Hands-On with Windows 10 on ARM x64 Emulation

Posted on December 11, 2020 by Paul Thurrott in Windows 10 with 75 Comments

Apple’s ARM-based M1 chipset has thrown the Windows community into a tailspin, given the failure to date of Windows 10 on ARM (WOA). But this week, Microsoft took a major step toward fixing one of WOA’s biggest problems when it introduced the first pre-release version of x64 app emulation. How does it measure up?

It will be a while before I can evaluate an M1-based Mac: I’ve traded in my 2020 (Intel-based) MacBook Air and am awaiting the Apple gift card I’ll get in return so that I can order one. (Most likely a Mac Mini with 16 GB of RAM, but we’ll see.)

But what I can do is see how this improves the application situation on WOA. And do so using a PC, the Lenovo Flex 5G, that is powered by a modern Snapdragon 8cx 5G system on a chip (SoC).

I reviewed this PC in mid-summer, noting that its great battery life, 5G connectivity, and versatile convertible form factor were, unfortunately, undone by its terrible performance and software compatibility, and its expensive pricing. Fortunately, I have my own shortcomings, among them procrastination: I should have sent the Flex 5G back to Lenovo by now, but since I didn’t, I can at least see what Microsoft’s first stab at x64 emulation for WOA looks like.

Before getting to that, let’s review what I see as the three key problems with WOA, two of which are fatal. Listened in reverse order of importance, they are:

Driver compatibility. WOA-based PCs cannot take advantage of any x86/x64-based drivers, like those that hardware makers ship with scanners, printers, mice, keyboards, and so on. I describe this as the least problematic WOA problem because the platform does come with what Microsoft calls class drivers, so most peripherals will still work. They just won’t give you any of the custom capabilities that the manufacturer might provide. So for some, this could be hugely problematic. But for most people, it’s probably not a huge issue.

Performance. WOA systems have struggled since the beginning, and after what I’ll call 3.5 generations of Qualcomm chipsets, performance has definitely improved, just not enough: Even the most modern WOA-based PCs, like the Flex 5G, are sluggish. The good news? This is a solvable problem, and while I can’t accurately predict when some combination of hardware and software improvements finally puts WOA over the top, at least for mainstream users, that will definitely happen. Hopefully sooner rather than later.

Application compatibility. This is WOA’s Achilles Heel. When this platform first launched, WOA could run 32-bit x86 applications in emulation, but very slowly, and ARM32 applications natively, though there were none to few available. Over time, Microsoft added 64-bit ARM64 support to the platform, and this year it finally admitted that it was working to add 64-bit x64 emulation to WOA. That’s what we’re looking at today: The ability of this ARM-based platform to install and run 64-bit Intel-type applications.

Looking at that list, you can see why adding x64 support is so important, even if it’s slow at first. As I’ve observed in the past, slow is better than not working all, and the performance issues are solvable. Just running the applications that users rightfully expect of anything named Windows is thus job one. And so here we are.

Well, here we are finally. Just installing the Windows Insider Preview build that enables this feature was time-consuming, and it didn’t help that I experienced a “green screen of death” (the Insider version of the blue screen of death) while I was trying to install the build. The slow pace of the install isn’t entirely WOA’s fault—Insider builds often take a long time to install, even on x64 systems—but I was reminded of the leisurely performance of this particular PC in getting it ready.

Anyway, after an afternoon of slowly installing the new build, I signed-in and then installed the two prerequisites suggested by Microsoft: A preview version of a  new Qualcomm Adreno graphics driver specifically designed for the Flex 5G and a preview version of the ARM64 C++ redistributable. Then, after a reboot—I wasn’t prompted, I just thought it was prudent—I opened up Microsoft Store and displayed the list of applications I’d previously purchased and installed. There are two key x64 apps in the Store that won’t normally appear when viewed on a WOA-based PC (because the Store filters out incompatible apps): Adobe Photoshop Elements (I’ve purchased versions 15 and 2020) and Affinity Photo. Both now appear in the Store.

So that’s good. I then installed each application, of course, though I’m only using Affinity these days.

