Maria Montero

SpiNNaker, the million-core supercomputer, finally powered up

After 12 years in the making, the “brain computer” designed at the University of Manchester is finally turned on. What does this computer do? How is it done? And who is Steve Furber?

AI systems have developed rapidly in the last decade with the use of deep learning, neural networks, and large computers to test and simulate neurons. But AI is not the only area of ​​interest when using such techniques; Scientists and engineers are also interested in trying to simulate the human brain to better understand how it works and why.

Simulating the brain is not a trivial task. The complexity of the human brain is difficult to replicate, which is part of why the SpiNNaker computer is important.

The challenges of simulating a brain

One of the first fundamental differences between the brain and computers is how their “smallest units” work. Brain neurons can have multiple connections and react to impulses in different ways. Computer transistors, by comparison, are switches that, while they can be connected to other transistors, can only have one of two states.

Neurons can also forge links between other neurons and react to stimuli differently (which is a definition of “learning”), whereas transistor connections are fixed.

Because of these differences, scientists have to “simulate” neurons and connections in software rather than hardware, severely affecting the number of neurons and links that can be simulated simultaneously.

What about simulation neurons in hardware?

Neurons and transistors have little in common, but a better comparison would be simple microcontrollers and FPGAs; Microcontrollers are similar to neurons in that they can process external signals quickly while being comparatively simple in architecture, while FPGAs provide the ability to break and create connections between microcontrollers.

Could hardware simulation be the key? A team of researchers believe so and have spent the last 12 years on the idea.

The SpiNNaker

A research team at the University of Manchester has spent the last 12 years creating a computer that will simulate neurons and connections with the use of many simple cores interconnected in a massive parallel system and the computer, called SpiNNaker, was finally turned on.

The million-core computer is designed to simulate up to 1 billion neurons in real time to allow scientists to study neural networks and pathways in a realistic way using hardware rather than software.

Unlike traditional methods for simulating neurons, SpiNNaker has individual processors that simulate up to 1000 neurons that transmit and receive small packets of data to and from many other neurons simultaneously.

Hexagonal topology between processors and a 48-processor SpiNNaker computer – Image courtesy of University of Manchester

The Spiking Neural Network Architecture (SpiNNaker) system consists of 10 19-inch computer racks with each rack containing 100,000 ARM cores. This core density is achieved with the use of a custom IC containing up to 18 cores. Each board in a rack has 48 chips, resulting in each board containing 864 processors.

Unlike typical software systems, cores are arranged in a hexagonal pattern with data transmission handled entirely in hardware. It is this topology that allows the system to simulate one billion neurons in real time. The system uses ARM9 processors that contain a total of 7TB of RAM and 57K nodes, while each processor has 128MB out-of-array SDRAM memory and each core has 32KB memory and 64KB DTCM memory with tightly coupled data.