How Intel's faith in x86 cost it the mobile market
And why Paul Otellini's successor needs to embrace corporate heresy
You can’t fault departing Intel CEO Paul Otellini by claiming he didn't spot the way personal computing was becoming more mobile. He certainly did. But you can argue that his strategy for adapting the chip maker to the trend really wasn’t the right one. But as a 40-year Intel veteran it was never very likely he would reject one of the company's articles of faith. Perhaps he should have.
That Intel hasn’t adapted to the mobile world is clear. We’re all buying fewer traditional computer form-factors, but never more smartphones and tablets.
Early in his reign, Otellini attempted to position Intel at the forefront of this change. In 2006, the company was quick to promote first the Ultra-Mobile PC (UMPC), a Celeron-based low-end PC in a thick tablet form-factor.
A year later, UMPCs having failed to win a thumbs up from consumers, Intel strove to persuade World+Dog to buy netbooks instead, compact laptops designed for easy portability and Wi-Fi net access. Initially based on Celeron chips, netbooks soon became flag bearers for Atom, Intel’s new family of x86 processors for ultra-mobile devices.
At the time, Otellini forecast Atom would go on to become the foundation for the Mobile Internet Device (MID), another tablet, this one smaller - and much thinner - than the UMPC but more internet friendly than the phones of the time. Indeed, in 2009 the CEO said Intel would, in five years' time - so by 2014 - be selling more Atom system-on-chips (SoCs) than regular CPUs.
x86 for mobility
The common thread running through all these initiatives - apart from mobility, of course - is the use of Intel’s x86-based processors. It’s now driving the largely unsuccessful push to Ultrabooks.
Intel had ARM chips in its portfolio, but one of the first things Otellini oversaw on becoming CEO was to shift his engineers’ attention away from the ARM instruction set architecture (IA) and back to Intel’s core IP, the x86 instruction set. Eventually, he sold Intel XScale ARM family to Marvell.
x86 holds a special place in long-term Intel executives’ hearts. Alongside with Moore’s Law, it is one of the two ‘self-evident truths’ that define the post-1970s Intel: that chips will double their transistor count every 18 months or so, leading to ever more processing power, year in year out - and that those chips will use Intel’s IA.
Neither, of course, are scientific principles, merely a hypothesis and a business strategy. But Intel had a nice run through the 1980s, 1990s and early 2000s demonstrating that these two notions are more fundamental than they actually are.
There’s no reason Moore’s Law can’t continue to define Intel’s chip roll-outs for some years yet. But the rise of ARM, the de facto standard architecture for the world’s phones, smartphones and tablets, shows that x86 does not reign supreme.
Intel executives continue to claim that the key benefit of x86 is compatibility: with Windows and with the vast library of application software already coded for the platform. Otellini banked on buyers wanting to run that software on their mobile devices too.
May be they do, though there’s no real evidence to show that that’s the case. Certainly, users have exerted little pressure on makers of ARM-based mobile devices to develop x86-based versions that can run Windows and Windows apps. Yes, Microsoft is offering an x86 version of its Surface tablet, but that’s as much about Redmond playing all the angles as a firm sense that some folk don’t want an ARM-based Surface.
In the middle of the last decade, Otellini knew that technology was going mobile and that ARM was then best suited, from a hardware perspective, to deliver greater mobility. He pledge to aggressively drive down Intel chips’ power consumption to a point where x86 processors could match the energy efficiency of ARM devices.
Eight years on from Otellini’s elevation to CEO, Intel can argue that it’s there. It has x86 in smartphones and, unlike previous attempts, they don’t deliver a much shorter battery life than ARM devices. The problem now, though, is that the rise of ARM in the interim has eroded the compatibility value of x86.
x86 for compatibility
Manufacturers have used ARM for a variety of reasons, some financial, others technical, such as the platform’s energy efficiency and the way ARM’s licensing model makes for greater system-on-a-chip customisation, the better to differente your chip from your rivals, and the better to tailor your design for for role it will be put to.
Intel could have delivered such a foundation years ago, but its belief in the primacy of the x86 instruction set and the software compatibility that came with it slowed its development work down. Google’s development of Android primarily on ARM, and Apple’s choice of ARM for the iPhone and iOS, shows the didn’t see Intel’s chips as viable alternatives in the mid-2000s and, more importantly, that they were willing to invest in a new IA. Had Intel not been so wedded to x86, it would have had energy efficient mobile CPUs out sooner and might have become the one that defined that IA.
Now it can match ARM on energy efficiency, at least, it’s too late, and the market had shifted even further from the x86 compatibility requirement.
Apple is a pragmatic company. It migrated to Intel from the PowerPC platform because it saw no benefit in staying, and it made the the move a success, just as it had when it went from Motorola 680x0 to PowerPC. Rumours have it that it’s now investigating whether it can make and sell laptops with ARM chips. Its clear processor agnosticism will mean that, if it sees a benefit - financial, technical and/or practical - in doing so, it will shift away from Intel.
Rival firms don’t need to, of course. Samsung, for instance, can happily go on making and selling Intel-based laptops for customers who want one, and ARM-based mobile devices for everyone else. These are not, after all, separate groups. Laptop owners also use phones and many of them have tablets too.
There will, then, still be a role for Intel, as the key manufacturer of Windows - and Linux, since the open source OS runs primarily on Wintel machines stripped of their shipping OS - laptop processors. But as ARM moves up the line into laptops and, soon, mini-desktop ‘living room PCs’ - it’s already putting footholds in the server world - Intel may come to rue its insistence on x86 above all, not because it can't compete technologically, but because the once implicit link between x86 and personal computing has been broken.
Otellini’s departure, two years before it was anticipated, is, he says, about making room for new blood, new thinking. Intel needs to be sure his replacement is not as wedded to the x86 faith as he was. ®