Affinity installed more quickly, which makes sense as it’s the (much) smaller application. So I ran that immediately and, perhaps not surprisingly, the performance is very slow. The application takes a long time to start, and operations within the application—like loading a file, or even displaying some menus and dialogs—are likewise slow.

But it does work, and that’s huge. After all, this PC is slow even when running native applications. Of course it’s going to run emulated x64 apps slowly. So it’s pokey. But it gets there.

Adobe Photoshop Elements 2020 was also slow, of course. It’s a beast of an application, after all. Just setting it up for first-time use was slow, and then launching what I think of as the application—the Photo Editor bit—took somewhere between 45 and 60 seconds to even display the first time.

But I was pleasantly surprised by the performance: Elements seemed to just work normally, and my normal image load, cropping, and exporting workflow proceeded normally. Nice. (One wonders if Microsoft somewhat prioritized for Photoshop Elements, given its popularity.)

OK, this wasn’t a particularly comprehensive test. But even in this early state, x64 emulation on WOA seems to achieve its goals. And in doing so, Microsoft has finally, after three long years, eliminated the single most debilitating problem with this platform. In doing so, it only has one other major hurdle to pass, general performance, which, again, seems like a solvable problem. WOA, finally, is on the path to being viable.

And that, folks, is great news.

Tagged with

Join the discussion!

BECOME A THURROTT MEMBER:

Don't have a login but want to join the conversation? Become a Thurrott Premium or Basic User to participate

Register
Comments (81)

81 responses to “Hands-On with Windows 10 on ARM x64 Emulation”

  1. cjhawkins3rd

    Also, it sounds like WOA is ahead of Apple M1 for compatibility at this point in time, based on reading other parts of this site. (Although neither is ready for the "mass" market at this time.)

  2. jblank46

    One thing I keep seeing about Rosetta is that it translates the app binary on first launch and then performs mostly like native arm from that point forward. I swear I read somewhere that WOA does something similar, caching instructions on disk. I've seen people reference that WOA is all JIT but is that really the case?

  3. aretzios

    Interesting report!


    However, I am wondering why any of this is necessary at all. Who would even buy these machines? Why are they even being produced? It is crazy!! It seems to me that this is an experiment in which the users should be paid to use these computers, not the other way around.


    I understand the Microsoft need to have Windows for various chip platforms. OK, but it should come out with a product that works on par with the Intel platform. It should stay in the lab until it achieves this and it should not be hoisted to consumers!!

  4. wright_is

    In reply to ianbetteridge:

    Leo Laporte gave a whole list of macOS applications he regularly uses, which don't work at all. One was emacs and another was a development environment that he wanted to use for the month of code competition, he said he ended up using his Linux powered Dell laptop instead.

  5. SYNERDATA

    Fortunately, AMD has reportedly indicated they plan to dust off their ARM license and make a new ARM PC chipset to compete directly with Apple Silicon, including putting the RAM right on the ARM64 PC SoC, and so hopefully we may yet see an SPX2 with a 12 core AMD ARM64 PC SoC which is then competitive with M1 and M1X.


    We need 8 performance cores and 4 efficiency cores, Radeon GPU and on-SoC RAM for Windows on ARM to compete with MacOS on ARM and Windows on M1.


    I put buying an SPX1 (SQ2) on hold.

  6. behindmyscreen

    The only way Microsoft can get off x86 is if they stop allowing windows to install on x86 chips.

  7. anderb

    It would have been nice to include an observation on the hit to the battery by running the emulated x64 apps. After all, if x64 emulation hammers the battery then what is the point of buying an ARM-based laptop again?

  8. navarac

    Just reading the review I can almost feel the slowness! It definitely seems NOT fit for purpose yet. Certainly not something I would invest in for some considerable time yet.

  9. hlbuck

    About third party drivers - what is the underlying reason for their being incompatible with WOA? Is there any possibility of Microsoft addressing this?

    • wright_is

      In reply to hlbuck:

      The drivers are only available in x64 code and the drivers have to be native. The x32 and x64 interpretation layer runs on top of the Kernel and drivers.

      Drivers take a lot of development and testing and it isn't as easy to take a driver and re-write it for WoA as it is to take an application, especially one written in .Net or UWP. That is a big investment that the numbers just don't make worthwhile at the moment.

