BrainChip Pledges to Showcase "Revolutionary" Neuromorphic Processor at the Linley Fall Conference

Claimed to offer unbeatable performance through spiking neural network technology, can BrainChip's SoC bring AI to the edge?

Semiconductor startup BrainChip has announced it is to officially unveil its Akida neuromorphic system-on-chip processor at the Linley Fall Processor Conference later this month, promising a "revolutionary new breed of neuromorphic computing devices" for everything from the industrial Internet of Things (IIoT) to augmented and virtual reality (AR/VR).

Unlike traditional processors, neuromorphic processors are designed to mimic the operation of the human brain. While this means they're not that good at running Doom, they offer a range of advantages for accelerating neural networks at the heart of everything from machine vision to natural language processing. Based on the concept of a spiking neural network, BrainChip's Akida SoC is claimed to draw power only when actively processing spikes — meaning they're energy efficient enough to do the processing at the edge, right on an embedded device.

"Spiking neural networks have many attractive characteristics, such as rapid learning capabilities, and low compute and memory overhead," explains BrainChip's chief development officer Anil Mankar, who will be presenting the Akida SoC at the event. "I am excited to have the opportunity to share with attendees of the Linley Fall Processor Conference how Akida’s capabilities open up hundreds of new possibilities for applications operating at the edge."

BrainChip claims its Akida processor dramatically reduces the power, memory, and host CPU requirements of running a range of spiking neural network workloads on an embedded device, highlighting the surveillance, advanced driver assistance systems (ADAS), autonomous vehicles (AV), vision guided robotics, drone, augmented and virtual reality (AR/VR), acoustic analysis, and industrial IoT markets as being particularly suitable for Akida deployment.

BrainChip isn't the only company looking at spiking neural network technology, however: Back in July Intel announced Pohoiki Beach, a 64-chip supercomputer based on the company's earlier Loihi neuromorphic processor design. At the time, the company claimed it was around 1,000 times faster and 10,000 times more efficient at running selected workloads than the company's own traditional CPUs. Intel's Loihi, however, has yet to progress past "research chip" status — giving BrainChip the chance to hit the ground running.

BrainChip is currently operating in semi-stealth mode, though in addition to formally launching the Akida design at the Linley Fall Processor Conference has announced a workshop using the chip to take place in Perth early next month. Full details are available on the registration page.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles