Anmou Technology WAIC update: AIoT combo punch, NPU amplification move, spoilers for the next generation VPU, and taking the lead in launching the open source AIOS alliance

Zhidongzhi Author | Yunpeng Editor | Xinyuan Zhidongzhi reported live at Shanghai WAIC on July 18. Just now, Anmou Technology, a leading manufacturer in the field of domestic chip IP design and services, officially released two blockbuster products, the "Star 300" AIoT prototype platform and the "Zhouyi" X3-Pro series NPU. At the same time, it revealed for the first time the core technical characteristics of the next-generation "Linglong" VPU "Wudang". At the press conference, ARM Technology took the lead in launching the "Open Source AIOS Alliance". Xinfa Technology, Unisoc, Huixi Intelligent, Rockchip, Wallface Intelligence, RT-Thread, AutoCore, Tsinghua University Intelligent Industry Research Institute, and the Hong Kong University of Science and Technology Physics AI Research Center have all joined this alliance. Last year, ARM Technology proposed the All in AI strategy for the first time. ARM Technology's technical support can be seen in a series of cloud-edge solutions demonstrated on site, from edge, cloud to physical AI. 1. Aiming at the direction of AI for all things, showing off the heterogeneous AIoT computing power platform, and hitting the CPU+NPU+Helium combination punch. Currently, the miniaturization and intelligence level of models are constantly improving, and small models are accelerating to approach commercial standards, promoting the large-scale implementation of end-side AI. This has also accelerated the evolution of IoT, which used to be centered on "connection", to AIoT, which is marked by "cognition". According to ARM Technology, the focus of sensory data is on the edge-end side, real-time and autonomy, privacy and data sovereignty, and low-power lightweight chip technology. These are the four driving forces that accelerate the transformation of AIoT. Zhu Xiaoming, product manager of Amou Technology's "Star" CPU product line, mentioned that AI deployment in embedded devices always faces multiple resource constraints such as power consumption, latency, and storage, which puts forward higher requirements for the underlying design of the chip. CPU instruction extension, general small computing power NPU, and SRAM in-memory computing are three technical paths. Currently, for the field of embedded computing, Arm has built a complete IP solution covering Cortex-M CPU processor, Helium vector extension technology and Ethos NPU dedicated AI accelerator. Arm Helium is Corte