      Heck, our laboratory equipment suppliers want over $100,000 per device in some cases to move from Windows 7 to Windows 10 compatible drivers and software! What do you think the chances are of them re-writing them for ARM?

  10. joloriquelme

    Microsoft came late to the party, again. Probably a decade late. They are testing a technology that Apple already has finished, working and it's selling it, and that tech outperforms every other single actor in the industry.


    I am sorry, but Microsoft needs a miracle, again. A miracle that actually never happened in the past with their failed products. And it will not happen. Unless they want to be humilliated again, they need to silently kill this WOA and Surface X thing, and focus on x86 market.

  11. wright_is

    In reply to lvthunder:

    Here is a review of the M1 Macs, with the question, is it ready and should you buy?

    https://www.forbes.com/sites/patrickmoorhead/2020/11/21/apple-macbook-pro-13-m1-reviewwhy-you-might-want-to-pass/?sh=85c4789786a8


    This site lists known issues with Mac software, it shows whether there is an M1 optimized version and whether it runs without issues on M1. When new versions are released that are compatible, the list is updated to show that. There is an impressive number of applications that have since been made M1 compatible, but there are still many that don't work or have problems.

    https://isapplesiliconready.com/

  12. Scsekaran

    In reply to sammyg:

    You could see it from a different logic as well. With an iPhone, you will have most if not all the benefits of Apple ecosystem. With the addition of a Surface Pro or any other Windows system of your liking, you get the benefit of both worlds.

  13. Greg Green

    In reply to sammyg:

    I think this is why the Nadella plan, “I want people to love windows” didn’t work. For the cubicle class windows reminds them of work and few people love their work. The smartphone reminds them of social media, and many love that.


    And most gamers just want the os to get out of the way so they can play. Nadella’s plan had no chance at all.

  14. melinau

    The universal good reception for M1-based devices & their stunningly good performance highlights just how "Intel-like" & complacent Qualcomm seems to have become. Their latest & best effort to produce a "PC & Notebook" SoC is clearly significantly inferior to Apple's 1st attempt. If we believe what is being bruited around the Web, Apple is to release much more powerful SoC's next year. Without the computing horsepower available to Apple via M1 & successors, MS & Windows simply can't even start to compete properly.

    • wright_is

      In reply to melinau:

      The problem is, Qualcomm sees making a competitive laptop or desktop chip as a waste of time and money, as the RoI is much too small to bother about.

      For Apple, it is all-in on ARM.

    • mattbg

      In reply to melinau:

      One difference is that Apple is doing translation while Qualcomm/Microsoft is doing emulation.


      I wonder why Microsoft is not doing translation. Given the work they've done with Windows Subsystem for Linux and virtualization in general on the desktop and cloud side, you'd think they would be in a great position to do something like that.

  15. angusmatheson

    In reply to ianbetteridge:

    But why release it as a product that doesn’t work well. Why not get the bugs ironed out - like drivers, 64 bit apps, and performance and then release it. It is never going to get developer support if it doesn’t work well. And people who buy it aren’t going to be happy. (I’ve never used one, or seen one in use. Although I love the look of tut surface X)


    • tboggs13

      In reply to Angusmatheson:


      It of course depends on the device and the apps you run, but the Surface Pro X (SQ1) is my primary work computer and I love the device. I have it connected to a Surface Dock 2 and drive a 4K 43" second monitor day in and day out. Also, I have the 16gb model, which the extra memory may help with performance. I work in IT and don't use photo editing tools outside of what's built into windows.


      It was rockier at first, mostly due to Microsoft Teams. but now even that has a native Arm64 client. The only app I would like to run that I can't is our corporate VPN solution, but there are alternative arm native clients.


      So let's answer your question. "Why not get the bugs ironed out - like drivers, 64 bit apps, and performance and then release it."


      Microsoft included tons of drivers in the box, the ones they can. I can think of only two things I would like that I don't have a driver for, the VPN and BlyncLight (It technically works, but I have issues with it when it's running). But there are millions of devices out there and it's up to the OEM's to write the fully featured drivers. They will never write drivers for a device that doesn't exist. It's the same Chicken or the Egg problem that Microsoft has had with every new platform.


