Intel Hires Top AMD Data Center GPU Exec Ogi Brkic For HPC Role

The semiconductor giant confirms to CRN that it has hired Ogi Brkic as vice president and general manager of the Super Compute product line within the chipmaker’s Accelerated Computing Systems and Graphics Group. The hire was made ahead of Intel’s plan to challenge AMD and Nvidia with its Ponte Vecchio GPU next year.

ARTICLE TITLE HERE

Intel has hired a top data center GPU executive from AMD to lead the product side of the chipmaker’s new high-performance computing group.

The Santa Clara, Calif.-based company confirmed to CRN that it recently hired Ogi Brkic as vice president and general manager of the Super Compute product line within the chipmaker’s Accelerated Computing Systems and Graphics Group that was formed in June to accelerate Intel’s GPU and HPC efforts.

[Related: AMD Continues To Steal Top 500 Supercomputer Share From Intel]

id
unit-1659132512259
type
Sponsored post

The hire was made ahead of Intel’s plan to challenge AMD and Nvidia in HPC and AI applications with its Ponte Vecchio GPU, which will power the U.S. Department of Energy’s Aurora exascale supercomputer that is scheduled to go online next year. Brkic is one of several executives or senior staff members from AMD to join Intel over the past few years in its efforts to build a discrete GPU business.

Brkic, who revealed Monday that he started at Intel in October, was previously corporate vice president and general manager of the data center GPU business unit at AMD, where he worked for 15 years, according to his LinkedIn profile. An AMD spokesperson confirmed his departure but did not provide any additional information.

According to Intel, Brkic is part of the new Super Compute Group led by Jeff McVeigh that was formed within the Accelerated Computing Systems and Graphics Group in September to focus on HPC applications for the chipmaker’s CPUs and its upcoming slate of discrete GPUs.

Brkic has his own team, which is responsible for both CPU and GPU products and businesses for supercomputing and visual cloud applications, Intel said. This includes “defining market needs, product positioning and ecosystem,” the company added, and his team will work closely with the Super Compute Platform Engineering team led by Brijesh Tripathi, which focuses on product design.

Brkic departed from AMD before the chipmaker in October revealed the Instinct MI200 series, a new line of GPU accelerators it said will provide up to 4.9 times faster HPC performance and up to 20 percent faster AI performance compared with the 400-watt SXM version of Nvidia’s flagship A100 GPU.

Intel has large ambitions in the GPU space—with plans for products in several segments, including gaming —and it has relied on veteran executives, architects and designers from AMD to build out its expertise. This started with the hiring of AMD’s former chief GPU architect, Raja Koduri, in 2017 and continued with several other big hires over the past few years, including Masooma Bhaiwala in 2019, Ali Ibrahim in 2020 and, most recently, Vineet Goel in October.

Dominic Daninger, vice president of engineering at Nor-Tech, a Burnsville, Minn.-based HPC system integrator, told CRN that Intel’s latest hire shows that the chipmaker is serious about competing with AMD and Nvidia in the GPU space after previous false starts with accelerators like Xeon Phi.

“It’s pretty clear that with Intel, somebody tickled the bear there and they’re awake again,” he said. “I’ve seen that happen over and over again with Intel: They’ll get a little comfortable there, and then AMD will come up and beat them to 1GHz or something like that. And then Intel reaches into its pockets there and catches up and passes them again.”

It’s a tall order to make a new line of GPUs that will compete with AMD and Nvidia for the years to come, Daninger said, but what might make Intel’s efforts different this time is the fact that new advances are enabling CPUs and GPUs to share memory.

“The unified memory that multiple vendors are going for is going to help, and I think just the physics of this is pushing all of the vendors that direction,” he said.

Daninger said he is already seeing the benefits of AMD’s increased competitiveness against Nvidia in the HPC space with its latest GPUs, and that means the introduction of Intel as the third major player in the market could bring even more advantages for channel partners by giving them an additional choice.

Nor-Tech is currently working on what could be the company’s most significant project yet with AMD’s data center GPUs, Daninger said, and the price-performance benefits of AMD’s GPUs is one major reason why Nor-Tech’s customer in this case is going with AMD versus Nvidia.

“We could just get a lot more compute power for X dollars then we could going with the default, which would have been Nvidia,” he said.

Intel’s entry into the GPU market also means that channel partners will likely have more supply to work with as GPU shortages continue to hamper projects, Daninger said.

“That’s one of the real pain points right now: the availability of GPUs,” he said. “We’ve had projects sitting for four or five months waiting for some GPUs, so it’s extremely tight.”