20 years ago today - the making of ARM
ARM turns 20 at the end of November. The company nearly didn't make it past its first birthday but now produces the most widely used microprocessor architecture. E&T talks to Sir Robin Saxby and finds out how ARM pulled it off.
Although he left the company several years ago, having spent close to 18 years as chief executive and then chairman, Sir Robin Saxby is still a keen ARM-spotter.
'For years, if there was something with an ARM chip in it, I'd buy it at the airport,' says Sir Robin.
As ARM processors sit inside a dizzying array of consumer products today, as well as cars and even giant neutrino detectors, this could grow into an expensive habit.
'You only need so many devices: what I am looking for today are changes,' Sir Robin explains, holding up his iPad as an example of a device that, based on ARM, is changing the way he views reading. 'This is the replacement for a book or newspaper while travelling. It lets me get information quickly and easily although, the trouble is, when doing work I find I still have to take my PC with me.
'Work is still the PC. For reading, it's this [iPad], and for communication it's the phone. But they all have ARM technology in there somewhere, even the PC.'
Mike Muller, another of ARM's founders and currently chief technology officer at the company, boasts: 'Our ambition is absolutely the entire spectrum of computing from very small to very big. With the exception of the desktop - which I see as a fairly static market - if you skirt around that, everything else is there for ARM to take. It's not particularly modest, but it is the reality.'
Twenty years ago, few would have bet that a tiny Cambridge-based startup would have reached the point where it is now, representing the first-choice processor architecture for anyone not building a PC. In 1990, MIPS and Sparc provided the momentum behind what their respective owners saw as the revolution in processor architecture - reduced instruction set computing (RISC) - that would unseat Intel.
Telecom giant AT&T was spending millions trying to establish its Hobbit processor for portable devices, including the e-reader being developed at the time by Acorn co-founder Hermann Hauser at his startup Active Book. Yet it was the Acorn-developed processor, developed on the cheap, that would ultimately win.
Muller, who was on the original design team together with Steve Furber, Sophie Wilson and others, recalls: 'It's more like 27 years since we started work on the first ARM project: back in 1983. We had been struggling with the performance of conventional processors. They were all large, complex and slow.
'Hermann had been on a world tour and visited a lot of universities and he said they were talking about RISC. Nobody said it was supposed to be difficult,' says Muller, so Acorn decided that it could develop a processor architecture that would suit the computer maker's needs. 'We felt a small team could build an architecture from the ground up.'
The start of ARM
The architecture made its debut in the Archimedes series of computers that launched in 1987. After the launch, 'Acorn was looking at its business and realised that it had an asset in the processor architecture,' Sir Robin explains.
The processor attracted the attention of Apple Computer. 'Larry Tesler was chief scientist at Apple and he was a big fan of the ARM technology,' Sir Robin claims. 'They had been working with AT&T but Larry wasn't particularly happy with the way things were progressing. Acorn was looking to form a joint venture around the ARM processor and Apple, and Larry in particular, were interested in forming one.'
The other company involved was VLSI Technology - later bought by Philips Electronics and folded into what is now NXP Semiconductors - which manufactured the key chips for Acorn's computers.
'I could see from this little glimmer that it could become this global standard,' says Sir Robin. 'The whole idea of a changing landscape in the industry was getting there.
'At the start, most of the engineers thought we would go bankrupt,' Sir Robin recalls. He asked a friend, Jonathan Brooks, to put together a spreadsheet to show how the company might fare financially. 'I bought him a curry. He said 'the company couldn't afford my salary, let alone yours'.'
But Brooks said he would join the company once it was big enough to be able to hire a full-time chief financial officer and he did. A corporate lawyer advised part-time, and ultimately joined.
'The board said we should hire a marketing director. But that was too expensive. I put Mike Muller in charge of marketing and Jamie Urquhart in charges of sales. A lot of the engineers didn't have a lot of commercial experience. I taught them how to do it,' says Sir Robin. 'In December 1990, I got the 12 engineers to do a SWOT analysis: I got the whole company involved. There were other competing architectures out there. ARM was very much the small one. But we realised that the strength lay in MIPS per watt, at a time when everyone else was going after raw MIPS.
'If you look at the original company brochure, what it says is still true today. It has scaled all the way to 2010. That's unusual for startups because they normally don't know what they are going to do at the beginning.
Muller recalls the prevailing view of the time: 'Real men had fabs. Real men made chips.'
That was not the future for ARM, Sir Robin says: 'It was clear that we should not make chips, because we were so short of money.
'In the short term, we had to make chips work for Apple and, EF at the same time, deliver for Acorn,' adds Sir Robin. 'As well as doing things to pay the rent, we had medium-term things to fix. One of the things Apple loved was that the Acorn team had created a very good C compiler. But it only ran on Archimedes, so we had to port that.
