ARM vs. Atom: The battle for the next digital frontier
- 28 October, 2009 21:50
For once, Intel knows how it feels to be the underdog.
Over the past 25 years, Intel has risen to become the leading supplier of microprocessors for home and business computing, commanding a virtual monopoly in the market for desktop, laptop, and server CPUs. Even Apple has joined the choir.
But CEO Paul Otellini isn't content to stop there. He envisions a world in which Intel chips power every device, from the grandest server to the humblest media appliance -- a "continuum of computing" that spans many tiers of processor power, all united by Intel's x86 architecture.
Key to this vision is Atom, the most recent entry in Intel's processor line. Compact and extremely energy-efficient, Atom is already the leading CPU for netbook computers. With its latest, ultra-low-voltage versions of the chip, Intel is poised to take x86 even further down Otellini's continuum, away from PCs and into the world of handsets, media players, smart TVs, and other digital electronic devices.
It won't be easy. Intel may be the reigning king of PCs and server CPUs, but in the world of mobile devices, that title goes to an unlikely rival: a small, unassuming company called ARM Holdings, based in Cambridge, England.
Most consumers have never even heard of ARM. You won't see ARM ad campaigns in magazines or on TV. There are no stickers proclaiming "ARM Inside!" The company employs fewer than 1,800 people, and at $3 billion, its market capitalization is a mere fraction of Intel's. But make no mistake -- ARM and Intel are on a collision course. What happens next could determine the shape of the computing industry for years to come.
The next digital frontier The stakes are high in the market for electronic devices, but the opportunity is massive.
Consider: Intel sold its 1 billionth x86 chip in 2003. Its closest rival, AMD, broke the 500 million mark just this year. ARM, on the other hand, expects to ship 2.8 billion processors in 2009 alone -- or around 90 chips per second. That's in addition to the more than 10 billion ARM processors already powering devices today.
Pick up any mobile phone and there's a 95 percent chance it contains at least one ARM processor. If the phone was manufactured in the past five years, make that 100 percent; that goes for standard handsets as well as smartphones.
The same is true for portable media players. Whether the label says Archos, iRiver, or Sony, inside it's ARM.
You'll also find ARM chips in wireless routers from D-Link, Linksys, and Netgear; printers from HP, Konica Minolta, and Lexmark; graphing calculators from HP and TI; GPS devices from Blaupunkt, Garmin, and TomTom; and countless other devices. Even the flight information system on Burt Rutan's SpaceShipOne was powered by ARM.
Every one of these applications is a potential opportunity for Intel, but until recently x86 chips were generally considered too power-hungry -- and too expensive -- for use in embedded applications. Atom is changing that, but Intel still needs to convince device manufacturers that it can be as good a partner as the existing ARM-based ecosystem.
ARM: The mouse that roared ARM may hardly be a household name, but to insiders it's one of the most recognized brands in the semiconductor industry. By volume, it's the most successful 32-bit processor architecture in the world, accounting for more CPUs in more devices than any other -- without so much as a single ad on TV. This seemingly paradoxical success is the direct result of ARM's unique business model.
Intel rose to the top of the computer industry the old-fashioned way: fighting tooth and nail. It guards its processor designs jealously. Even where Intel licenses its technology to other companies -- such as AMD -- it still competes head to head with those licensees for the same markets.
ARM, on the other hand, is all about partnership. It owns no fabrication plants and sells no chips under its own banner. Instead, it licenses its CPU core designs to more than 200 semiconductor companies worldwide. Prominent American licensees include Freescale, Marvell, Qualcomm, and Texas Instruments.
Each licensee is free to package the ARM technology with its own custom modifications and market the resulting chips under its own branding. For example, the CPU that powers the iPhone 3G S is sold as a Samsung S5PC100, but inside it's a 600MHz ARM Cortex A8 core coupled with Samsung's proprietary graphics, signal, and multimedia processing units.
That's why there are so many different kinds of ARM processors in so many different kinds of devices. ARM isn't just a single CPU; rather, it's an entire ecosystem, comprising not just processors but also development tools and other connecting technologies, allowing many competing manufacturers to bring a variety of products to serve various market niches, all based on ARM architecture.
In particular, this flexibility makes ARM an ideal platform for building complex, densely integrated system on chip (SoC) products, which typically combine processor cores with memory, signal processing circuits, timers, and external interfaces such as USB and FireWire, among other components.
Intel enters the Atom age If Intel wants to win a share of the lucrative digital electronics market, ARM is the vendor to beat. But relations between the two companies haven't always been so frosty. In fact, Intel even designed and marketed its own ARM-based chips for several years, under the XScale brand. You can still find XScale processors inside BlackBerry 8000-series handsets.
Intel sold its XScale division to Marvell in 2006, however, during a period of general restructuring. At the time, a company spokesperson described the division as a "nonperforming business unit" and went as far as to claim that the handheld market served by XScale was "not a good fit for [Intel]."
Less than two years later, Intel unveiled the chip that would become Atom.
Atom is a brand-new take on the x86 architecture. Working in conjunction with leading netbook vendor Asus, Intel designed the chip from the ground up to offer good performance at extremely low voltage.
The earliest Atom design was still too power-hungry for ultraportable devices like smartphones, but it took off in the netbook market like a rocket. Today, Atom chips power more netbooks than any other CPU. Newer iterations of Intel's netbook-centric Atom line have increased speed and added features, with the latest models offering dual cores.
