Microsoft's Windows on ARM Better for Partners Than Us
By: Mark Hachman
LAS VEGAS—Microsoft's partners are undoubtedly pleased with its decision to port future versions of Windows to the ARM architecture. But should consumers be? I'm much less sure. Rumors of the Windows-ARM partnership slipped out in December, prompting derision from some quarters. Steven Sinofsky, president of the Windows division at Microsoft, explained that the decision was made at a time in the development process where Microsoft could accommodate other designs. "We are making this announcement now to allow greater collaboration across our expanded partner ecosystem so we can bring to market the widest possible set of PCs and devices, from tablets on up, with the next generation of Windows," he said in a published internal Q&A that Microsoft released. "We're at a point in engineering the next release of Windows where we are demonstrating our progress and bringing together an even broader set of partners required to deliver solutions to customers." And that's why the partners are happy: pushing the concept of a "personal computer," which has traditionally been synonymous with Windows, out to a broader market. That's undoubtedly why Intel isn't displeased to see Windows once again to expand to RISC architectures. Windows NT, the code base for Windows XP, Vista, and Windows 7, once ran on ARM, MIPS, and even the PowerPC (although that version never shipped). But the eventual dominance of the X86 architecture pushed the others aside. "We're looking forward to it," said Bill Kircos, a spokesman for Intel who becomes a vice president of sales and marketing for its tablet division on Monday. "By the time that OS starts to ramp in a couple of years, we will have the lowest power, highest performance chips in the market. One need only look at the creation of the netbook category to see what both companies can do. " Nvidia also has to be ecstatic about the whole thing. Once, a PC was assumed to have an X86 CPU, a chipset, and a graphics chip. As the chipset and GPU became integrated, the X86 remained the most valuable piece of real estate. Likewise, the combination of GPU, CPU and some core logic into the AMD Fusion and Intel "Sandy Bridge" CPUs still froze out any vendor without an X86 architecture and a GPU. Nvidia, lacking that X86 architecture – as well as an X86 bus license – was out in the cold. No longer. Nvidia's "Project Denver" allows Nvidia to essentially build an "Nvidia Nside" machine in time with the release of Windows 8. Jen-Hsun Huang, whose designers have been forced to craft lower-profile embedded and automotive designs, has received a new lease on life. In a press conference Wednesday, Huang called it a "game-changer". It is. For Nvidia. Consumers, however, may see things differently. At this point, I have a hard time believing that Microsoft, whose Windows complexity grows with each passing iteration, can develop a version of Windows designed both for many-core, 3.0-gigahertz processors with gigabytes of RAM, as well as a low-power ARM architecture. To expect to do both implies either a modular architecture, where capabilities can be added and subtracted around a single, efficient framework; or a game of marketing, where "Windows" will be applied as a generic label to what will essentially be two separate operating systems, like Windows 7 and Windows Phone 7. And then there's the performance issue. "People are used to how cell phones operate," said Tom Starnes, an embedded processor analyst with Objective Analysis. "They used to be able to blame it on the bad Internet connection, the bad cellular connection, that sort of thing. 4G kind of takes away some of that blame. Now there are some pretty fast ARM processors out there, but are they the equivalent of Intel's 1-GHz massive cache?" Consumers can be distracted away from poor performance. Consider the relative lack of capabilities of the original Android phones, compared to their blisteringly-fast LTE-based cousins. But if consumers come to expect substantial performance upgrades and additional features, a manufacturer that falls short of that bar will end up with disappointed customers. How do you measure speed? Consumers don't run benchmarks; if a phone "feels" fast, it is fast. Windows doesn't feel fast, not even on a desktop PC. I'm already frustrated enough when my laptop drops a Wi-Fi or cellular connection. Windows suddenly pop into focus, demanding attention. It's chaos. And then there's the power issue. "Texas Instruments (TI) is pleased to collaborate with Microsoft to make Windows on mobile devices a reality," the company said in a statement. "TI has long believed that low power is the fundamental of fundamentals when it comes to mobility, and we look forward to taking our popular OMAP platform, with the ideal balance of high performance and low power, into exciting new mobile computing devices." That's the right approach. If Windows is simply designed to run in a low-power environment and graciously accepts being plugged into a wall socket or docked, congratulations. Windows does this today, so I'm hoping that this isn't a problem. TI, Qualcomm, and especially Nvidia, however, have bet their Windows future on Microsoft. That's not always the best bet. Page back a few years, and numerous features that Microsoft once promised – WinFS comes to mind – never make it in the final OS. As with most announcements, I'm cautiously optimistic. There are opportunities here for consumers and manufacturers alike. But in two years consumers are going to expect faster, more connected, infinitely more streamlined experiences. I'm not sure that the Windows legacy lends itself to that. http://www.pcmag.com/article2/0,2817,2375250,00.asp