Apple demanded changes to the architecture, says Muller, to create the ARM6 processor. 'They wanted us to have a full 32bit addressing range. We had a 32bit architecture but the program counter was only 26bit. That was the first architectural change; the break with the past.'
Finding new customers was a priority if the company was to survive past its meagre £1.75m seed funding. 'If you control your costs you have a longer life. So the idea was the everyone at ARM would have share options. And we didn't bribe anyone to come onboard, we just said we had share options.'
The original engineers had been promised a raise by Acorn. But, close to one year on, there was nothing to pay them with. ARM was running out of cash. 'We had to freeze salaries. But we signed a deal with Plessey in December and the engineers got a backdated cheque.'
As resources were so limited, every deal had to make a difference. Sir Robin remembers having to tell Wilf Corrigan, then chief of VLSI's prime competitor LSI Logic that he would be unable to sell the company a licence, at least for now.
In the early days, selling the concept of a licensable microprocessor was practically impossible, Sir Robin says. 'There was a very small number of people who believed the vision might come off.'
Before he became the president of design-tools company Mentor Graphics in 1993, Wally Rhines was executive vice president of TI's semiconductor group. 'Wally was the first person who said: 'I think this company is going to make it,' recalls Sir Robin.
The reason why companies such as Sharp and TI pushed ARM was because of what the processor could get them into, says Sir Robin.
Muller expands: 'Quite a few of the early customers were what you might call 'CPU-light'. The industry was only just starting to do SoC. It was a time when you had people who sold devices such as discrete disk controllers that would ultimately be integrated into SoCs. They would be locked out if they didn't have a CPU.'
Sir Robin says: 'The other idea that evolved was the voice of the customer into the business model. ARM got its income from semiconductor partners but it was very apparent you needed end customers to drive demand.'
The markets that appeared first were not necessarily the ones that either party expected. The initial push for ARM at TI was in automotive. Sharp America was the driving force behind the Japanese company's decision to license. 'They were interested initially because of the Apple Newton. But it was when we went over to Japan that we identified Nintendo as an opportunity,' Sir Robin says. 'Warren [East, now ARM chief executive] was hired from TI. One of his first projects was Sharp on Nintendo.'
For TI in Europe, what ARM could get them into was Nokia, as the Finnish company began to carve out its dominant position in GSM handsets. Nokia's decision to license the ARM architecture put the company on the map. But it almost didn't happen.
Muller explains: 'TI was taking us in for meetings. Nokia wanted to move away from its 16bit core. It was quite late in the process when they said they had done a code analysis on the processor. We had the best code density for a RISC core but they said it's not as good as our existing designs, which means we will have to use more memory. Sorry, you have lost the socket'.
Thumb a lift
'It was on the way back that the Thumb architecture got dreamed up,' says Muller, referring to a preprocessor tacked onto the ARM7 processor's pipeline that converted a subset of instructions that were just 16bit long instead of 32bit into full ARM instructions. 'It meant that we had even better code density than we had had. It was an innovation that transformed the company, and the one architecture change that you could directly tie to a hardware design win.'
The personal involvement of people in each of the big customers was crucial in the development of ARM, Sir Robin says: 'Every semiconductor partner when they license something has to put a business justification to the board to spend the money - and someone needs to drive that.
'If you are trying to establish an architecture, you need evangelism. We needed ARM disciples. We had people in large corporations willing to bet on this little company. They could have fun at ARM. For a senior person at a big company, the involvement was more engaging. They could roll their sleeves up and get on with things.'
ARM was something of a culture shock for those raised on large-corporation politics. 'In meetings, people tend to be polite to each other. When they came into our meetings they thought we were the rudest people who shouted at each other. But we were just being honest,' says Sir Robin. 'Even though I was chief executive they could tell me I was wrong. It's only by doing things that you learn. You can spend ages talking about it. You learn by your mistakes. If you don't make mistakes, you are not trying hard enough. You want to push the boundaries.'
As well as successes such as TI with Nokia and Sharp with Nintendo, Sir Robin says: 'There were some customers who fell by the wayside. There was 3DO, which was developing a games console. And even before Nokia, VLSI was doing business with Ericsson.'
Gradually, ARM created a community of users. 'People moved on and spread the ARM architecture,' says Sir Robin.
Initial public offering
With licence revenue increasing from a lengthening list of customers and royalties beginning to mount up, the ground was laid for an initial public offering (IPO) in 1998.
'One of the key decisions were made was to dual list: on both the UK and US markets. We had a beauty parade of banks and chose Morgan-Stanley. They said we should list on both as valuations for companies in the US were higher but we could remain a UK entity,' says Sir Robin.
