Champagne and chips: Arm found the perfect recipe
Producing a new microprocessor from scratch at Acorn Computers seemed a mad project doomed to failure, writes Hermann Hauser, co-founder of Acorn Computers and Arm. How could a medium sized British home computer company take on mighty Intel in the microprocessor field?
Well, Acorn had a lot of confidence in its design capability and we had examined all the different processor designs of the day (Intel, Motorola, National Semiconductor and new 16-bit versions of the 6502 which we used for the BBC Micro).
We asked Intel to modify their 80286 so that we could use it in a more efficient way but they politely told us to get lost. If they had not, maybe we would have never embarked on the Arm project.
After a visit at Western Digital by Steve Furber and Sophie Wilson we realised that if they could produce a microprocessor then so could we. So, we started the Arm project with two major advantages over all other microprocessor teams in the world:-
- No money. (We didn’t have any).
- No people (well two for the original architecture and instruction set). Intel etc typically had teams of 50 people working on their next processor.
We were greatly helped by an architectural breakthrough called RISC (Reduced Instruction Set) which was spotted by Andy Hopper who gave us the papers written by John Hennessy at Stanford and David Patterson at Berkley.
This led to the unusual feat of a Silicon Valley invention being first implemented on a working chip in Cambridge UK. Hennessy later created MIPS and became president of Stanford.
On the 25th April 1985 the first ARM chip was delivered to Acorn and I had bought two bottles of champagne to celebrate – being the perennial optimist.
We plugged the chip in and it did not work! However, using his green fingers it only took an hour or two for Steve Furber to find the circuit board errors and a completely new microprocessor designed using BBC Basic sprang to life with a “Hello world I am an ARM” message on the screen.
In retrospect this seems totally unbelievable as it meant the silicon chip worked the first time, the circuit board worked after minor tweaks, the software that was written blind, not having a processor to test it on, also worked first time. Now we could have our champagne. I still have one of the bottles at home.
After this rather pleasing event, Steve Furber wanted to measure the power consumption by connecting an ammeter to the power pin – only to discover that it was not connected!
This was an early indication that the design was rather low power, but it surprised us that it would not consume any power at all.
The explanation lay in the fact that the design was indeed so low power that the leakage current from the other connections to the chip was enough to make it work.
At that time, we had no idea that first Nokia phones and then 95 per cent of all phones in the world would appreciate this low power feature for their battery-operated devices – a feat that Intel never managed to achieve.
More than 180 billion Arm chips later we can rightly be proud of a British success story which I very much hope will not be sold to NVIDIA to become yet another US monopoly.