Meta has done something that will get Nvidia and AMD very, very worried — it gave up on GPU and CPU to take a RISC-y route for AI training and inference acceleration

The Meta logo on a smartphone in front of the Facebook logo a little bit blurred in the background
(Image credit: Shutterstock / rafapress)

The RISC-V movement has been given a massive boost after Meta revealed it is primed to begin mass development of hardware and products with the technology at its core.

RISC, an alternative Instruction Set Architecture (ISA) to the x86 architecture used by the likes of Intel and AMD in their best CPUs, is dominated by Arm. This ISA is largely able to replicate many of the functionalities of the best GPUs and processors out there. 

RISC-V is an open source alternative that appeals because it lowers the barrier of entry due to its open nature and the fact organizations that use don't need to pay fees to license use of CPUs. In recent months, companies from Alibaba to Qualcomm have thrown their weigth behind the RISC-V ecosystem – announcing plans to migrate to varying degrees.

Why RISC-V makes sense for Meta's AI push

Now, Meta is migrating away from CPUs to RISC-based components due to the power efficiency of the ISA, performance, lower latency, and the flexibility to support different workloads. This is according to its senior director of engineering, speaking at the RISC-V Summit, as reported by Next Platform.

Meta's gamble on RISC-V is in a phase of acceleration, following four years of planning, and the firm is not only rolling out production hardware – but getting ready to manufacture future custom RISC-V silicon. 

The firm has, for instance, built some of its video transcoding hardware on RISC-V alone – replacing 85% of the CPUs. Known as the Meta Scalable Video Processor (MSVP), this piece of kit is already deployed and handles all video uploads across a swathe of social media services, including Facebook, Messenger and Instagram. 

But the firm is also eyeing up RISC-V for use in developing a chip that could be used in AI training and inference, at a time when most of the industry is clamoring to get its hands on Nvidia's industry-leading GPUs. 

The A100, followed by the H100 and H200 chips have been pivotal to the rise of generative AI, but Meta is planning to skip this stage altogether and instead throw itst weight behind RISC-V when it comes to building AI processors.

Meta's first in-house RISC-V silicon for AI acceleration is a 7nm component that operates at a frequency of 800MHz, has 128MB on-chip memory and supports up to 128GB LPDDR5 RAM. They're currently dedicated to accelerating models in both training and inference.

More from TechRadar Pro

Keumars Afifi-Sabet
Channel Editor (Technology), Live Science

Keumars Afifi-Sabet is the Technology Editor for Live Science. He has written for a variety of publications including ITPro, The Week Digital and ComputerActive. He has worked as a technology journalist for more than five years, having previously held the role of features editor with ITPro. In his previous role, he oversaw the commissioning and publishing of long form in areas including AI, cyber security, cloud computing and digital transformation.