Meta (formerly Facebook) is now building an AI supercomputer called AI Research SuperCluster (RSC). The company claims that the supercomputer is already among the fastest machines and, when completed by mid-2022, will be the world’s fastest.
The RSC is a high-speed computer explicitly designed to train machine learning systems. Meta will use this AI supercomputer to train business systems for content moderation algorithms and augmented reality features.
“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.”Meta engineers Kevin Lee and Shubho Sengupta
Meta engineers started designing the AI supercomputer a year and a half ago. The RSC is already in its first phase and has over 760 Nvidia GGX A100 systems containing 6,080 connected GPUs. These types of GPUs are great for solving machine learning problems.
Phase two of the RSC will be completed by the end of 2022 and will contain over 16,000 total GPUs. For comparison, Microsoft’s AI supercomputer built with research lab OpenAI only has 10,000 GPUs. For a list of the world’s fastest supercomputers, take a look at Top500.org.
AI supercomputers are better than other high-performance computers as they can carry out more calculations per second using the same hardware. Even with this much raw computing power, Meta needs to up its game to compete with companies like Google, Microsoft, and Apple.