POWERING THE METAVERSE
As we know, the metaverse has been one of the most recent buzzwords across the tech space and almost every digital-based brand has made mention of it to date. What most will not mention is the enormous amount of computing power needed to power such a concept. According to senior vice president of Intel, Raja Koduri stated that:
… our computing, storage and networking infrastructure today is simply not enough to enable this vision
Koduri goes on to predict that we will need a one-thousand times increase in computational efficiency. This is where Facebook’s parent company, Meta steps in. Meta announced earlier this year that it would be building the world’s most powerful AI-based supercomputer. The computer will feature enhanced speech recognition tools, automatic translation and additional software to aid in the development of the three-dimensional metaverse.
The AI Research SuperCluster (RSC) , although currently incomplete, has already replaced Meta’s previous supercomputer. The computer has 6,080 graphic processor units (GPUs) and a current performance score on par with the Perlmutter supercomputer at the National Energy Research Scientific Computing Center in California. Upon completion, the RSC is set to consist of 16,000 GPUs and estimated to be three times more powerful than it currently is.
Meta also seems to be taking security very seriously with this project as RSC will feature encrypted AI training and is wholly isolated from the larger internet. RSC has no inbound or outbound connections and the only allowable traffic will be from Meta’s production data centers.
So far, Meta has maintained relationships with a consistent roster across its three systems; Penguin Computing for managed services and architecture, Nvidia for systems/GPU, and Pure Storage for the majority of its storage solutions.
In conclusion, Meta’s research supercomputer is being built through 2022 and is expected to be the world’s fastest AI supercomputer performing at almost 5 exaflops of mixed precision computing according to the company.