'We did a roadshow and were 30 times oversubscribed. But we hit a big problem. We arrived at a hotel in San Francisco and received a fax from the Securities & Exchange Commission saying our share-option scheme had not been approved, even though it was fine for the London Stock Exchange. We had to reissue our prospectus saying that we would take a non-cash charge: that was a particularly stressful time.'
Despite the attraction of the US markets, ARM remained a UK company. 'There was pressure to move to America,' says Sir Robin. 'But there were disadvantages to becoming an American company. I felt that ARM would be less global if American, and we needed to be a global company.'
As ARM moved into its second decade, a new team began to form. Brooks decided to leave in 2001 - being CFO of a publicly listed company was less interesting to him. Urquhart moved into venture capital and Sir Robin took up the role of chairman, making way for East as chief executive, before retiring in 2007. Of the founders, 'Muller and Tudor Brown remain on the technical staff at ARM.
'To be successful in companies you have to keep rejuvenating the team. There was a great founding team and now the company has got some more great people. In another ten years' time there will probably be another rejuvenation,' says Sir Robin.
'The volumes that ARM partners are going to ship in the next ten are probably going to be bigger than I ever thought they would be. There are nearly seven billion people in the world and ARM touches a lot of those people.'
With what is now a dominant position in embedded microprocessors - and the opportunity to go after Intel in compute servers - is there a danger that ARM will be infected with the hubris associated with the world's largest chipmaker? 'I hope not,' says Sir Robin. 'You have to be humble. The danger is when you are successful, people see you as that and maybe you don't try quite as hard. But they have to keep working on that and push new application areas and push in the unknown direction rather than taking the easy option.'
Partnership is easy to say
Partnership is a word that ARM uses a lot and has become something of a cliché in the chipmaking business where executives talk warmly of embracing partnership in place of traditional customer/supplier relationships. But for ARM, this approach was a necessary way of doing business.
Sir Robin Saxby, ARM's former chairman, says: "ARM created a community. For ARM it was the only way to do it. But it is not for everyone. As a startup company with a niche or as a fabless chip company you may be better off doing a lot of things yourself.
"What ARM did is develop is a true business partnership model. What ARM understands is partnership and risk and opportunity sharing. That is in its lifeblood. We were creating standards that everyone can use and work with and that can only be done with partnership."
As research and development costs increase, Sir Robin says more companies should look at the model to see how it suits them. "I do say people should be better at partnership. Technology companies should only do what they are better at and work with partners for the rest."
However, the biggest problem with putting it into action, says Sir Robin, is that "they pretend".
"People have to work together as a team and every individual team has to deliver beyond expectations," says Sir Robin. "You need honesty, integrity and to be very challenging. You mustn't be lazy. You should think beyond the possible and then back off to the feasible. A true partnership is where you try to do the impossible. It involves a lot of hard work and brutal honesty.
"You don't just sign a piece of paper; you assign people. Creating some space for communication helps a lot. People do more real communication on the way to the airport than in the boardroom."
Hits and misses
To expand its position in microcontrollers, ARM took a gamble on compatibility with the ARM7, which, up to that point, was the company's most successful core. The core for the Cortex-M3 - one member of a reworked architecture - needed to be small to fit the cost plans of microcontroller suppliers. The engineers could make one small enough that ran the Thumb code but not the full ARM7 instruction set.
"People said we were mad to do it, because ARM7 had that massive installed base," says Muller. "We agonised over it for a long time and, in the end, made the right choice because it opened up the microcontroller market. However, we understood that in other areas, you have to have 100 per cent backward compatibility so we maintained that for other processor cores."
The company was less successful around the same time with a plan to extend its reach beyond the processor core. The company packaged a number of common peripherals under the Primexsys banner. But the customer base did not embrace it: believing that ARM was moving into their territory.
"If I look at where it went wrong, it was ahead of its time," says Muller. "We were building the same systems over and over again. We saw the same issues coming up, so why not standardise? Our lead partners said:
‘We don't want to be standardised'.
"We were a little early because now people are saying that they don't see these functions as their differentiation.
Muller adds: "You can't expect all your production to be successful all the time. As with all your failures, you learn a bit about why you went wrong and how you could do better. Now we have a broad community involved in the activity, so people can understand when they will see benefits. That's what we learned from the Primexsys engagement."
|To start a discussion topic about this article, please log in or register.|
"Africa is abundant with engineering opportunity. We look at some of the projects and the problems."
- UK just six hours from running out of gas in March [09:02 pm 24/05/13]
- Ideas for a final year university project [05:55 pm 24/05/13]
- Fourth Generation Nuclear: Molten Salt Reactors [10:39 am 24/05/13]
- LED bulb efficiency - its all about the drivers not the LEDs? [09:52 am 24/05/13]
- Marketing from Engineers' perspective [02:18 am 24/05/13]
Tune into our latest podcast