GlobalFoundries子公司MIPS发布三款面向物理AI嵌入式设备的RISC-V开发平台

MIPS said on Sept. 4 that it has launched three development platforms designed to enable physical artificial intelligence (AI) in embedded devices.
Physical AI refers to AI technology that senses and interprets the physical world and takes actions directly. Applications include robotic arms, autonomous driving (AD) vehicles and autonomous factory machinery. Embedded systems are small computer control systems installed in products such as home appliances and automobiles to perform specific functions.
The new products are MIPS Acies, Actus and Aegis. Each is a dedicated development platform tailored to specific types of physical AI devices. Their computing capabilities are optimized for different workloads, including AI inference, real-time control and safety-critical systems. MIPS calls the approach "workload-native."
The platforms focus on delivering class-leading energy efficiency. That means achieving high tokens-per-watt, a measure of how much AI computational output, or tokens, a system can generate for each watt (W) of power consumed.
Acies is a developer platform for AI inference workloads at the edge. It features an open, standards-based neural processing unit (NPU), an open software stack and an easy-to-adopt form factor. Edge computing processes data on industrial, commercial and other on-site devices rather than sending it to cloud servers.
Actus is a real-time development platform for embedded systems that need to respond quickly to external signals and changes, including motion control, sensor fusion and predictive maintenance. Sensor fusion combines signals from multiple sensors, while predictive maintenance detects potential equipment and facility failures in advance.
Aegis is a development platform for mission-critical physical AI systems where failure-free operation is essential. It is suited to physical AI systems directly related to life and safety, including autonomous vehicles and medical applications. It supports low latency and high levels of stability and reliability.
"As AI expands into machines, vehicles and other physical systems, compute platforms must evolve to meet the characteristics of each workload," MIPS Chief Executive Officer (CEO) Sameer Wasson said. He said MIPS provides open, software-first platforms that allow customers to develop and optimize physical AI on a common foundation and deploy it in real-world systems.
A software-first platform is a development approach in which hardware architecture is designed and optimized around the software and AI algorithms that will run on it. Traditionally, software was developed and optimized after the hardware was built.
Semiconductor industry partners including AMD and MediaTek also expressed support. Manuel Uhm, director of silicon marketing for AMD Embedded, said the growth of RISC-V is driven by industry collaboration and expansion of the software ecosystem, adding that he welcomed efforts to broaden developer access and encourage the use of open-standard designs.
MIPS is a subsidiary of semiconductor foundry company GlobalFoundries (GF). Using the open-standard RISC-V architecture, it offers a portfolio integrating AI and processor intellectual property (IP), workload-optimized analysis, or workload modeling, and solutions for physical AI platforms.