Steve Furber: developing ARM with no people and no money
Steve Furber recalls the birth of the ARM processor at Acorn, despite the lack of money and personnel.
Well I mean ARM was a completely speculative game, we were looking for a processor to succeed the 6502 from our mainstream products and Sophie had been and, well we’d actually brought in 6800s and 32016s, played with them and we didn’t like them. And Sophie had started doodling with instruction sets and then Hermann brought us some papers from – on Berkeley RISC and so on. And gradually the project went from a sort of random investigation onto an official footing. But that was very much driven by us, with Hermann’s backing, okay, so it was a very strange thing for the company to do. It was fuelled by the fact that the company had been extremely successful with the BBC Micro and related stuff and things we did tended to work. And the company had sort of grown a kind of self confidence and possibly an arrogance about its ability to do technology which meant that this apparently crazy thing was just about conceivable as an activity. You know, we’d visited Nat Semi in Haifa as I mentioned and they were on Rev H of the 32016, so they’d gone through about a dozen iterations and it still wasn’t quite working right and they had teams, you know, teams of I don’t know 100 people working on this. So these were very expensive developments and Acorn was still a small company, I mean it was three or 400 people but you still couldn’t find a team of 100 people just to go off and do speculative processor design. But the Berkeley ideas were key, so the RISC idea of actually going complicated is the wrong way to go, that trying to keep it simple is a better idea and using the silicon resource differently, meant that there was a chance this made sense. We actually thought through the ARM development we thought, well this RISC idea is so obviously right that, you know, all the big boys will pick up on it fairly soon and they’ll just trample us under foot but at least by that point we’ll then know quite a lot about it and then be better placed to choose which one of them we buy. And I think nobody was more surprised than us when eighteen months later we had a working RISC processor and mainstream industry was still sort of eschewing the idea of RISC, they still thought RISC was a bad idea. Because we didn’t have the resources, you know, standard microprocessor designs absorbed hundreds of man years of work and there’s no way we could afford that. We had to do it in a tiny fraction of the resources that most companies used to solve a similar problem and you know in the end it was eighteen months, I reckon it was probably about a dozen man years went into developing the first ARM design. So we did deliver with much less resource and the RISC idea was key to that. But also the, you know, the quality of the people doing the design and Hermann’s argument we had these two big advantages over the competition, one was no people and the other was no money and these are kind of true, that this really forced us to look for simple solutions and simplicity is really the core of what we got right.
- Interviewee: Steve Furber
- Duration: 00:03:26
- Copyright: British Library Board
- Interviewer: Thomas Lean
- Date of interview: 10/22/2012
- Shelfmark: C1379/78