Intel Corp. unveiled a series of new server processors on Tuesday that it claims are 1.65 times faster than its predecessors.

Intel’s Xeon Scalable chip series, which ranges from eight processing cores all the way up to 28 cores, are designed to counter a new threat to the company’s server chip business from rival Advanced Micro Devices Inc. AMD recently launched its own Zen-based server chips. Like AMD, Intel says its new lineup is its most competitive breed of server chips in the last decade.

In a blog post, Lisa Spelman, Intel’s vice president and general manager of the data center marketing group, said the new chips have the potential to boost the core computing foundations of almost every industry thanks to their enhanced artificial-intelligence processing capabilities.

With the new chips, Intel is aiming to dominate compute, storage and communications across the data center. For example, Spelman said the Intel Xeon Scalable family can support more than four times as many virtual machines, and as many as five times more transactions per second than the company’s previous-generation data center chips. When configured for storage, the chips deliver five times more operations with 70 percent lower latency, she claimed. Meanwhile in communications, the chips are designed to power next-generation 5G networks with a 2.5 times performance boost.

The performance gains come thanks to a new architecture design that changes the way data flows across Intel’s processors. Previous generation chips used something called a “ring architecture” that sent data in “laps” around the circuit to meet the desired processing core or memory interconnect. With the new “mesh architecture,” data can now flow directly between cores and interconnects.

At its unveiling event Tuesday, Intel reeled off the names of a pack of early adopters that say they’ve experienced some serious performance gains since adopting the new chips. These include mobile operator AT&T Inc., which is already running production traffic and benefits from using 25 percent fewer servers per cluster with the new chips.

Meanwhile, five customers using Google Inc.’s compute platform said they’ve seen performance benefits of up to 40 percent for regular workloads. However, when those applications include optimization for Intel’s AVX-512 instructions, which are accelerators designed to boost performance for workloads and usages such as scientific simulations, AI, 3D modeling, image and audio/video processing, those customers saw performance more than double.

Customers have reported benefits in other areas such as virtual reality too, with media firm Technicolor SA saying it can render VR content almost three times faster than before.

Those performance gains should come in handy for Intel as it faces a fierce struggle to retain its dominance of the chip market for corporate servers and cloud data centers. The company currently has about a 90 percent share of that market, but it’s coming under serious attack from AMD, which says its new Epyc chips are cheaper, provide better performance and also use less energy than Intel’s offerings.

Intel is also feeling threatened by Nvidia Corp., which is pushing its graphics processing unit cards as a better platform for running emerging technologies like AI and machine learning. Intel is no doubt also concerned at the news that companies like Google are said to be working on their own, customized data center chips as well.

With regards to AMD, Intel yesterday claimed that the Xeon Scalable family, particularly the high-end Intel Xeon Platinum 8180, offers a 28 percent faster performance than AMD’s newest Epyc 7601 chips.

Those claims cannot be immediately verified, but Patrick Moorhead, president and principal analyst at Moor Insights & Strategy, said it was important to note that Intel is pitching more than just a central processing unit with its latest effort.

“Intel is bringing CPUs plus chipsets, accelerators, Optane SSD, buses and optimized software required for workload-optimization,” Moorhead said. “Intel is even bringing near-engineered solutions to the table with Select Solutions, an indication that it is moving up the food chain and also enabling a much larger market-basket.”

Because the Xeon family is built for targeted workloads such as high performance computing, cloud and AI, Moorhead believes that Intel will not only ward off the threat from AMD, but actually increase its market share. “By increasing the ‘market basket’ with workload-optimized technologies and Select Solutions, I see growth potential for the company in spaces they haven’t historically achieved high degrees of market share while defending their CPU turf,” he said.

Intel still needs to come up with something to deal with the threat from Nvidia and other companies’ graphics chips, which are still viewed as a superior platform for most AI workloads. Intel’s answer to that threat is said to be a work in progress, as the company continues to integrate technology it acquired when it bought out the two-year-old AI chip startup Nervana Systems last year.

Image: Intel