Today, Intel’s x86 architecture reigns supreme on PCs (and millions of servers, such as Google’s, that use the PC organ bank). Anywhere else, the ARM processors have won; they’re in billions of devices, regular cell phones, smartphones, entertainment devices, navigation systems and legions of other embedded applications.
Understandably, perhaps, Intel didn’t want to play in the low end of the processor market. But we now see the emergence of RPCs, Really Personal Computers, more commonly called smartphones. Nokia, RIM, Apple and the fast-rising army of Android licensees all use high-end ARM derivatives.
‘Wait’, says Intel, ‘over time, our proven semiconductor design and manufacturing capabilities will allow us to reduce the power consumption and cost of x86 processors. That’s how we’ll win this emerging market, just as we won the PC.’
Easier said than done. The older and more complicated x86 architecture is inherently disadvantaged against the more modern ARM architecture. And, as we’ll see, there is more to this fight than semiconductor design and manufacturing prowess.
For context, let’s go to Mary Meeker’s latest (June 7th, 2010) Internet Trends presentation.
By 2012, she predicts, smartphones shipments will exceed PC unit volumes. Approximately 480 million smartphones versus 430 million PCs, going to 650 million next generation devices by 2013:
Just as important, by next year, smartphones unit volumes will overtake “feature phones”:
Smartphones, feature phones? Without losing ourselves in taxonomy games, let’s turn to the popular Blackberry devices: they are good examples of the smartphone category. Anything less is a feature phone, sometimes called a regular phone, or a “dumb phone”.
Going back to the inflexion/transition taking place, the iPad usage pattern shows it “usurping” the desktop’s Internet usage:
Next, Mary Meeker points to this year’s big surprise, the iPad’s unforeseen fast rise:
So far, Intel plays no measurable role in this new computing genre. Smartphones (and feature phones as well) all use ARM processors.
Does this mean Intel didn’t see this coming? Has the king of PC microprocessors given up on smartphones and tablets? This isn’t like the feisty company we know. Tim Jackson’s book, Inside Intel, makes for enlightening and, at times, disquieting reading as it explores Intel’s culture. The company’s take no prisoners ethos resulted in brushes with US and European courts, leading to costly settlements, such as the latest $1.25B payment to its competitor AMD.
Intel people dislike the Wintel word: it points to their being prisoners of their good fortune, to their Microsoft-dictated role as the supplier of microprocessors that run Windows. To better see their problem, consider two equivalent processors: same computing power, same silicon technology, same power dissipation, same size, same manufacturing cost. But with one difference: one runs Windows, the other doesn’t. Which one will fetch the better price?
This “running Windows” advantage killed all competitors: the PowerPC from IBM and Motorola, DEC’s Alpha, HP’s Precision Architecture (smothered into Intel’s Itanium, properly nicknamed the Itanic), MIPS.
Because of a fundamental philosophy difference.
Traditional Intel does “everything”: it designs, makes and sells a range of x86 CPUs.
ARM, on the other hand, doesn’t make and sell chips; it licenses microprocessor designs. Over the last two decades, designing electronic circuitry has largely become a software affair thanks to EDA tools. With such design software, you take ARM’s licensed designs, presented as ‘‘libraries’’, and adapt them to your needs. When done, you turn to a manufacturer, a “fab”, such as Samsung, TSMC, Global Foundries and many others. One crucial advantage of the ARM world stems from the customization process, often with additional libraries licensed elsewhere. The result is what we call a SOC, a System On a Chip. A SOC reduces cost, physical dimensions and power consumption by adding “non-processor” functions to the chip. This is what all smartphone makers look for. Hence ARM’s omnipresence in the emerging world.
Intel’s absence is puzzling, especially when one remembers the company got an ARM StrongArm license as part of another settlement, with DEC this time. StrongArm became XScale and, surprisingly, was sold to Marvell in 2006.
Four years ago, Intel bet against ARM.
As stated at the beginning of this note, Intel’s view is they’ll win against ARM using the sheer force of their semiconductor technology. Today, ARM processors are smaller, cheaper and consume less power than Intel machines, but Intel thinks their design and manufacturing might will win the day, some day. For low power applications, Intel deploys its Atom sub-family of x86 machines. This works for Windows netbooks, but power consumption is still too high for smartphones or tablets such as the iPad, which can live for about 10 hours on a 24.8 watt-hour battery:
That’s an average 2.5WH consumption, including a big display, radios (WiFi, Bluetooth, 3G), storage — and the A4 ARM derivative. According to EE Times Europe, the iPad’s A4 consumes between 450mW and 800mW, depending upon the application it runs. Comparable Atom processors, with a graphics unit, consume about two or three times as much. (Typical netbooks come with a 7200mAh, 7.4V battery; that’s 50 watt-hours, lasting from 3 to the ‘‘spec sheet’’ 5 hours, versus the ARMed iPad’s 25 watt-hours lasting 10.)
Intel seems to believe they can close the gap. But there is little reason to believe their highly motivated competition won’t keep raising their game.
There might be another factor to Intel’s bet against ARM: Microsoft. Following the “Everything is a PC” motto, the Redmond company believes tablets will run a version of Windows 7, hence the recourse to x86 processors.
So far, x86-based PC tablets haven’t taken the world by storm. And it looks like Nokia will continue to use ARM processors for its future tablet based on the merger between Intel’s mobile Linux, Moblin and Nokia’s Linux derivative, Maemo, the whole thing being now called Meego; ChromeOS tablets are also likely to use ARM processors, not x86 CPUs as Google acquired Agnilux, a processor design firm somehow related to Palo Alto Semiconductor, a firm acquired by Apple, which later also acquired Intrinsity, another processor design team. It also appears HP’s tablet, based on their acquired WebOS, will use ARM processors.
As discussed here two weeks ago, the PC isn’t going away. But, as Mary Meeker sees it, a new era has begun and Intel isn’t part of it.
This absence might be a simple question of profit margins: the revenue by ARM processor (or SOC) is much smaller than with x86 chips, less than $10 per unit in most examples. Intel might harbor the usual cannibalization fear: cheaper ARM processors might displace x86 CPUs.
Another speculation is Intel’s “other half” objects to processors that don’t run Windows. That assumption doesn’t hold water: Microsoft’s first Xbox used an x86 engine but the newer Xbox 360 uses a… PowerPC derivative. Microsoft’s Zune media player uses an ARM derivative, the Tegra chip, made by Intel’s “legal nemesis” Nvidia. The same applies to smartphones powered by Windows Mobile, in the past, or by Windows Phone 7, later this year: all ARM-based devices.
This sends us back to the bet on the “technology brute force” as the only reasonable explanation, coupled with a “Windows for ever” belief. As we can see below, Intel shares, like Microsoft’s, haven’t done much since 2000 and the bursting of the Internet bubble:
Wall Street isn’t showing much belief in Intel’s reliance on its x86 architecture to win in the new personal computing world.
We’ll see if and how they cave in and get a new ARM license.