Uh oh, Intel and AMD – Meta is making strides towards its own custom chips

Meta has poached a senior silicon engineer from Intel in a move that appears designed to kick-start the development of custom solutions. server chips for its data centers.

Jon Dama joined Meta earlier this month as Director of Silicon for the Infrastructure Hardware Group, a role in which he will support multiple design teams focused on “large-scale data center innovation.” “.

Meta did not state specifically what Dama will be working on, but its role in the development of Intel’s infrastructure processing units (IPU) – which free up CPU performance by offloading certain functions – offers a clue. Tech Radar Pro asked the company for comment.

Bad news for Intel and AMD?

Although Meta’s ambitions to develop custom silicon for its upcoming AR glasses are already public knowledge, the new hire is among the first indications that the company intends to develop chips to power its data centers as well.

Given that Meta recently signed a long-term agreement with AMD, it is unlikely that the intention is to replace a third-party data center processors but rather to develop specialized silicon to enable specific use cases.

This decision can be seen as the continuation of an emerging trend that has seen several cloud Vendors have been designing their own ARM-based silicon over the past few years. In 2018, AWS unveiled its Graviton line (now in its third generation), which turned out to be a huge success, and Alibaba Cloud is also working on a powerful new line of custom chips. recent activity suggests Microsoft has similar plans to support its own cloud computing business.

The primary incentive for large-scale data center operators to develop their own custom silicon is the ability to more closely match the hardware to the requirements of their specific workloads.

In conversation with Tech Radar Pro earlier this year, AWS HPC specialist Brendan Bouffler explained that the Graviton series has given the company far greater freedom to optimize its operations that Intel and AMD simply cannot provide.

“Generally speaking, removing unnecessary functionality means you have much tighter control over the failure profile, and silicon control gave us a similar advantage,” he said.

“Graviton3 is optimized for our data centers, because we are the only customer for these things. We know our terms, while other manufacturers have to support the weirdest and most unusual data center setups.

Intel, for its part, says it is out of step with the emergence of powerful custom silicon for the data center. In fact, the company sees an opportunity here for Intel Foundry Services, its new pure-play foundry arm.

Speaking at Intel Vision in Dallas last week, Data Center and AI Manager Sandra Rivera said the following:

“Hyperscalers are among our biggest volume drivers, and they rely on us to deliver sustainable, sustainable capacity to their data centers. For their large-scale deployments, they still rely on Xeon.”

“But they also want sustainable innovation. We can deliver some of that with Xeon, and there’s also the opportunity with our foundry business to enable hyperscalers to innovate on a single IP.”

Going through The register

Leave a Comment