Edge AI is one of the biggest trends in chip technology. These are chips that run AI processing on the edge — or, in other words, on a device without a cloud connection. Apple recently bought a company that specializes in it, Google’s Coral initiative is meant to make it easier, and chipmaker ARM has already been working on it for years. Now, ARM is expanding its efforts in the field with two new chip designs: the Arm Cortex-M55 and the Ethos-U55, a neural processing unit meant to pair with the Cortex-M55 for more demanding use cases.
The benefits of edge AI are clear: running AI processing on a device itself, instead of in a remote server, offers big benefits to privacy and speed when it comes to handling these requests. Like ARM’s other chips, the new designs won’t be manufactured by ARM; rather, they serve as blueprints for a wide variety of partners to use as a foundation for their own hardware.
But what makes ARM’s new chip designs particularly interesting is that they’re not really meant for phones and tablets. Instead, ARM intends for the chips to be used to develop new Internet of Things devices, bringing AI processing to more devices that otherwise wouldn’t have those capabilities. One use case ARM imagines is a 360-degree camera in a walking stick that can identify obstacles, or new train sensors that can locally identify problems and avoid delays.
As for the specifics, the Arm Cortex-M55 is the latest model in ARM’s Cortex-M line of processors, which the company says offers up to a 15x improvement in machine learning performance and a 5x improvement in digital signal processing performance compared to previous Cortex-M generations.
For truly demanding edge AI tasks, the Cortex-M55 (or older Cortex-M processors) can be combined with the Ethos-U55 NPU, which takes things a step further. It can offer another 32x improvement in machine learning processing compared to the base Cortex-M55, for a total of 480x better processing than previous generations of Cortex-M chips.
While those are impressive numbers, ARM says that the improvement in data throughput here will make a big difference in what edge AI platforms can do. Current Cortex-M platforms can handle basic tasks like keyword or vibration detection. The M55’s improvements let it work with more advanced things like object recognition. And the full power of a Cortex-M chip combined with the Ethos-U55 promises even more functionality, with the potential for local gesture and speech recognition.
All of these advances will take some time to roll out. While ARM is announcing the designs today and releasing documentation, it doesn’t expect actual silicon to arrive until early 2021 at the earliest.