The desktop computer as we know it could be in danger! Our hobby is doomed!
Let’s face it. We’re all going to be reading these words a trillion times during the next 12 months. So we decided to head this one off at the pass. Is the onset of ARM a real threat to desktop computing, or is it more of an evolutionary force?
Over the past several years, the demand for desktop computers in the home has been steadily shrinking, displaced by notebooks and all-in-one systems. In fact, 69 percent of PCs sold last quarter were notebooks, compared to the falling 22 percent that were desktops. This has been the trend ever since notebook sales caught up to desktop sales back in 2008. (Hey, it could be worse.) These days, all-in-one systems and notebooks are capable of offering near-desktop performance (so long as you don’t mind giving up the ability to play modern games and upgrade in an affordable, modular fashion).
Something a bit more worrisome is the recent explosion of ultra-mobile products, specifically smartphones and tablets. In just a few years, smartphones and tablets have made stupendous jumps in performance. Formerly pokey computing devices have become pocketable powerhouses capable of many of the same basic functions as an ordinary desktop. The kingpin behind all of this development? ARM, a versatile processor microarchitecture that emphasizes power efficiency, yet is surprisingly high in performance.
What’s the impact of these ARM devices’ ascendancy over PC desktop and desktop component sales? It’s not 100 percent clear, but it’s theoretically possible that ARM could migrate up to the desktop level. Principal analyst Rob Enderle of the Enderle Group agrees, noting, “If current trends hold, there should be at least one credible ARM player in the [desktop] space in five years.” One thing is clear, however—other mobile devices such as laptops and netbooks are suffering. Notebook sales have already fallen 11 percent compared to last year.
But how likely is it that ARM’s success will exact blood from traditional desktops? We’re going to take a long, hard look at all the facts to try to settle the question once and for all.
In the beginning, the ARM architecture was specifically developed for use in a PC—the Acorn Archimedes to be precise. In 1987, the Archimedes hit the market, powered by the ARM2 processor with up to 4MB of RAM and a 20MB hard drive. With only 30,000 transistors (less than half of the Motorola 68000’s 68,000), the ARM2 was one of the simplest 32-bit processors of its time. This lower transistor count, paired with the efficient reduced instruction set computer (RISC) architecture, allowed ARM2 to outperform Intel’s 80286 while consuming less electricity.
ARM’s inherent power efficiency and embedded-system approach made it ideal for mobile devices. Toward the end of the 1980s, ARM Ltd. partnered with Apple, and the result was the ARM6 architecture, which, in the form of the ARM 610, powered the world’s first PDA, the Apple Newton.
ARM also licensed the ARM6 architecture to DEC, which led to the development of StrongARM in the mid-1990s. Its purpose was to provide a faster ARM processor for high-end, low-power embedded systems, such as set-top boxes and PDAs. In 1996, the first StrongARM processor, the SA-110, was released. Capable of speeds up to 200MHz and performing at about 1.0 DMIPS/MHz, it was the fastest available processor for portable devices throughout the entire year (until the StrongARM SA-1100 came out). StrongARM became so overwhelmingly popular in mobile devices that Microsoft eventually dropped support for all other architectures in its Pocket PC software. Shortly thereafter, Intel was able to purchase intellectual rights to StrongARM and developed the XScale architecture, which still exists today in many Marvell products.
This initial licensing contract with DEC pushed ARM Ltd. to continue licensing the ARM architecture on a much larger scale, resulting in the widespread proliferation of ARM applications we see today, from routers and NAS devices to smartphones and MP3 players.
You know the rest of the story. Advancements in ARM processors, such as dedicated graphics chips, multimedia instruction sets, and embedded flash memory, have allowed smartphones to evolve into the high-end multimedia devices they are today.
