Facebook teases AI supercomputer built to power the metaverse

Meta, the artist formally known as Facebook, is showing off its new AI Research SuperCluster which it claims will be the fastest AI supercomputer on the planet.

What’s under the hood of this absolute beast? 760 Nvidia DGX A100 systems comprising of 6080 GPUs communicate with each other via NVIDIA Quantum 1600 Gb/s InfiniBand two-level Clos fabric, and then there’s 175 petabytes of Pure Storage FlashArray, 46 petabytes of cache storage in Penguin Computing Altus systems, and 10 petabytes of Pure Storage FlashBlade.

Woof. And as if that’s not enough firepower, Meta plans to stuff in more components over the course of 2022 bringing the total number of GPUs up to 16000 and upgrading the storage system to be capable of a delivery bandwidth of 16 TB/s. The firm reckons all this will represent the fastest AI supercomputer in the world, performing at nearly 5 exaflops of mixed precision compute.

“Developing the next generation of advanced AI will require powerful new computers capable of quintillions of operations per second,” said the company in a blog. “Today, Meta is announcing that we’ve designed and built the AI Research SuperCluster (RSC) — which we believe is among the fastest AI supercomputers running today and will be the fastest AI supercomputer in the world when it’s fully built out in mid-2022. Our researchers have already started using RSC to train large models in natural language processing (NLP) and computer vision for research, with the aim of one day training models with trillions of parameters.”

So, that’s the AI Research SuperCluster. A massive system for AI programmes to live on that represents some sort of a landmark in terms of raw computer power. Why does Facebook want that in its basement caning the electricity bill? Does suggesting you might like to connect with old school friends (incorrectly) really require so much processing grunt?

There’s some nods towards it being able to help with language processing to help tackle harmful content, but the real drive of the project seems to be about powering the metaverse. Which makes sense really since the company has literally changed its name to Meta.

“We hope RSC will help us build entirely new AI systems that can, for example, power real-time voice translations to large groups of people, each speaking a different language, so they can seamlessly collaborate on a research project or play an AR game together,” the company continued in its blog. “Ultimately, the work done with RSC will pave the way toward building technologies for the next major computing platform — the metaverse, where AI-driven applications and products will play an important role.”

The metaverse is being hailed as the next stage of the internet by many proponents, and the way its mostly described at the moment boils down to a network of VR spaces that will open up new ways of communicating and collaborating, and there’s some heavy overlap with the gaming sector which is basically already doing a lot of that.

It is a bit of a buzz word still, and like when companies grandstand about 5G and smart-things, it has the tendency to feel a bit like the use cases aren’t exactly nailed down yet, but the spokesperson wants you to know that this company is guiding the revolution. Whatever it turns out to be.

Facebook has put more chips on the table than most with regards to the metaverse, and to be fair spending what must have been a fortune putting together this supercomputer to power it is certainly more than just verbal gesturing. Unfortunately aside from the fact it will underpin the AI is clearly thinks is going to be key to the whole affair, there’s not much more revealed as to what Meta thinks the metaverse is, other than on the horizon and cool. The question of what exactly we will be doing in ten years when we ‘go on the metaverse’ is the wall we hit with any discussion on the subject.

If you’d like to know more about Meta’s general outlook – here’s a video it put out last October.