Q&A: How ARM stays six years ahead of the game

 

Gallery

By Barry Collins

Posted on 30 Jul 2010 at 11:21

Another set of record profits has underlined ARM’s status as one of Britain’s biggest tech success stories.

However, the bottom line wasn’t the most interesting thing in ARM’s financial report. It also contained details of the company’s next-generation Eagle processor and the firm’s new deal with software giant Microsoft.

In an exclusive interview, we put ARM CEO Warren East on the spot about these two important issues.

Q. Your annual report talks of the new ‘Eagle’ processor that you’re preparing to launch. Can you tell us more about that?

A. We’re going to launch this processor later this year and we’ll talk in more detail at that time. But basically it’s a top-end processor aimed at the apps processor-type chips, running a big open operating system.

We’re looking for more miles per gallon. This processor will either run at the same sort of speed as a Cortex A9, but give you the same amount of compute functionality for less electrical power consumption. Or for the same power consumption, you’ll be able to do more computing in the same period of time.

Q. Do smartphone manufacturers want more computing power or less aggressive power consumption?

A. Well, a bit of both really. Everybody wants utopia and unfortunately you can’t have it. But when we bring out a new processor like Eagle we’re taking a step forward in terms of efficiency, so we’re helping people to beat that trade off a little bit more each time. That’s why somebody should want to buy an Eagle processor rather than simply continue with Cortex A9.

Q. When can we expect to see Eagle in products?

A. We have three lead partners signed up at the moment. We will give them early access, the very first deliverables from which they can make silicon, at the end of this year or early next year. But you won’t be able to go and buy a phone with this in until 2013, at the very earliest.

Q. Obviously working that far in advance you need to build in a certain amount of future-proofing and guesswork about what people will want in three or four years’ time. How do you do that?

A. The way we do it is by being fairly generic. Nobody can guess that far in advance. If you go back to when we started developing Eagle you’re talking about six years typically before something we start developing ends up in an end product. Nobody has that visibility. And so what we create is something that is actually pretty generic and spans a range of expected requirements, and as you get closer to the time, you find maybe a few variants emerging around the one processor that we kick off.

Q. Your report talks about the move to 22nm and eventually 20nm parts. Do you have any idea when we’ll see those and will Eagle be a 22nm processor?

A. Ultimately, I’m sure there will be Eagles implemented on 22nm, but the first implementations will be 32nm. In fact, a lot of the development work has been done on 40nm.

Our designs are all transferable, and you can take an ARM 926 that we designed 10 years or more ago and you can build that on a 32nm process today with the appropriate physical IP. We know these microprocessors have very long life cycles.