traffic and a three-chip PCIe card designed to improve server security.

Even more significantly, Intel plans to launch its first line of standalone graphics cards next year. CXL will give the company an answer to the NVLink interconnect that Nvidia Corp. ships with its competing chips, which dominate the category.  

Photo: JiahuiH/Flickr

A message from John Furrier, co-founder of SiliconANGLE:

Support our open free content by sharing and engaging with our content and community.

Join theCUBE Alumni Trust Network

Where Technology Leaders Connect, Share Intelligence & Create Opportunities

11.4k+  
CUBE Alumni Network
C-level and Technical
Domain Experts
15M+ 
theCUBE
Viewers
Connect with 11,413+ industry leaders from our network of tech and business leaders forming a unique trusted network effect.

SiliconANGLE Media is a recognized leader in digital media innovation serving innovative audiences and brands, bringing together cutting-edge technology, influential content, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios — such as those established in Silicon Valley and the New York Stock Exchange (NYSE) — SiliconANGLE Media operates at the intersection of media, technology, and AI. .

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a powerful ecosystem of industry-leading digital media brands, with a reach of 15+ million elite tech professionals. The company’s new, proprietary theCUBE AI Video cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.