Inventor of the ARM chip Stephen Furber

27 June 2010 | By Stuart Nathan

Readers of a certain age will have fond memories of the BBC Micro. Purchased in bulk by UK schools in the 1980s from Acorn, it gave many their first taste of computing and set plenty of programmers on their career path. Many credit this machine for the UK’s enviable software industry.

One of the main architects of the BBC Micro, Stephen Furber went on to develop the ARM microprocessor – the first commercial chip based on the principles of Reduced Instruction Set Computing (RISC). Now, the ARM processor is at the heart of virtually every example of mobile computing. Over the past 25 years, almost 20 billion ARM chips have been produced; they operate 98 per cent of mobile phones and over a quarter of electronic devices.

Furber, who was nominated for the 2010 Millennium Prize, had an unconventional path to the computer industry.

’I started off with maths and did my PhD in aerodynamics, from the theoretical side,’ he said. ’The theoretical work suggested some experiments and, to do those, I needed data logging. Concurrently with that, I was interested in flight simulation and, as I was thinking about how to build a flight simulator, I became aware of the Cambridge Processor Group – a student society that built computers for fun. I started working with them and the things I built merged with my aerodynamics work.’

At that point the founders of Acorn Computing, Hermann Hauser and Chris Curry, were looking for prospective staff. ’Hermann, who was at Cambridge at the time, came along to the CUPG and asked me if I would like to get involved with Acorn. I did some part-time work for them and, when my fellowship expired, joining Acorn seemed to be the best option for me.’

After working on the BBC Micro, the development process for Acorn’s next machine took an unexpected turn. The team decided to shift from 8-bit to 16-bit architecture for the new computer’s CPU, but all of the commercial chips were ’deeply unimpressive’, Furber said. ’They were based on 1970s minicomputers that had evolved to do relatively complex things in one instruction, which made them slow. Also, they couldn’t move data into and out of the memory as fast as the memory chips themselves were capable of working.’

The development team was disconsolate until Hauser drew its attention to some research papers from Stanford and the University of California at Berkeley, which looked at a new kind of chip based on RISC. ’The idea was to do simple things very efficiently,’ Furber said. ’That meant you’d get good real-time response because you won’t have complex instructions that block the processor out and, if it’s simple, it should be able to keep up with the memory.’

The problem was that chip design was seen as an expensive and labour-intensive task, and Acorn had relatively little cash and a very small team – none of whom had ever designed a chip before. ’We thought the Berkeley research was so obviously a good idea that the industry would pick it up and run with it, and they’d flatten us. But, for two years, the industry defended its position in complex instruction sets, because of what they perceived as ’the risk of RISC’. And we, quite unexpectedly, ended up with a working RISC processor.’

Furber still believes that, in many cases, big companies’ conservatism holds them back. Companies would rather improve their existing products a little at a time than risk a complete paradigm shift, such as the one represented by ARM. ’Some of that’s justified – it’s hard to move the world a lot at once. It’s much easier to do it in small steps. Evolution is easier than revolution.’

In the UK, Furber thinks that big innovations are more likely to come from small start-ups. ’There are very few companies playing the R&D game and most of the research money is in the universities. But the goal of the academic is to get results you can publish, not make money. The wackier it is, the more likely people are to be interested in the paper. Every so often, you get lucky, and it’s wacky and effective.’

Furber’s current project certainly has the air of wackiness. Called Spinnaker (Spiking Neural Network Architecture), it aims to use ARM chips to model a proportion of the neurons of the human brain (see Q&A panel). Furber, however, is keen to point out that he’s not trying to solve philosophical problems. ’We’re nowhere in sight of finding out what consciousness is; there’s a huge gap before we can even formulate questions on those matters.’

Furber’s happier looking at concrete matters and he’s most proud of the way his processor is changing lives in the developing world. ’In Africa, the mobile phone is the main way of connecting to the internet,’ he said. ’ARM chips are making it available to so many people and it changes their lives for the better.’