Arm seeks to deliver AI-optimized CPU/GPU designs and ready-to-manufacture blueprints, aiming to improve smartphone AI adoption and product development.
Today, Arm Holdings plc, a British semiconductor and software design company based in Cambridge, has released two new designs: the Cortex-X925 CPU and Immortalis-G925 GPU. These will provide power efficiency, performance, and security for Android, Linux, and Windows devices, mainly targeting the 3-nanometer process nodes.
Arm’s Compute Subsystems for Client integrates CPUs and GPUs, utilising the latest Arm architecture, Arm v9.2, for enhanced performance. The Cortex-X925 CPUs boast up to 3 megabytes of L2 cache per core, improving data retrieval speed and overall processing performance. Alongside lower-end core designs, the Cortex-A725 and Cortex-A520 trade processing speed for reduced power usage, catering to varying performance needs.
Senior Vice President Christopher Bergey states that using Arm’s prepackaged core blueprints streamlines development processes, though customizing processors still requires significant effort.
Arm refreshes its top-end CPU core design annually, with the Cortex-X925 boasting the largest speed increase in the product line’s history. Its single-threaded application performance improves by 36%, while AI performance sees a 25% boost.
Arm’s flagship smartphone GPU, the Immortalis-G925, debuts alongside the new CPUs. This GPU promises 37% faster mobile game performance and a 34% improvement in AI model processing.
This GPU, along with the Mali-G725 and Mali-G625, caters to varying device needs, from high-end smartphones to entry-level handsets and smartwatches.
The Cortex-X925 CPU and Immortalis-G925 GPU are manufactured using the three-nanometer process, contributing to their performance prowess.
Additionally, Arm introduces a feature dubbed fragment prepass, which enhances GPU efficiency by reducing the need for data sorting computations.
Arm’s latest offerings signal a pivotal moment for the mobile industry. They provide chip designers with frameworks for AI-enabled mobile devices.
This initiative aims to streamline system development and accelerate time to market, solidifying Arm’s position as the foundation for future computing performance demands.