Modern desktops and laptops still hold a hefty lead over smartphones in performance as well as functionality, with features that make them suitable for server operation, home theater use, and high-end gaming. But for the mainstream user, who only uses his computer to get online, send emails, and watch the occasional YouTube video, a lot of that functionality is overkill. Herein lies the potential threat; these are all things that smartphones can already do; the biggest limitation is the awkwardly cramped interface. But even this is starting to change as new interfacing options enter the market.
Earlier this year, AT&T launched the Atrix 4G. The smartphone uses one of the most powerful ARM processors to date, Nvidia’s dual-core Tegra 2, and comes with a plethora of accessories. One accessory, the Laptop Dock, literally turns the Atrix into a netbook-caliber laptop, complete with an 11.6-inch display and full keyboard. Another, the Media Dock, allows you to connect an HDTV or monitor and several USB devices (such as a mouse and keyboard), effectively turning the Atrix 4G into a super mobile desktop .
We found that while snappy in a smartphone, Tegra 2 struggled a little bit in powering a netbook platform. This fall, however, Nvidia’s quad-core Tegra 3 (code-name “Kal-El”) will up the ante with performance estimated to be close to that of an Xbox 360. That includes graphics. It’s easy to speculate that, within a year or two, we’ll see many Atrix-like devices on the market. It’s also not hard to imagine that the sales of these systems will negatively impact desktop sales.
The truth is that neither Intel nor AMD have anything that can quite match this combination of mobility and performance. Not yet, at least. In the meantime, ARM appears ready to take the battle to Intel and AMD by penetrating the desktop and laptop market.
ARM isn’t the only company plotting to mobilize the world and liberate people from their desks. Intel has plans to enter the smartphone and tablet market with its latest version of the Atom processor. The Z6xx series is a system-on-a-chip (SoC) design that takes the Atom’s processing core and combines it with the system’s memory, memory controller, and graphics processor.
Early results indicate that this, combined with newly designed processor components may allow the Atom Z6xx to compete with, and even outperform most current ARM offerings, all while consuming significantly less power than previous versions of the Atom. Intel is predicting a peak TDP below 3W, which is still significantly higher power consumption than competing ARM processors. The Z6xx also has the added benefit of x86 compatibility, making it possible to run more advanced applications and even full-fledged versions of Windows on tablets and smartphones featuring the Z6xx processor.
While Intel’s intentions in the ultraportable field are clear, AMD is still on the fence. It wants (perhaps even needs) a share of the smartphone and tablet market, but to do so, it needs viable chips, and soon. According to both Rob Enderle and industry analyst Tom Halfhill, AMD has two options: license the ARM architecture and develop its own ARM chips (like Nvidia’s Tegra line) or follow Intel’s lead and scale down its current x86 architecture. AMD has publically stated that it intends to develop its own APUs, much like Intel’s Z6xx processors. It’s a risky move, considering Intel’s deeper pockets, but AMD’s previous acquisition of ATI could prove useful here. Nvidia’s Tegra platform utilizes an advanced graphics core, while Intel benefits from x86 compatibility. AMD’s unique situation allows it to combine both of these traits. We should see the end results with the company’s rumored “Desna” APU.
Rumors are swirling that by the second half of 2013, Apple will switch to ARM in its MacBook and MacBook Pro lines, and possibly even its desktops. As Enderle puts it, “This would follow Apple’s strategy to own the technology they use.” Assuming price/power/performance ratios work out, it’s easy to imagine other PC manufacturers following suit for their lower-end systems, such as all-in-ones and net-tops. Appearing to confirm this migration, research firm IDC’s quarterly forecast in May 2011 predicted that ARM will hold a 13 percent share of the PC market by 2015.
The ARM PandaBoard is a TI OMAP 4330–powered prototyping and development board.
Obviously, this will be an uphill battle. And how to measure success? Maximum PC CPU columnist Tom Halfhill, who is also a senior analyst with The Linley Group and senior editor of Microprocessor Report, says, “Intel will do everything it can to keep x86 processors in Macs, but Apple has a history of switching CPU architectures when it’s advantageous to do so. Each time, critics predicted doom. Each time, Mac sales went up. But the professional graphics market is still important to Apple, so high-end Macs won’t switch to something less powerful than the x86. It’s possible that low-end Macs—especially notebooks—will switch to ARM in the future.”
