Inside Arm CSS for Mobile 2: The AI-native compute platform for mobile
Arm unveils CSS for Mobile 2, an AI-native compute platform integrating new CPU and GPU engines to support agentic AI workflows and neural graphics on mobile devices.
Agentic AI transforms mobile interactions by converting user intent into actions through context gathering, planning, and coordination across applications and services. This shift demands a compute platform optimized as a complete system, where speed, data movement efficiency, and cross-engine coordination determine performance. Arm’s CSS for Mobile 2 addresses these needs with two new compute engines: the Arm C2 Ultra CPU featuring Arm Scalable Matrix Extension 2 (SME2) for responsive on-device AI, and the Mali G2-Ultra NX GPU with dedicated neural accelerators for native neural graphics execution.
The platform integrates CPU, GPU, system IP, software, and developer tools into a unified system designed for sustained performance within mobile power and thermal limits. The Arm C2 CPU cluster combines C2-Ultra and C2-Pro CPUs with two SME2 units, doubling SME2 capabilities compared to the previous generation. This configuration accelerates single-thread and multi-thread performance, reducing latency in agentic workflows such as speech processing, memory retrieval, reasoning, and application execution.
Mali G2-Ultra NX introduces neural accelerators tightly integrated into shader cores, supported by a new execution engine and third-generation ray tracing unit to handle demanding neural graphics workloads. The platform’s SI L2 system interconnect enhances data movement with lower-latency access, higher bandwidth, and coherency features, ensuring responsive execution across concurrent AI and graphics tasks. Physical implementation tools allow partners to optimize performance, power, and area before SoC integration.
Arm CSS for Mobile 2 is backed by a software ecosystem including Arm KleidiAI, framework integrations, and the Arm AI Portal, which provides optimized models, performance data, and deployment resources. The platform connects hardware capabilities with developer tools, enabling over 22 million Arm developers to leverage SME2-based AI acceleration on C2 CPUs and neural acceleration on Mali G2-Ultra NX for building on-device AI experiences.