Meta has partnered with Arm in a multi-year deal that will see the social media giant use Arm-based CPUs in an effort to improve and scale up its AI systems.
Per the agreement, Meta will use Arm’s Neoverse platform to power its AI-based search ranking and recommendation systems and work with the British chip designer to optimize its existing AI infrastructure software stack.
How Arm is building infrastructure for a systems-level world
Chip design giant has gone from playing catch-up to being a market-leading triple threat in 15 years
In a statement, Arm said that by running its systems on Neoverse, Meta will be able to achieve “performance-per-watt parity,” with the platform supporting the delivery of higher performance and lower power consumption when compared to x86 systems.
The partners said they also worked closely to optimize Meta’s foundational AI software technologies, including Facebook GEneral Matrix Multiplication (FBGEMM) and PyTorch’s ExecuTorch Edge-inference runtime engine, in order to produce “measurable gains in inference efficiency and throughput.”
These software improvements have been made available to the open source community, allowing others to make similar changes to their software stacks.
“From the experiences on our platforms to the devices we build, AI is transforming how people connect and create,” said Santosh Janardhan, head of infrastructure, Meta. “Partnering with Arm enables us to efficiently scale that innovation to the more than three billion people who use Meta’s apps and technologies.”
Rene Haas, CEO, Arm, added: “AI’s next era will be defined by delivering efficiency at scale. Partnering with Meta, we’re uniting Arm’s performance-per-watt leadership with Meta’s AI innovation to bring smarter, more efficient intelligence everywhere - from milliwatts to megawatts.”
Arm first launched its Neoverse-based CPUs in 2018, and divides the offering into three separate groups: The V series, which focuses on high-performance, general-purpose compute; the N series, which targets the server market; and the E series for Edge compute.
In 2023, the chip designer introduced Neoverse CSS to simplify and accelerate the adoption of Arm Neoverse-based technology into new compute solutions by enabling its partners to build specialized silicon more affordably and quickly than previous discrete IP solutions.
Comments