Since ARM processors are RISC based and don’t feature full x86 or x64 instruction sets, there’s no chance of getting any current versions of Windows to run properly on an ARM platform. This will change when Microsoft releases Windows 8. Microsoft plans to include ARM support with the release of its next OS, expected to launch in 2012. At CES this year, Microsoft had a rough, but working version of Windows running atop ARM processors.
Given the dominance of Windows in home and corporate settings, it’s likely that the presence of Windows tablets—Microsoft is even rumored to be working on its own proprietary model—will send sales figures rocketing. But just like Apple’s transition to ARM, this doesn’t come without some headaches (and conflict). According to Halfhill, “Legacy Windows apps simply won't run on an ARM processor without emulation. Apple has proven—twice—that software emulation can ease the difficult transition to a new CPU architecture. But each time, the new processors were more powerful than the old processors, making software emulation practical. That probably won't be the case with an ARM-based Windows PC. At best,” he says, “the ARM processors will be equally powerful.”
Enderle agrees, expecting legacy application compatibility to be “pretty poor” and goes on to say that “the ARM version [of Windows] is likely not where you’d want to run [legacy applications] anyway.”
The million-dollar question is this: Will Microsoft’s involvement with ARM benefit the desktop crowd? It’s a mixed bag. While the downwards migration of Windows will certainly cannibalize some of the desktop market, it won’t eat into the high-end market share Maximum PC users care about. The other way of looking at it is that this migration could result in a proliferation of purpose-specific ARM-based PCs.
In reality, ARM may seem like a threat, but it still has a long way to go before it will be a suitable desktop platform. And even then, it won’t be capable of meeting the graphics demands of gamers, the processing demands of video and graphic specialists, and the flexibility demands of programmers.
Should the day come that ARM (or Intel or AMD) manages to create a mobile platform that can match the performance of desktop computers of the time, people will welcome it with open arms, but the transition won’t happen overnight. Despite the inroads laptops have made, desktops are still more powerful than their notebook counterparts.
“The lines between ARM and x86 will blur in the middle ground,” Rob Enderle says. “But ARM will sustain smartphones, and x86 will have high-performance PCs and workstations.” For the short- and mid-term, ARM appears to be an additive dimension of computing rather than a substitute.
If ARM plans on providing an adequate platform for devices like Apple’s MacBook line, it has a few performance gaps to fill. Currently in the works is the Cortex-A15 core, which, clock-for-clock, is up to 40 percent faster than the Cortex-A9 found in the latest smartphones and tablets. That alone isn’t enough of a boost to compete with Intel in notebooks; however, the Cortex-A15 can house up to eight cores, sharing a total of 8MB of L2 cache, and can even be clocked up to 2.5GHz.
Although the specifications call for a maximum of eight cores (four cores per cluster, two clusters per chip), a specialized AMBA4 interconnector can be used to connect two chips together, potentially running 16 cores together, at the full 2.5GHz.
Unlike ARM’s other cores, the Cortex-A15 is tuned for performance, rather than power-efficiency, though it will still achieve a healthy balance of both, suggesting that ARM intends it to be used in more than just handheld devices. Cortex-A15 chips are expected to reach the consumer market by the end of 2012. This syncs up nicely with Apple’s predicted switch to ARM by 2013, but as of yet, there is no evidence that Apple has licensed this new core.
Thus far, Texas Instruments, ST-Ericsson, and Nvidia are among the licensees, and TI actually already has a chip in the works: the OMAP 5430. As just a dual-core chip, the OMAP 5430 doesn’t appear to push the limits of the Cortex-A15’s potential, but at 2GHz, it does boast a 3x performance boost over TI’s previous OMAP 4330. If anything, Nvidia will be the one to watch for high-performance, desktop-potential Cortex-A15 chips.