Arm has unveiled its first computing subsystems for consumer devices, simplifying the development of processors for desktops, laptops and tablets running Android, Windows and Linux.
Building on its existing Total Compute solutions, Arm’s new CSS for Client platform includes the latest Armv9.2 Cortex CPU cluster (featuring the powerful Cortex-X925 CPU, efficient Cortex-A725 CPU and updated Cortex-A520 CPU), Arm Immortalis and Mali GPUs, and CoreLink Interconnect system IP.
This combination is said to provide advanced performance for AI and other compute-intensive workloads. The CoreLink Interconnect system reduces DRAM bandwidth usage and improves security through advanced memory management, while the Immortalis-G925 GPU, built on the 5th generation architecture, delivers high performance and energy efficiency.
Significantly faster inference times
The Cortex-X925 CPU and Immortalis-G925 GPU are said to deliver significantly faster inference times, enabling seamless user experiences for a range of AI applications. Arm says the new CSS for Client achieves a 42% faster time-to-first-token when running LLMs locally.
CSS for Client is also Arm’s fastest platform for Android, with significant improvements in key benchmarks and common compute use cases compared to the TCS23 platform. These include a 36% improvement in peak performance, 33% faster application startup times, 60% faster web browsing and a 30% increase in maximum graphics performance. These improvements support demanding tasks such as ray tracing and variable-speed shading.
CSS for Client is scalable across consumer devices and targets high-end consumer technology, including – no surprise here – next-generation AI PCs.
“We are now delivering physical implementations for Arm, CPU and GPU, making it easier to build and deploy Arm-based solutions and leaving nothing to chance, enabling new performance points and compute capabilities and increasing time to market to accelerate,” said Chris Bergey, SVP and GM of Arm’s Client Line of Business.