Ina Fried follows up on last month’s rumors about Microsoft Windows (not the smartphone version already running on ARM) running on the ARM CPU Architecture which powers nearly all the smartphones of the world with 5 billion devices projected to ship in 2011 by iSupply.  Two weeks ago, Peter Bright had given probably the best explanation of why such a move would be pointless.  The basic problem is that ARM based Windows will run 0% of all existing Windows based software.

Fried’s report is based on “a source familiar with Microsoft’s development plans” which doesn’t give us much to go on.  But for the sake of argument we’ll assume this source to be accurate and Microsoft is indeed planning an ARM based Windows.  Fried explains the reasoning behind this as superior battery life on ARM based systems, but that’s one of the smaller problems with Windows on mobile devices.  Microsoft’s main problem with Windows is its lack of a finger-driven user interface and its existing Tablet PC extensions are designed for mouse or stylus input.

The battery life difference between x86 Intel Architecture versus ARM architecture is rapidly shrinking and so is the performance difference.  The question is whether Intel will close the power consumption gap faster than ARM closes the performance gap, but Intel seems to be getting their sooner with their “Moorestown” product launched last year.  Moorestown type products hold a significant performance lead over ARM based products and its energy efficiency could probably achieve 8-10 hours of battery life on a 7-10 inch tablet form factor with 14 to 25 watt*hour batter.

So why doesn’t Microsoft use Moorestown?  Because Moorestown stripped out a lot of legacy hardware support that would prevent the stock version of Windows 7 from booting up.  But if Microsoft was interested in better battery life in ultra-mobile products, it would be far easier for them to create a version of Windows that runs on Moorestown and future legacy-free x86 products than to create and ARM based version.  The biggest difference is that most existing software written for Windows will run on Windows on Moorestown but zero percent of them will run on Windows on ARM without being rewritten or performance sapping emulators.

But the problem again is that existing Windows software is designed for keyboards and mice, not fingers so they’ll have to be rewritten anyways, at least the user interface portion.  But at least on Moorestown, they don’t have to start from scratch.

[Cross-posted at Digital Society]