AI models are rapidly evolving and outpacing hardware capabilities, which creates an opportunity for Arm to innovate across the entire compute stack.
Arm recently announced new chip blueprints and software tools aimed at making smartphones more capable of processing AI tasks. But what's more, Arm is also making changes to how it delivers these blueprints that could speed up adoption.
Arm is evolving its solution offerings to maximize the benefits of key process nodes. The company has announced Arm Compute Subsystems (CSS) for Client, its latest cutting-edge computing solution tailored for AI applications on smartphones and PCs.
This CSS for Client promises significant performance improvements, with over 30% improvement in compute and graphics performance and as much as 59% faster AI inference for AI, machine learning, and computer vision workloads.
Arm's technology has powered the smartphone revolution, but it's also garnering attention in PCs and data centers, where energy efficiency is at a premium. Smartphones remain Arm's largest market, and it supplies IP to rivals including Apple, Qualcomm and MediaTek, but the company is expanding its offerings.
The company announced new CPU designs and new GPUs optimized for AI workloads, as well as software tools that make it easier to develop chatbots and other AI apps on Arm chips.
But the real game changer is how these products are delivered. Historically, Arm has provided specifications or abstract designs that chipmakers had to translate into physical blueprints – a monumental challenge when it comes to putting billions of transistors on the line.
For this latest offering, Arm worked with Samsung and TSMC to provide manufacturing-ready physical chip blueprints, saving significant time.
Samsung's Jung-Wook Kaye praised the partnership, saying that the combination of its 3nm process with Arm's CPU solutions will enable “early and close collaboration” with DTCO in maximizing PPA, ensuring on-time silicon delivery to meet performance and efficiency demands and address the growing demand for generative AI in mobile.
Dan Koch-Pacharin, head of ecosystem and alliance management at TSMC, agreed, calling the AI-optimized CSS a “great example” of Arm and TSMC collaboration that helps designers push the boundaries of semiconductor innovation to deliver unmatched AI performance and efficiency.
“Working with Arm and our Open Innovation Platform® (OIP) ecosystem partners, we are helping customers accelerate their AI innovation using cutting-edge process technologies and design solutions,” Koch-Pacharin emphasized.
Arm is not trying to compete with its customers, but rather aims to accelerate time to market by providing optimized designs for neural processors that deliver cutting-edge AI performance.
“We're putting together a platform that allows us to very tightly couple these accelerators to our customers' NPUs,” Arm's Chris Bergey said.
In essence, Arm offers more sophisticated, “baked” designs that customers can integrate with their own accelerators to rapidly develop powerful AI-driven chips and devices.
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London – this comprehensive event will take place alongside other major events such as Intelligent Automation Conference, BlockX, Digital Transformation Week and Cyber Security & Cloud Expo.
Find out about upcoming enterprise technology events and webinars hosted by TechForge here.
