Last year, Arm introduced Lumex Compute Subsystem (CSS) – a turnkey solution for chipset designs. Today, the company is unveiling CSS for Mobile 2, an AI-native compute platform for smartphones.
It is built around the Arm Mali G2-Ultra NX GPU, which has built-in AI acceleration hardware and an Arm C2 CPU cluster including Ultra and Pro cores with SME2 (Scalable Matrix Extension) for agentic AI tasks.
The Arm Mali G2-Ultra NX is the first Mali GPU with dedicated AI acceleration. This enables 4x higher performance per watt for neural graphics – that is graphics that are enhanced, reconstructed or even fully generated by Arm Neural Technology.
The Arm C2-Ultra CPU cores are the company’s fastest cores – they can deliver 15% higher single-thread performance than the C1-Ultra or 38% power savings at the same performance level. Meanwhile the C2-Pro cores focus on efficiency. Two SME2 units enable low-latency AI compute on the CPU itself – they bring twice the compute power of the C1 generation design, which results in a 70% speedup on Small Language Models.
SME2 has been adopted by Android and iOS devices, with work done by Alipay, Google AI Edge Gallery, Oppo and vivo.
So these CPU cores are still part of ARMv9.3-A architecture? Also, I wish they iterated on C1-Nano and C1-Premium too, it's always weird how ARM refreshes enhanced version of previous gen CPU cores.
There is no premium core this year, and the C2 Pro is just a renamed C1 Pro. That's why Xiaomi picks C1 for its O3...
Tip us
2.1m 180k
RSS
EV
Merch
Log in I forgot my password Sign up