World’s Fastest Supercomputer – Facebook

Facebook is building and designing the World’s Fastest Supercomputer, the AI Research SuperCluster. RSC is an AI supercomputer that the company believe will be the world’s largest and fastest when fully built out.

So in early 2020, the company started with a clean slate to design a supercomputer that facebook envisioned to be the largest in the world. It will also meet the security requirements of our production systems and will provide the flexibility and developer experience that researchers need.

Facebook believe that a project of this complexity has never been attempted in the high-performance computing community.

One does not simply buy and power on a supercomputer. RSC was designed and executed under extremely compressed timelines without the benefit of a traditional product release cycle.

Additionally, the pandemic and a major industry chip supply shortage hit at precisely the wrong moment in time. The company had to fully utilize all of its collective skills and experiences to solve these difficult constraints.

Fundamentally, they’ve leveraged the best that everyone had to offer across people, technology, and partnerships to deliver and light up the ultimate in high-performance computing.

There are so many rewarding features of this project. One of the company’s favourite parts by far was when we rolled in the first rack. It was a breath of fresh air. It made this project feel real because we’ve just been sitting at home and planning and executing from our home space. To actually come to the site and see a rack rolling from the truck right into its final position was a really exciting moment, not only for any single employee but for all the cross-functional teams.

Every employee should be proud of is doing this with the team completely remotely. It is insane doing this without ever meeting anybody, the level of complexity.

With phase one, the company already has one of the largest and most performant supercomputers out there.

By the July timeframe, Facebook plans to triple its data centre in capacity and be up to 16,000 GPUs. The fact that Facebook is building this large a scale of a GPU cluster and turning it over to researchers who can use actual data on it is something we think is gonna be really interesting in the future of Metaverse.

More On TheFlashUpdate:

Facebook now called Meta announces CEO Mark Zuckerberg

Facebook, Twitter to be banned in India tomorrow? New guidelines