Your Ad Here

Tuesday 20 December 2011

The Engadget Interview: ARM co-founder John Biggs


You've heard of ARM, right? The little chip design company that started out as twelve engineers in a barn in Cambridge, UK, but is now responsible for 25 billion of the microprocessors on this planet? Good, so now you need to meet John Biggs, who's been there since the very beginning -- since before the barn, in fact, when ARM was just a pesky little project within Acorn Computers. Read on for John's straight-up perspective on the company's frangible beginnings, its deep rivalry with Intel, and what those three letters actually stand for. 

You played around with computers and chips as a teenager, but how did you first become involved with Acorn, the parent of ARM? 

A friend of my brother got a job at Acorn to write the manual for one of their new computers, called the Electron. He bluffed his way into it, but then realized he had very little idea about computers, so he asked me to help him with the technical stuff. So I spent the summer of 1982 at Acorn, and then during university breaks I went back there to do different things, until they finally promised that they'd give me a proper job when I graduated.

Once that was lined up, I didn't even bother looking for other jobs. But then the whole home computer market went 'pop' between '84 and '85 and Acorn really suffered. Someone called me up during my finals and said: "you know that job we promised you, well, it doesn't exist anymore." Hermann Hauser [Acorn's CEO] felt bad about it and wrote me a glowing reference, which helped me to get a job at Cray Research. I didn't enjoy the corporate environment there at all, so a year later I used Hermann's own reference to get myself back into Acorn.

That was in just in time for you to help Acorn develop its first chip -- the "Acorn RISC Machine." How did that get off the ground?

Acorn needed more CPU power in its machines. 8-bit was no longer enough, so it was looking at 16-bit. Initially, we asked Intel for samples of their 80286 processor, but they refused. That's funny, because ARM is now perceived as a competitive threat to Intel, and you can trace that all the way back to the seminal moment when they refused to give us those samples. We somewhat arrogantly thought 'we can do this ourselves' and we decided to build our own 32-bit silicon. If Intel had been more generous, Acorn might have built its Archimedes computer around the 80286 processor -- it would have changed the whole landscape.

Why did Acorn let its chip division split away in 1990 to become a separate company, Advanced RISC Machines? 

It was a bit like Hitchhiker's Guide to the Galaxy, when this great civilization decided to get rid of all its insurance salesmen, telephone sanitizers and other useless people, by ferrying them off to another planet. It convinced them that their home world was about to be hit by a terrible catastrophe -- a giant mutant star-goat -- and that they needed to escape on a space ship called the 'B Ark' to save themselves.

Our chip design unit was losing too much money and Acorn wanted shot of us, so they sent us off in our own B Ark: twelve engineers who went away to start the new company. Luckily, we jelled with Apple, who really wanted access to ARM processors and so decided to invest in us. Another company called VLSI Technology became our foundry, contributing the design tools and then manufacturing our chips. As it turned out, there really was a mutant star-goat -- we survived and Acorn didn't.

When did things start to turn around? And when did your shares become valuable?

It really took off in the mid '90s. That's when the ARM7-TDMI was licensed by Texas Instruments and designed into the Nokia 6110, which was the first ARM-powered GSM phone. The company moved out of our little barn and then we were listed on the stock market around 1998. At first, none of us believed our shares would go anywhere -- if you looked at examples like Acorn and Tadpole, many Cambridge technology companies were financially not very successful. Nothing was certain, but we were lucky to be in right place at the right time.

That must have been a pretty good period in your life? 

Yes, it was. I bought an old rectory and renovated it. I got married in 2001 and we traveled to places like the Galapagos, Ecuador and Peru. And then in 2003 we had twin boys, the first of our four sons.

When the company floated, it stopped being Advanced RISC Machines and simply became ARM Holdings. But will ARM always design reduced instruction set chips?

Yes, we'll always be RISC. For a given amount of money, you get a great deal more performance. Or you can have slightly more performance for a lot less money. The low-power thing was a happy accident, which of course became important for handheld devices with limited battery life, but it wasn't initially the prime concern when we started at Acorn.

The idea of RISC is that you focus on instructions that the chip does 80 percent of the time, and you make those instructions run really fast. You don't worry so much about the other, less common instructions. So you end up with a machine that runs a bit slower 20 percent of the time, but is faster on aggregate. With a complex instruction set computing [CISC], you build silicon that can cope equally well with all instructions, but it's slower on aggregate.

The main reason Intel is doing complex instruction set chips is because of its legacy -- the whole RISC concept only came on stream after it had established 8086. It couldn't easily switch due to backwards compatibility issues with existing software.

How does it feel to be in a company that doesn't actually make anything? Isn't it risky to reveal your designs to other companies and potential customers who can build chips themselves?

No, because of patents. In order to make an ARM chip, even though we call it an open architecture, we believe you cannot implement an ARM core without violating our patents. So you need a license agreement with us. Then we negotiate how much you pay for the license agreement, which may be a few million dollars, plus something else for royalties, but that compares to the hundreds of millions of dollars you'd need to design a chip from scratch. Then our customers can get on with their product differentiation.

How can you design chips when you can't make them? What about trial and error?

We can simulate how a chip will behave: its performance and energy consumption. It's not completely predictable, but we can factor uncertainties into our models. Plus, we do build stuff too -- it's not completely cold. We just don't make stuff to sell it. We work with our primary licensee partners to develop test chips.

What's the best thing about being at ARM? I mean, in your career, you only have one other company -- Cray Research -- to compare it to.

The atmosphere here. Cray Research was all suits and ties, and it had a hierarchy based on length of service. ARM is more like a Californian company; shorts, t-shirts, sandals, whatever you want. It's a culture that goes all the way back to Robin [Saxby, ARM's first CEO]. We had great parties and Robin would always sign off bar bills because he believed the best deals were done over a pint.

No comments:

Post a Comment