N4-Edge neuromorphic processor

AI for devices that run on a battery.

N4-Edge is designed to do work only when the signal changes, so detection can run inside the sensor itself.

Smoke rising from a wildfire above a forested hillside

The problem

Remote analysis is not always an option.

A sensor on a hillside, a pipeline or a drone has a small battery and a connection that comes and goes. Sending every reading away to be analysed spends power it does not have.

How it works

Sensor events in, a result out.

N4-Edge die with the compute tiles highlighted

N4-Edge die with the working memory highlighted

N4-Edge die with the management highlighted

N4-Edge die with the host interface highlighted

Architectural visualisation

Why it uses less energy

Most chips never stop. Ours only wakes for the signal.

Conventional processor0

N4-Edge0

What we have built

Built, and run on hardware.

4processor generations designed and validated on FPGA

90.28%accuracy on the Spiking Heidelberg Digits test, run on the hardware

2earlier reference designs released as open source

See the research

Applications

For places a wire will never reach.

Defence and counter-droneDetection on the platform, where power and communications are limited.

Environmental sensingWildfire, gas and equipment monitoring without constant site visits.

Space and industryOn-board analysis and remote condition monitoring.

Programmes

Converge Net Zero finalist

ACCEND cohort two

Techscaler member

What do you need to detect?

If a device has to listen or watch all the time on a battery, that is the conversation worth having.

[email protected]