But Intel isn't content to limit Atom to low-end laptops. Even as the netbook market has accelerated, Intel has worked to refine Atom to suit a new niche, one even further down the Otellini's proposed continuum. According to reports, Intel's latest, hush-hush project, code-named Medfield, aims to produce a version of the Atom that's so tiny and operates at such a low voltage that it can be used in a whole range of consumer electronics devices.
And that's not all. Where historically Intel has manufactured and sold its CPUs as off-the-shelf parts, with Atom it's attempting something new. In March it inked a deal with Taiwan Semiconductor Manufacturing Co. that will allow TSMC and its customers to build custom SoC products based on Atom cores. In other words, Intel is borrowing a page straight out of ARM's playbook.
The compatibility question Atom has its work cut out for it if it wants to win customers away from ARM, but Intel may have one ace up its sleeve: its x86 architecture.
Although ARM has gained almost universal acceptance in the embedded systems market and supports a thriving developer ecosystem, it's not without its faults. Programmers who are accustomed to more traditional PC software development need to learn new tricks to be productive in ARM environments.
This is due in part to ARM's novel history. A unique, RISC-based processor design, ARM grew out of the quirky British computer industry of the 1980s. It was compact and efficient mainly because it had to be -- its British backers lacked access to the kind of capital that propelled Intel to the top of Silicon Valley. But when it became clear that x86 would dominate the PC market in the United Kingdom as it had in America, ARM's efficient design quickly earned it a place of favor among digital device manufacturers.
Atom, on the other hand, is a full-blooded x86 CPU. It's smaller and uses less power than Intel's mainstream PC chips, but it supports the full x86 instruction set and its accompanying programming model. As any netbook owner can attest, an Atom CPU can execute any binary that will run on a Core 2 Duo without modification, albeit more slowly.
Intel is banking that this compatibility will appeal to developers making the transition from PC environments to mobile devices. It means they will be able to use the same compilers, tools, and code libraries to build software for Atom-powered mobile devices as they do for PCs.
That's not to say ARM lacks software. The catalog of operating systems and applications available for the platform has been growing for decades, and includes several complete Linux distributions. Google's Android OS runs on ARM, and so will Chrome OS when it ships. Even some commercial software vendors support it; for example, Adobe recently announced it would ship versions of Flash Player 10.1 for ARM and Intel simultaneously.
One thing ARM doesn't have, however, is Windows. While various flavors of Windows CE will run on ARM devices, Microsoft says it has no plans to port the genuine article. And even if the OS itself did boot, it wouldn't be much use unless major applications vendors ported their software, too.
Admittedly, Windows is overkill for many embedded applications. But the inability to run Microsoft's flagship OS could be enough to stymie ARM's plans to compete on the low end of the netbook market.
Intel: A babe in the woods? If Intel has misjudged the market, however, its emphasis on Windows and x86 architecture could backfire.
Not so long ago, Upside magazine hailed a company called Transmeta as "the most important company in Silicon Valley." Its product sounded remarkably similar to Atom. Transmeta CPUs used advanced, proprietary technology to execute the x86 instruction set in a way that consumed much less power than traditional Intel desktop and laptop chips.
When the first Transmeta chips began appearing in consumer laptops, however, they were a disappointment. Transmeta-powered laptops weren't much smaller or lighter than standard ones, but their performance was perceptibly worse.
The netbook category didn't exist at that time, and battery technology was less advanced than it is today. To the gigahertz-obsessed buying public, the minutes that Transmeta's technology added to their battery life simply weren't worth sacrificing performance.
The situation is similar today, only now consumers demand both speed and power savings. Who cares if a chip uses x86 architecture, so long as it has enough juice to decode high-definition video and won't drain the battery before the movie ends?
Atom's performance is good, but Intel has yet to demonstrate a model with power characteristics comparable to those of the current generation of ARM chips.
Meanwhile, ARM recently demonstrated a version of its Cortex A9 processor running at 2GHz, proving that ARM chips can scale to handle high-performance applications. And a forthcoming ARM product promises to consume one-third as much power as current offerings.
Given numbers like those, Intel's talk about a universal x86 architecture could end up falling on deaf ears -- especially since there's no shortage of ARM programmers in the embedded market.
Let the chips fall where they may Ultimately, as is so often the case when large companies enter new markets, Intel's real best weapon is its deep pockets. Atom doesn't have to be an overnight success in the device market, as long as Intel can prop it up with sales from its lucrative PC and server lines.
But deep pockets won't matter if Intel can't make its Atom business profitable in the long term. While ARM has made its home in the embedded market for years, Intel has grown accustomed to businesses with higher margins. Intel may find it lacks the nerve for the battle to come.
According to sources, each Atom CPU retails for around a tenth of the price of one of Intel's Penryn chips for standard laptops. As ARM makes its chips ever faster and more versatile, Intel will be pressured to follow suit with Atom. But the more powerful Atom becomes, the more Atom sales will cannibalize sales of Intel's higher-margin traditional chips -- leading some analysts to wonder whether this is really a market where a company with Intel's business model can succeed.
But then, it's not as if Intel has a choice. The rise of netbooks, the decline of desktop PCs, the green IT movement, and the smartphone explosion all point to a sea change in computing. Intel can embrace the shift, but it can't stop it.
For ARM and its many partners, however, the world of Intel's future must seem like a very familiar place.
This story, "ARM vs. Atom: The battle for the next digital frontier," was originally published at InfoWorld.com. Follow the latest developments in ARM and Atom, as well as netbooks and mobile technologies, at InfoWorld.com.