Low power chip designer, ARM is putting up to a million of its processors into a new breed of computer that aims to replicate the way the brain works and several million of its dollars into a new Cambridge cleantech venture.
Prof Steve Furber, who co-designed the ARM processor with Sophie Wilson while at Acorn Computers in Cambridge is leading the SpiNNaker (Spiking Neural Network architecture) project – a massively-parallel chip multiprocessor system that mimics how nerve cells in the brain interact.
Meanwhile, ARM has co-led a $7m Series A investment into Amantys, a one year startup developing power control technology that reduces the amount of energy lost in the power conversion process. The startup says it can “address power losses all the way from wind and solar photovoltaic modules, transmission grids and transformers through to electric motors and electric vehicles.”
Amantys, which is staffed by a team of former ARM execs and Dr Patrick Palmer of Cambridge University’s department of engineering, says it is aiming to release its first products by Q4 of this year. The funding round was co-led by Moonray Investors, part of Fidelity International.
Amantys says it is looking to recruit ‘analogue design gurus’, embedded software and power electronics engineers
ARM has quietly assembled an investment portfolio worth $40m and containing 15 companies.
Principal designer of the BBC Microcomputer as well as the ARM 32-bit RISC microprocessor, Prof Furber is now ICL Professor Of Computer Engineering at University of Manchester. He is working with scientists from the universities of Cambridge, Southampton and Sheffield as well as industrial partners – foremost among them, ARM – to develop a massive computer, nicknamed, the ‘brain box.’
By emulating the networks of billions of neurons in the brain using ARM processors, the hope is that scientists will gain a greater understanding of how processing in the brain works – including how damage to the brain interferes with it – but also that these biological models will lead to more efficient and fault-tolerant computers.
Professor Furber said: “Developing and understanding the information processing in the brain is the key. We are actively engaging with neuroscientists and psychologists, both here at the University and elsewhere.
“This could ultimately be of great help for patients, for example, who have presented with reading problems caused by strokes or similar brain injuries. Psychologists have already developed neural networks on which they can reproduce the clinical pathologies. At present they are limited in the fidelity they can achieve with these networks by the available computer power, but we hope that SpiNNaker will raise that bar a lot higher.”
The project has received funding of £5m from EPSRC.
The chips that will power the system – designed in Manchester and manufactured in Taiwan – were delivered from the foundry last month and with 18 ARM processors on board every chip, they will dramatically increase the number of brain cell interactions that can be modeled compared to earlier test systems.
Although there will eventually be up to one million Arm processors in SpiNNaker, making it capable of modelling a billion neurons in real time, this is still only around 1per cent of the human brain.
In the brain, neurons emit spikes which are relayed as tiny electrical signals. Each impulse is modelled in SpiNNaker as a ‘packet’ of data, which is sent to all connected neurons. Neurons are represented by simple equations which are solved in real-time by software running on the Arm processors.
Acorn Computers co-founder, Hermann Hauser, who describes Prof Furber as one of the smartest people he has met, told New Electronics in December 2010 that he was “keeping an entrepreneurial eye” on his latest work. He told the publication: “There is potential in the way the demonstrator works that one can build a computer that can do certain things others cannot. The sort of the things humans are very good at and computers are not.”
Arm was approached in May 2005 to participate in the SpiNNaker project. A subsequent agreement paved the way to make ARM processor IP available to the project, along with ARM cell library IP to aid design and manufacturing.
Mike Muller, CTO at ARM said: “SpiNNaker seeks to create a working model of the ultimate smart system, the human brain. Steve is part of the Arm family, so this project was a perfect way to partner with him and Manchester University, and for ARM to encourage leading research in the UK.”