Moore’s Law Is Becoming Irrelevant
The CEO of ARM says power-efficient chips for mobile devices will move into desktops, laptops, and servers.
As computing devices proliferate, energy efficiency is becoming more important than raw processing power.
Companies like Apple and Samsung are the public face of the smartphone and tablet boom, but they all rely on ARM, the British company that licenses the energy-efficient processor designs required by mobile devices. Those chips were once considered significantly less powerful than the x86 processors found in desktops, laptops, and servers—a market dominated by Intel—but that gap appears to be closing. Microsoft is exploring a switch to ARM’s technology for traditional computers, suggesting that ARM’s technology will soon shape more than just mobile computing. ARM’s CEO, Warren East, met this week with MIT Technology Review’s senior IT editor, Tom Simonite.
For decades the computing business has been guided by Moore’s Law, which predicts the rate of improvements in computing power. You have a different focus.
We have always been about efficiency, miles per gallon instead of top speed. That’s actually what matters. Mobile is an easy example: you know that phone is constrained because it’s battery powered.
But [even if] you can plug [a computer] into a socket, [efficiency] is a serious issue for the world. Servers use huge amounts of power. Data centers get located in strange regions of the world where it’s naturally cooler. More and more of this mobile stuff [also] means more and more servers are required. We’ve actually changed the way people design servers [by making them smaller and lower-powered]. Instead of being restricted to big data centers where you know you can get massive amounts of power in, you can distribute these things. You could have many more servers. The analogy I would use is routers. Once upon a time, routers were effectively mini-computers in a massive box. Cisco managed to reduce that to things you have in your home. There’s no reason it shouldn’t go that way for servers.
Microsoft just launched the Surface tablet, with a version of Windows for ARM-based chips. Is that a warm-up act for your technology appearing in many more places currently dominated by Intel-compatible chips?
Maybe. I said a few years ago that you would eventually see ARM in these sorts of places, with or without Microsoft, so we’ll just let it happen. It’s a business model thing. Someone building a PC or a tablet or whatever has a choice of ARM-based supplier. That kind of competitive environment is good because suppliers are competitive in terms of price, but more importantly, they’re forced to be innovative. There’s been a lot more innovation in the world of mobile phones over the last 15-20 years than there has been in the world of PCs.
Will those technological innovations benefit traditional PCs?
Absolutely. To me a PC is really just a smartphone in another form factor. It’s got a bit more local storage, and it’s got some types of interfaces that are a bit different. You wouldn’t have a DVD player in your phone because there’s no room for a DVD, but actually there’s no reason why you couldn’t have a connector to an optical reader to have a DVD. TVs are the same. TVs are big smartphones. Computers are kind of medium smartphones.
Bloomberg has reported that Apple is considering switching its laptops and desktops to ARM-based chips.
Sorry, you’d have to talk to Apple about that. We provide microprocessors to enable you to do everything from tiny little sensors up to supercomputers. Apple operates in a couple of different places in that spectrum. Apple can make up their minds about what they want to do.
It’s often said that ARM-based chips can’t be powerful enough to support everything PCs need to do. For example: image editing and processing.
That’s rubbish. There’s nothing intrinsic in the architecture which stops you [from] being at the high end of performance. Traditionally ARM has found a lot of opportunity in things like mobile phones, where you don’t want to have something which is super high-performance because it consumes more power or real estate. But if you pick another design point, like a computer, the battery’s going to be a lot bigger so you can use more energy. It’s like having a car with a bigger fuel tank. If the only fuel tank you want to put on your car is a small one, then you have to be quite efficient. If someone wants to build a Ferrari and have a bigger fuel tank, then fine, you can make the engine do more. The ARM microprocessor was never designed for mobile in the first place. It ran a computer with a Windows-type operating system before Microsoft ever had Windows, called RISC-OS. There’s nothing inherent in the microprocessor architecture that says you can’t have computers and keyboards and mice.
But software built for Intel-compatible systems can’t be run on ARM-based systems. How are you fixing that problem?
You have to apply the pragmatism filter and say there is all sorts of legacy software out there that’s never going to be ported to a new architecture. Nobody’s going to rewrite Lotus Notes to run on ARM. We don’t believe there’s going to be a massive switch to ARM-based PCs overnight. It’s more of a gradual process as legacy applications become replaced by newer applications that are more up to date. Microsoft’s RT is a good example of that. Successive generations of Windows have been variations on a theme for the last 15-20 years, around the point-and-click-with-a-mouse form factor. They had to do a fundamental redesign if they wanted to be part of tablets. That was an opportunity for people to write new applications that are done in a different way.
Is it getting easier to write software to run on an ARM architecture?
It is, in that once upon a time it was a bit of a niche sport. Now there are many phones and other devices with ARM-based chips inside.