On Tuesday, Intel released its first public declaration on the metaverse, an often hazy future of computing that promises a constantly connected virtual world that exists in parallel with our physical one. However, while Intel is optimistic about the metaverse’s potential, in theory, it presents a significant challenge with accomplishing any metaverse ambitions: there isn’t nearly enough computing power to go around.
“After the web and mobile, the metaverse may be the next big computing platform,” writes Raja Koduri, senior vice president and head of Intel’s Accelerated Computing Systems and Graphics Group. However, Koduri swiftly dispels the notion that the metaverse is just around the corner, writing, “Our processing, storage, and networking infrastructure today is just not enough to enable this goal.” Most importantly, Koduri doesn’t believe we’re even near. He claims that a 1,000-fold increase in computer power is required above our existing capabilities.
Much of the hoopla surrounding the metaverse has centred on what you’ll be able to do there, whether it’s virtual reality meetings, digital concerts, or, of course, blockchain and NFT-based integrations. And, whether it’s Meta’s Quest devices (previously known as Oculus) or Apple’s long-rumoured headset, there’s plenty of excitement about the future of virtual and augmented reality headsets.
But the metaverse’s actual building blocks won’t just be software and virtual places (which, of course, is a separate battle, given how self-contained today’s digital worlds are), or even the headsets and gadgets people need to “get” there. It’ll be in the computers and servers that power the metaverse’s massive shared virtual worlds, which the metaverse envisions as the technological future. And it’s there that Intel delivers the most sobering reality check: today’s computers are simply not powerful enough to realise those goals. They don’t even come close.
On the one hand, the assertion is so apparent that it’s almost comical. Horizon Worlds, Meta’s flagship VR space, has a cap of 20 participants per space, and that’s for simple, Roblox-style animated worlds. The state of the art in VR still necessitates thousands of dollars in PC gaming hardware, with numerous drawbacks (such as the need for a tethered headset and graphics that fall short of what 2021’s top flatscreen games can provide). Even the most popular traditional video games, such as Fortnite or Battlefield 2042, which don’t have to deal with the additional demands of VR, can only manage 100 to 128 players at a time. With today’s technology, we can’t even place two individuals in a genuinely detailed virtual environment, as Koduri points out in his editorial. “Think about what it takes to put two people in a social setting in a completely virtual environment: convincing and detailed avatars with realistic clothing, hair, and skin tones – all rendered in real-time and based on sensor data capturing real-world 3D objects, gestures, audio, and much more; data transfer at super-high bandwidths and extremely low latencies; and a persistent model of the environment, which may contain both real and simulated elements.”
And that’s only for two people; scaling up to the hundreds of millions of users required by a Ready Player One, Snow Crash, or Matrix-style metaverse notion would necessitate a massive amount of processing power.
Intel, of course, has a vested interest in claiming that more and better computers and servers are required. Intel, after all, makes CPUs (and, shortly, GPUs) for both consumer electronics and data centres. And if the metaverse – the future’s trendiest buzzword technology — requires a 1,000x boost in computer power, well, that’s just smart business. It’s no accident that Intel’s metaverse brief specifically mentioned both its client computing and cloud processors, as well as its graphics offerings. The difficulty is that even Intel does not believe that hardware will be enough to get us to 1,000x. “We believe that a conventional Moore’s Law trajectory will only get us to roughly eight or ten times growth over the next five years,” Koduri said in an interview with Quartz. (Moore’s Law states that processing capability doubles every two years, which corresponds to the eight to tenfold growth predicted by Koduri.)
Instead, Koduri is optimistic that improvements in algorithms and software will close the gap. Machine learning-powered neural nets or AI-enhanced computing approaches like Intel’s Deep Link technology or the future XeSS supersampling it plans to unveil with its Arc GPUs early next year are examples. It’s a huge ask, though: Intel is banking on algorithms or artificial intelligence to provide a hundredfold (or more) increase in computer capacity, on top of the company’s current hardware strategy.
Improved software and algorithms, according to Koduri in the same Quartz interview, will be necessary not only to close the gap in the ambitious five-year timeframe he lays out but also to help mitigate the increased energy consumption that brute-forcing the problem would create, which he compares to the current problems with cryptocurrency mining today.
It’s simple to wave a hand and say that software will fill in any holes left by hardware (especially for a company like Intel, which primarily makes the hardware). Many major tech companies have flocked to the idea that AI and machine learning will solve their computation problems, from improving smartphone cameras to providing upscaled gaming visuals, and it’s tempting to believe they will. However, relying on them to 100x tomorrow’s computing, which is only expected to experience a 10x increase based on hardware upgrades alone, appears to be a tough order.
However, the fact that Intel is thinking about it now — and stating the problem — is a positive sign. It’s tempting to get caught up in the hoopla and start suggesting outlandish concepts for selling NFTs that will follow you about in various games and virtual environments. Building up server infrastructure and striving to minimise latency is less exciting, but as Intel’s presentation demonstrates, there’s a lot more foundational work to be done in the coming years if the metaverse is to achieve its sci-fi goals.