Intel sees metaverses needing '1000x more compute capacity'

After Nvidia outlined its own metaverse vision last month, it put a little bit of pressure on other semiconductor companies to say something about how they plan to support rapidly expanding virtual worlds, and the interactions between physical and virtual worlds.

This week, Intel stepped up with a metaverse-themed keynote discussion at the RealTime Conference 2021, describing a vision of multiple metaverses and metaverse types–social, gaming and professional, and all of them containing extremely realistic people, objects, structures and locations–to be enabled by multiple companies. 

Raja Koduri, senior vice president and general manager of the Accelerated Computing Systems and Graphics (AXG) Group at Intel, described these metaverses collectively as the concept of an “immersive internet” that will be available to billions of people around the world, with the aim that the usage experience for all users will be consistent and and “preserved exactly as users switch devices and change locations.

“If you think about your favorite vacation destination, everyone may have a different idea of where they want to go,” he said. “There won’t be just a single metaverse from a single company like in ‘Ready Player One’ because no one wants this to be proprietary to one company.”

If it sounds like a lot more computing power will be needed to realize that vision, that was Intel’s biggest point, too. Anton Kaplanyan, vice president of graphics research at Intel said, “Just imagine the plumbing that needs to be updated to support all that stuff, that kind of immersive Internet. We’ll need 1,000-times more compute capacity than we have today.”

Intel plans to help support the builders of those metaverses with new processors like the CPUs, GPUs and IPUs it announced last August. The GPUs, the first of which will be available in the first quarter of 2022, will be key to the realistic feel and these extremely detailed rendering of the metaverses, Intel officials said.

But they also made clear that even as processing power ramps up in the years to come, there is an even more immediate need to help companies access and put to work compute capacity that they already have, but is underused. Intel said it aims to enable an “intelligent infrastructure layer” of software that will identify unused compute capacity that can be reallocated to help applications and experiences perform better.

RELATED: Intel to invest $7B in chip fab for Penang, report says