Arm adds IP and processors to its to AI offering
For on-device intelligence for IoT applications, Arm has announced new IP and a neural processor to boost machine learning performance.
In a single announcement, the company introduced machine learning (ML) IP, the Arm Cortex-M55 processor and Arm Ethos-U55 neural processing unit (NPU).
The IP and supporting unified toolchain enable AI hardware and software developers what Arm describes as “unprecedented levels of on-device ML processing” for small, power-constrained IoT and embedded devices.
Explaining the need for AI, Dipti Vachani, senior vice president and general manager, Automotive and IoT Line of Business, Arm, said: “Enabling AI everywhere requires device makers and developers to deliver machine learning locally on billions, and ultimately trillions of devices”.
As the IoT intersects with AI advances and the rollout of 5G, more on-device intelligence means that smaller, cost-sensitive devices can be smarter and more capable while relying less on the cloud or internet for greater privacy and reliability.
Arm explains that this intelligence on microcontrollers designed securely from the ground up, reduces silicon and development costs and speeds up time to market for developers.
The Arm Cortex-M55 is described as Arm’s most AI-capable Cortex-M processor. It is the first to be based on the Armv8.1-M architecture with Arm Helium vector processing technology. Cortex-M55 delivers up to a 15-fold uplift in ML performance and a five-fold uplift in DSP performance, with greater efficiency, compared to previous Cortex-M generations, reports Arm.
Additionally, Arm Custom Instructions will be available to extend processor capabilities for specific workload optimisation.
The Ethos-U55 is Arm’s first microNPU (neural processing unit) for Cortex-M. It is designed to deliver a combined 480x leap in ML performance to microcontrollers.
Paired with the Cortex-M55, it can deliver a combined 480-fold increase in ML performance over existing Cortex-M processors.
The Ethos-U55 is highly configurable and specifically designed to accelerate ML inference in area-constrained embedded and IoT devices. Compression techniques save power and reduce ML model sizes significantly to enable execution of neural networks that previously only ran on larger systems.
The Cortex-M55 and Ethos-U55 are supported by Arm’s Cortex-M software tool chain. This ensures a unified development flow for traditional, DSP and ML workloads. There are specific integration and optimisations for leading machine learning frameworks (e.g. TensorFlow Lite Micro).
The processors and the accompanying Corstone reference design work with Arm TrustZone to ensure security can be more easily incorporated into the complete SoC.
Ecosystem partners, include Amazon, Alif Semiconductor, Bestechnic, Cypress, Dolby, Google, NXP, Samsung and STMicroelectronics.
