Microsoft Takes Another Shot at Windows on ARM

At its Windows Hardware Engineering Community (WinHEC) event in Shenzhen, China this week, Microsoft announced a partnership with processor vendor Qualcomm that will bring Windows 10 to the ARM platform. It’s an interesting, if widely expected, move that sets the stage for some unique new Windows products going forward.

Of course, Microsoft has been here before. Back in October 2012, when the company launched the ill-fated Windows 8 for traditional X86 platforms (Intel and AMD), the company also launched the Windows RT operating system, running on Surface RT (as well as a few partner devices). This ARM-based version of the OS looked like Windows and even mostly acted like Windows. But it couldn’t do the one thing everybody expected Windows to do: Run years of legacy X86 applications. While Microsoft tried to make up for this by creating a special Windows RT version of Office, other application vendors failed to follow and, after several years of poor performance in the market, Microsoft quietly ended support for WinRT.

Emulate This
Neither Microsoft nor Qualcomm has disclosed the specifics of how Windows 10 runs on Qualcomm’s Snapdragon processor except to say this time out they’re using emulation technology. Emulation is a process that essentially tells the operating system it’s running on X86 when in fact it is running on ARM. The industry has used emulation for years, often as a stop gap measure when shifting from one technology to another. Traditionally, emulation has solved the problem of legacy apps, but at a significant cost to performance. If we believe Microsoft’s online videos of Windows 10 running on Qualcomm, there appears to be plenty of headroom in the Snapdragon processor to account for this. Performance looks robust and during the demonstration, the presenter launched and used several legacy X86 applications with no apparent performance penalties. I look forward to testing this firsthand in the future.

Why Revisit ARM?
While Microsoft’s decision to do the work required to bring Windows 10 to ARM is significant, it’s worth noting it has had a version of Windows running on the platform continuously since well before it launched, and ultimately killed, Windows RT. It’s called Windows Phone and at present, its smartphone market share is hovering around 0.2% percent, as Microsoft has all but exited the market. Clearly, Microsoft realized Windows Phone was a dead end but it felt Windows 10 on the platform could still be viable.

It’s an interesting decision and one with multiple potential implications. A primary one, cited by Microsoft in its announcement, is the desire to enable highly mobile, power efficient, and always-connected cellular PCs. That last element—always connected—is something I’ve been arguing I’d like to see from PC vendors for ages. Microsoft apparently sees Qualcomm’s chips as a key element to making this more viable in the market. The company also announced it would enable eSIM technology it says will let vendors ship products with no SIM slot, as well as let customers activate data plans on the device.

Second, the company clearly sees value in driving competition in the silicon space. Back in 2012, Windows RT was most certainly Microsoft reacting to its relative displeasure with the slow progress of X86 toward more mobile, lower-power processors. That frustration likely hasn’t entirely abated. As in all markets, competition is good for consumers and, in this case, Microsoft (and its hardware partners) are the consumers of X86 and ARM processors.

Third, Microsoft is trying to anticipate the future of personal computers. One such future may comprise the ultimate in convergence where one device acts as phone + notebook + desktop. For some time, there have been rumors of a Surface Phone; rumors that didn’t make much sense when Microsoft seemed to be abandoning Windows Phone. That version of Windows has always been a non-starter for business because it can’t run the legacy X86 apps needed by most companies. HP’s Elite X3 Windows Phone addresses that problem with its own virtualization technology. Now, Microsoft has addressed it at the OS level. That said, this solution still doesn’t address the fundamental issue that few people (especially consumers) want to replace their iOS or Android phone with one that runs Windows.

2017 Is Getting Interesting
During its keynote, Microsoft also outlined its plans with Intel around the new Project Evo; updated its campaigns around Microsoft HoloLens and the Windows Holographic Platform; and added another vendor (3Glasses) to its list of head-mounted display partners expected to ship next year. But the Qualcomm news is likely to be the most impactful in the near term. The company was deliberately vague about when new products might ship, saying simply “as early as next year.” Once we see the first products, their performance, battery life scores, and price points, we’ll have a better idea of just how big a deal Windows 10 on ARM turns out to be. I, for one, look forward to what the PC industry will cook up with this new option.

