With its AI compute needs set to grow dramatically, Meta has announced custom chips for running AI models and video, a new AI-optimised data centre design, and phase 2 of its 16 000 GPU supercomputer for AI research.
Santosh Janardhan, head of infrastructure at Meta, writes: “These transformational efforts — and additional projects still underway — will enable us to develop much larger, more sophisticated AI models and then deploy them efficiently at scale.”
The company is also deploying Code Compose, a generative AI-based coding assistant developed in-house as a key tool to make developers more productive throughout the software development lifecycle.
“By rethinking how we innovate across our infrastructure, we’re creating a scalable foundation to power emerging opportunities in the near term in areas like generative AI, and in the longer term as we bring new AI-powered experiences to the metaverse,” according to Janardhan.
Meta has announce MTIA (Meta Training and Inference Accelerator), an in-house, custom accelerator chip family targeting inference workloads. MTIA provides greater compute power and efficiency than CPUs, and it is customised for Meta’s internal workloads.
“By deploying both MTIA chips and GPUs, we’ll deliver better performance, decreased latency, and greater efficiency for each workload,” Janardhan discloses.
The company also debuted MSVP, its first ASIC for video transcoding designed for the processing needs of the ever-growing video-on-demand (VOD) and live streaming workloads at Meta.
MSVP is programmable and scalable, and can be configured to efficiently support both the high-quality transcoding needed for VOD as well as the low latency and faster processing times that live streaming requires.