Microsoft Porting Windows to ARM But Drivers Will Take Time

One of the big announcements at CES this year will be from Microsoft; the company plans to announce (and possibly demonstrate) a consumer-centric version of  Windows running on ARM processors. The company's decision is a powerful endorsement. Other Microsoft products like Windows Embedded and Windows Mobile support ARM products but the software giant has never released a non-x86 mainstream version of Windows.*

It'll be quite awhile before we see an actual shipping OS. Microsoft has already sunk months of work into the endeavor, but vendor driver support is expected to take quite some time. There's also the question of software support—one reason the non-x86 versions of Windows NT ultimately died is because Microsoft didn't provide service packs or maintain application parity between the various RISC architectures and IA-32.

ARM's upcoming Eagle processor will outperform the Cortex-A9--but will also add certain features that the company previously avoided on account of their power consumption.

We know that the new version of Windows will target a range of devices from slates to notebooks but Redmond has yet to reveal if the next-generation product is based on WP7, Windows 7, Windows 8, or Windows CE. A number of analysts are betting that what Microsoft unveils at CES will be an early version of Win 8 running on ARM hardware—but if that's true, it means the ARM-compatible version of Windows is at least two years out.

A 2013 launch timeframe would give vendors two important options not currently available. First, ARM's Cortex-A15 is expected to ship in the 2012-2013 timeframe. According to ARM, the A15 will offer up to 5x the single-threaded performance of the current Cortex-A9, integrate a more-powerful FPU, run at speeds of up to 2.5GHz, and scale from 1-8 cores.

If the A15 is delayed, device manufacturers would still be able to opt for Cortex-A9 processors built on thoroughly mature 28nm technology. According to certain industry vendors we've spoken to in the past it's already possible to build an ARM SoC with sufficient CPU and GPU power to run a modern version of Windows but the hypothetical SoC's power consumption would make it less than ideal for a modern smartphone.Two years from now, that shouldn't be an issue.

Architecture Showdown Fast Approaching

ARM has made no secret of its plans to challenge Intel from mobile devices to low-end servers once the A15 arrives. In the past, Intel could count on Windows compatibility as a significant positive, at least as far as netbooks and notebooks were concerned. Windows ARM could potentially negate that advantage provided it isn't crippled by anemic software support. At that point, ARM and Atom would have to slug it out based on performance, power consumption, and battery life. In our previous discussions with the company, ARM representatives have indicated they believe the company's two decades of experience designing low-power processors and strong vendor relationships will be sufficient to compete in emerging mobile markets.

*Windows NT 4.0 supported Alpha, MIPS, and PowerPC but was never intended for mainstream/consumer use. The IA-64-compatible version of Windows XP doesn't count at all.