Published by

Tom Mainelli

Tom Mainelli has covered the technology industry since 1995. He manages IDC's Devices and Displays group, which covers a broad range of hardware categories including PCs, tablets, smartphones, thin clients, displays, and wearables. He works closely with tech companies, industry contacts, and other analysts to provide in-depth insight and analysis on the always-evolving market of endpoint devices and their related services. In addition to overseeing the collection of historical shipment data and the forecasting of shipment trends in cooperation with IDC's Tracker organization, he also heads up numerous primary research initiatives at IDC. Chief among them is the fielding and analysis of IDC's influential, multi-country Consumer and Commercial PC, Tablet, and Smartphone Buyer Surveys. Mainelli is also driving new research at IDC around the technologies of augmented and virtual reality.

10 thoughts on “Microsoft Takes Another Shot at Windows on ARM”

  1. I think the “always connected” part is mostly a red herring: Intel makes cellular modems that are good enough for Apple, they should be good enough for MS too; plus Qualcomm parts also work with x86. So you don’t need ARM/Qualcomm do make connected PCs, and there are already plenty of those.

    I’m uncertain about the “x86 mobility” part. x86 phones (there have been a few, mostly one from Alcatel, though Intel has since discontinued that line of chips) and tablets (there are *a lot*) were/are par for the course re: battery life and power. ARM as been catching up fast, but catching up is just that; overtaking is a whole other game, and then overtaking by a far enough margin to justify the headaches of switching or emulating, is even more difficult. Even if the x86 instruction set is less efficient the ARM’s, it is a vanishingly small part of a modern processor. Maybe MS got destabilized by Intel dropping “phone-x86” (and AMD lacking the means to even try).

    As for future-proofing… maybe. It doesn’t solve the main issue of “no Modern apps”, but does help with the secondary issue of “one missing Modern app”, if Modern apps ever begin to land. I’m kind of doing the same thing today with dual-boot Windows+Android tablets.

    I’m a bit puzzled as to what the target is though, both as regards devices and markets.
    Win32 apps will by definition require a keyboard and mouse, so a desktop dock for phones and tablets, and a standard for that hasn’t been announced… can the idea catch on when docks today are proprietary and cost as much as a midrange Android phone/tablet/desktop and low-end Windows PC (which will still perform several times faster than emulated x86, and be 64-bits compatible, and be compatible with all apps) ?
    In Entreprise, cloud and virtualization make emulation and single-device mostly superfluous…

    One thing for sure: it’s a massive warning shot at Intel. Apple seem to be dissatisfied with them too, with the excuse for slow updates to Mac “x86 CPUs haven’t moved forward that much”, and the ruckus about 16GB Mac Pros mostly due to Intel not supporting more low-power RAM.

    1. What’s up with Intel lately?
      I’m sure you’re aware (maybe not, this may be just a US issue), that there “fast” iPhone 7 cellular modems (Qualcom) and “slow” ones (Intel). Apple has even throttled back the Qualcom units so that both versions are “slow”, which is an absurdity!

      1. And you can’t even pay to unlock the Qualcomms. After the comments about Tesla being so bad for paywalling extra battery capacity, I don’t know if straight disabling is better or worse ;-p

