The Facebook and Instagram parent claims that the new chip ‘more than doubles’ the compute and memory bandwidth of its predecessor.
Meta has unveiled its next generation of custom-made chips to power its increasing AI workload across its family of apps.
In an update yesterday (10 April), Meta said that the new AI chip – known as Meta Training and Inference Accelerator, or MTIA – significantly improves performance compared to its predecessor and will help power its ranking and recommendation ads model across Facebook and Instagram.
“The next generation of Meta’s large-scale infrastructure is being built with AI in mind, including supporting new generative AI products, recommendation systems and advanced AI research,” the company wrote.
“It’s an investment we expect will grow in the years ahead, as the compute requirements to support AI models increase alongside the models’ sophistication.”
Meta said that the previous version of MTIA was designed specifically for its deep learning recommendation models aimed at improving “a variety of experiences” across its platforms and technologies.
Introducing the next generation of the Meta Training and Inference Accelerator (MTIA), the next in our family of custom-made silicon, designed for Meta’s AI workloads.
Full details ➡️ https://t.co/bF9tn4TfeJ pic.twitter.com/eti6fRCLTs
— AI at Meta (@AIatMeta) April 10, 2024
As artificial intelligence gets infused across multiple aspects of the business, the increasing workload and demand for compute power warrants a chip upgrade.
The company claims the new chip “more than doubles” the compute and memory bandwidth of its predecessor and that it can handle both low-complexity and high-complexity ranking and recommendation models – key components of its products.
“This chip’s architecture is fundamentally focused on providing the right balance of compute, memory bandwidth and memory capacity for serving ranking and recommendation models,” Meta wrote.
“MTIA has been deployed in our data centres and is now serving models in production. We are already seeing the positive results of this program as it’s allowing us to dedicate and invest in more compute power for our more intensive AI workloads.”
Meta recently shared details on two of its new GPU clusters that it is using to train its future AI models, including a successor to Llama 2. Both clusters contain more than 24,000 Nvidia H100 graphics processing units (GPUs) and are being used to support the company’s AI developments.
AI chips have been the centre of news this week. Both Intel and Google unveiled new, more powerful chips to power their AI workloads, while TSMC – one of the world’s largest chipmakers – posted soaring revenue amid higher demand for semiconductors.
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.
