AI Business is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 3099067.

Arm’s latest IoT processor supercharges machine learning performance

by Max Smolaks
Article Image

To build AI into all the things

by Max Smolaks 12 February 2020

Machine learning applications running on a broad range of consumer devices are about to get a massive boost: silicon designer ARM announced that its upcoming Cortex-M55 processor design equipped with the Ethos-U55 neural processing unit (NPU) will deliver up to 480 times more performance in ML tasks than its predecessor, Cortex-M35P.

The new silicon is aimed at IoT and embedded devices like smartwatches, smart speakers, and other smart paraphernalia; Arm’s blog post on the subject lists a smart cane for visually impaired people as one of the potential use cases.

The
Ethos-U55 NPU is among the smallest devices in its class and was
optimized around minimal power consumption. According to Arm, it will
enable AI on billions of next-generation smaller, power-constrained
devices.

Endpoint
AI

ARM is the heavyweight champion of the world in thinking up new silicon for mobile devices and the IoT; the company doesn’t make any chips, but its designs are powering smartphones from Apple, Samsung and Huawei, and that’s just the tip of the iceberg. More than 50 billion Cortex-M chips have been shipped to customers.

The latest entry in the series is Arm’s most AI-capable Cortex-M yet. It is the first to be based on the Armv8.1-M architecture with Helium vector processing – which is especially useful for enabling functions like vibration and motion, voice and sound, and vision and image processing.

On
its own, M55 delivers an ML performance boost of up to 15x; when
combined with the Ethos-U55, Arm says it performs 480 times faster
than any other existing Cortex-M.

The
secret to this performance hike is
not due to silicon alone,
but also developer software
optimized for ML performance across all configurations of the new
processors.

Arm Custom Instructions will be available to extend M55 capabilities for specific workload optimization, and Google’s TensorFlow Lite for Microcontrollers will be supported out of the gate.

“Enabling
AI everywhere requires device makers and developers to deliver
machine learning locally on billions, and ultimately trillions of
devices,” said Dipti Vachani, senior vice president and general
manager, Automotive and IoT Line of Business, Arm.

“With
these additions to our AI platform, no device is left behind as
on-device ML on the tiniest devices will be the new normal,
unleashing the potential of AI securely across a vast range of
life-changing applications.”

Practitioner Portal - for AI practitioners

Story

AI and self-service business intelligence – competing or complementing concepts?

7/8/2020

One term – data analytics – having two meanings – AI and SSBI – this is the classic setup for misunderstandings, failing project pitches, and failed projects. But what exactly are the differences between AI and SSBI? And are they complementing or competing concepts?

Story

Open source platform aims to speed up autonomous car development

7/6/2020

Project ASLAN promises easy to install, fully documented and stable self-driving software for specific low-speed urban autonomous applications

Practitioner Portal

EBooks

More EBooks

Upcoming Webinars

Experts in AI

Partner Perspectives

content from our sponsors

Research Reports

9/30/2019
More Research Reports

Infographics

AI tops the list of most impactful emerging technologies

Infographics archive

Newsletter Sign Up


Sign Up