Intel focuses on efficiency, AI wth upcoming Xeons

Dan Rodriguez, corporate vice president of Intel’s network and edge group

BARCELONA — Intel opened its presence at this week’s Mobile World Congress here by focusing on AI and power efficiency in the telco space with a pair of upcoming processors and a new development kit.

Appropriate for the MWC audience, Intel announced its next-generation Xeon processors for the 5G core. Code-named Sierra Forest and slated for release this year, this subset of Xeons will be heavy on high-efficiency cores, offering up to 288 e-cores on a single chip. It will be Intel’s first Xeon offering powered by efficiency cores, most commonly seen on mobile devices where power consumption is often valued over pure performance, and aims to use e-cores to fill the role the GPU has starred in thus far in the AI world — handling the massive amount of data to be managed and processed.

Dan Rodriguez, corporate vice president of Intel’s network and edge group, said those efficiency cores make it “just what operators need for greater efficiency,” offering up to 2.7x performance per rack compared to previous Xeon processors in this type of workload. The efficiency message should resonate well with telcos, for whom Rodriguez said energy consumption typically represents 20-40 percent of operators’ expenditures.

Intel also announced a 4th Generation Xeon processor that supports virtual radio access network (vRAN) technology. Code-named Granite Rapids-D, the new chips will debut next year, following the release of the more generalized Granite Rapids server chips slated for this year, but Rodriguez said Granite Rapids-D is “in testing now.”

Rodriguez said Granite Rapids-D would offer a 2x capacity boost and 20 percent power savings compared to previous-generation Xeons.

The company also announced a vRAN AI Development Kit to help network operators build, train, optimize and deploy AI models in vRAN use cases. Rodriguez noted that in the case of the mobile network, those use cases are different from the typical cloud-based model for AI applications.

“In telecom, you think about infusing AI into existing nodes and applications,” Rodriguez said. “Customers are looking at how AI can help with things like power efficiency and server utilization, and this development kit is targeted at those use cases. We believe it will jumpstart innovation in the value chain and help telcos better incorporate AI to make their networks more efficient.”

Perhaps most importantly, from an Intel perspective, it supports the company’s belief that carriers can do this by getting more out or more efficient CPUs, meaning telcos don’t need to add GPU horsepower to their core servers for many use cases.

Robert Dutt

Robert Dutt is the founder and head blogger at He has been covering the Canadian solution provider channel community for a variety of publications and Web sites since 1997.