        I think the main issues are that
        – the architecture is very mature (Instructions-per-clock are progressing slowly), and so are fab processes (plus Intel seems to be losing their lead there ?)
        – for client PCs, CPU performance is above good enough. I/O, Storage, graphics, power consumption are the limiting factors. A 5+ year-old midrange i5 is good enough for VR which was not even on the radar at the time. So Intel has nothing to do but watch ARM catching up to where 80-90% of users are needs are met, there’s not much market above that (but it will be Intel’s ^^).
        – there’s no reason for anyone in the Mobile space to pay an Intel tax, and because Mobile is so big Intel is being overtaken by all the money being poured in ARM R&D, production, ecosystems.
        – Intel is slow-mowing. Desktop parts with full support for HDMI 2.0 ie 10-bits 4K @ 60 fps (that’s a 2013 standard) are expected next year (2017). On the ARM side Amlogic for example released the S905 and S912 early 2015, so Intel is 2 years late. Ditto support for low-power memory (the MacBook Pro issue), TB3, USB 3.1gen2, UFS, latest wifi and BT revisions…
        – They’ve failed at getting into phones, Windows has failed at getting into tablets. So they’re left with the smaller convertible market (plus desktops and servers). Where is growth supposed to come from ?

    2. >> “no Modern apps”

      >> I’m a bit puzzled as to what the target is though

      You need ARM to build low-power and most importantly cheap and great tablets. And a tablet that connects to displays wirelessly, can be connected and be used as a windows computer , and that can be connected to the TV , and runs windows games(even in mobile form) , all that at a very cheap price – could be very appealing to a certain demographic: frugal consumers, teenagers and college students ,and gamers.

      But why bother with a niche tablet ? well, if the niche is big enough, suddenly you get interest from app makers. and once you built a 7″-8″ tablet ecosystem, it’s much easier to build a “good enough” 5″-5.5″ phone app ecosystem.

      Also, definetly, like you said, a massive warning shot for Intel.

      1. Those x86 tablets already exist though. Sticking with Cube which is considered the better-quality Chinese OEM, prices are $100 for 8″, $180 for 10″, $300 for 12″ ( ). I got their 8″, and a 12″ from a worse brand, they’re as expected, equivalent to ARM-based tablets from similar OEMs (so, not top quality ^^). ARM tablets do cost a bit less but have half the RAM&storage since they only run Android while the Cube x86s run both Windows and Android. Performance on an Atom tablet is rather higher than anything below ARM-A72, though battery life is a bit worse, rest is equivalent.

        You also have Windows TV boxes (calling them PCs is a bit of a stretch, though for simple needs…) around $100. Those do have a 4K issue, ARM does 4K over HDMI 2.0 at 60FPS in 10 bits, current Atom only does 30fps in 8 bits.

        The only rationale I can find for MS’s move is that they expect the situation to evolve further in ARM’s favor, because right now it’s a wash, and rather irrelevant compared to the apps issue.

  2. @Tom “For some time, there have been rumors of a Surface Phone; rumors that
    didn’t make much sense when Microsoft seemed to be abandoning Windows
    Phone. That version of Windows has always been a non-starter for
    business because it can’t run the legacy X86 apps needed by most

    I don’t think the lack of legacy x86 desktop Apps on a Windows Phones has been keeping them out of business, and I don’t think this move will help sell Windows phones.

    This is Windows RT version 2.0 for cheap laptops/tablets/convertibles. This time at least there will be some legacy support.

    Though I suspect when it really ships it will be a lot of complaints about speed in legacy apps, despite some slicked up videos.

    1. What bothers me is that an Atom Windows HDMI stick is $150 (brand name) / $100 (known Chinese OEM). That’s *less* than HP want for their bare phone dock (just to be clear, just the dock, w/o the phone, is over $150), and the stick is smaller/lighter/almost certainly more powerful and compatible than the dock.

      I’m sure there are some cases where single-device has value (admin costs). But probably fewer than cases where dual-device is better (multitasking, redundancy, performance….). I know I’d rather carry about a phone and a stick than a phone and a dock, and buy several sticks rather than several docks. If only because the sticks will remain usable forever compared to docks that get obsoleted with each phone upgrade.

      The issue is a bit different for tablets because you don’t need a dock or screen, just a keyboard and mouse. But there are plenty of x86 tablets/convertibles with full Windows, of which a lot also dual-boot Android. The value of x86-32 emulated on ARM compared to going straight for x86 is doubtful.

Leave a Reply

Your email address will not be published. Required fields are marked *