Arm officially announced on Wednesday, September 9th, the launch of CSS for Mobile 2, a new platform developed to equip the next generation of smartphones. This innovation was conceived by the company to manage complex tasks involving agent-based artificial intelligence and increasingly sophisticated graphics.
The architecture, which integrates CPU and GPU, focuses primarily on executing artificial intelligence directly on mobile devices. Arm states that this new system will deliver the same level of performance with lower power consumption.
A central feature of the platform is the introduction of the new Mali G2-Ultra NX GPU. According to Arm, this is the first Graphics Processing Unit in this line to incorporate dedicated neural accelerators. This project ensures superior performance for graphics that are enhanced, reconstructed, or generated by artificial intelligence.
In terms of conventional rendering, the achieved performance gain is 14% compared to the previous generation. Furthermore, the component features a cutting-edge ray tracing unit, enabling the creation of richer environments and extremely realistic lighting in mobile games.
The industry has already shown interest in the technology. The company reported that developers of graphics engines and studios such as Sumo Digital, Tencent Games, Unity China, and NetEase are preparing to integrate this neural technology into future titles, including Arena Breakout Infinite and Where Winds Meet.
With the advancement of AI, the processor acts as a central coordinator on the platform. Chris Bergey, an Arm executive responsible for this sector, clarified that current technology demands the simultaneous operation of multiple applications and the management of several language models, all while respecting the thermal and performance limits inherent to a smartphone.
To support this high demand, Arm combines C2-Ultra cores, focused on high performance, with C2-Pro cores, which prioritize energy efficiency, in addition to offering low-latency AI computing support directly on the CPU.
The presented improvements are significant on paper. There is a promise of a 70% increase in the processing speed of Small Language Models (SLMs) and 1.7 times better AI performance than the old C1-Ultra. If these figures are confirmed, users could benefit from up to a 15% gain in single-thread performance, accompanied by up to 38% energy savings, as disclosed by Arm.
To facilitate the rapid adoption of this hardware by manufacturers, the company emphasizes that the CSS for Mobile 2 platform already includes native support for essential development kits, such as Arm KleidiAI and the Neural Graphics Development Kit, which have been made available on the new Arm AI Portal.