      Performance to me is fine. I upgraded from a Surface Pro (5) w/LTE and this performs better. I also own a Surface Pro 7 (i7) for personal use and with the apps I use, I don't notice a difference when I am using it all day long and switching between the two. Yes, the SP7 is a bit snappier, but an i7 should also be snappier than an i5.


      To be fair, I bought the SP7 after the SPX. I really wanted an SPX for personal use also, I like it that much. But compatibility is a valid concern. I do work on the side that puts me in touch with some really old legacy software and I couldn't run the risk of not being able to run a utility when I needed to. If I did not have that particular use case, it would have been SPX all the way. Which is maybe Paul's point, but my point that for a lot of people it's not an issue in this day and age. People are perfectly happy with ChromeOS and I see WOA as beeing more fully capable than that.




  16. Martin Sjöholm

    In reply to ianbetteridge:

    I can concur. I have the same; SPX + Mac Mini M1. Same use case, same experience. The only application that is slow is 1Password. Yes, there are a few applications I would have preferred running on the SPX but I can easily switch over to another computer do to those things.

    • jatinder37

      In reply to RoundaboutSkid:

      May I please ask what apps you run on your SPX. I have a SP PRO 7 and I use the following:


      Netflix

      Prime video

      Office 365

      Nord vpn

      Nord pass

      MS Teams


      Would these work on a SPX

      • Scsekaran

        In reply to jatinder37:

        Apart from Nord VPN and Nord Pass everything else is ARM native. Office 365 is a Hybrid 32bit x86 /ARM application as explained by SWCetacean below which works well including extensions.


        I do use Nord VPN but I don't use Nord VPN app (I don't think it works). I configure built-in Windows 10 VPN client with IKEv2/IPsec protocol using instructions from Nord VPN website. Other options includes Nord VPN browser extension or Viscosity application, a paid ARM64 native app for configuration with any VPN service.


        I don't use Nord pass so I can't comment on that. Browser extension may be an option




  17. madthinus

    The real issue is that fundamentally, Windows itself needs to be reviewed and streamlined to make this switch really work and fly. That unspoken portion of the Mac transition is the re-engineering of MacOS itself over multiple years to make it viable today. That includes subsystems like Metal that is build for direct access to low level system. We don't have a unified OS level interface yet for menus, let alone for all apps to access.

  18. angusmatheson

    In reply to ianbetteridge:

    Thanks. I admit I love the look of the surface pro X. Out of curiosity, why did you get a surface pro X over a iPad Pro? Runs office and web fine. And would give you easier flow with your MacBook Pro 16

  19. rmlounsbury

    I do hope that Microsoft is able to improve overall performance of emulated apps on WOA over time (but not take 3 years to do so). I do have a Surface Pro X and the device itself is absolutely fantastic from the keyboard and pen to the larger screen and thing fanless design.


    I have no delusions that I'll ever do more than basic tasks such as light code editing in VS Code, PowerShell, Windows Terminal, Office, and hopefully some image editing via Elements/Affinity. But it would be nice if those emulated apps performance is close to that of the native applications. At least, I assume since Microsoft collaborated with Qualcomm on the SQ1 chip it was optimized for WOA -- maybe not as good as Apple's M1 chip since that was 100% designed in house for macOS.


    If nothing else, having serviceable 32 & 64 bit emulation and a decent library of native WOA is nice to have. With the potential of also getting Android apps running on WOA that should fill any leftover gaps.

    • wright_is

      In reply to rmlounsbury:

      The problem is that WoA has been a minor distraction for Qualcomm so far. It is very low numbers that really investing in optimising for the laptop market, let alone the desktop just isn't worth it. They optimize what they already have for smartphones and tablets, but actually take the market seriously, like Apple has? The market is too small for them, at the moment, for that sort of investment. They don't look long term, if they can't turn a profit in the next 6 months, it isn't worth investing in. Making an optimized laptop or desktop chip is something that will take years of R&D to pull off and they don't know if the market is worth it.

      Maybe, if apple pulls it off, they will rethink their strategy.

  20. Jorge Garcia

    Honest question for a software engineer - Is this emulation layer something that can be periodically swapped out with new and improved models, or do they reach a "point of no return" where they are "stuck" with the decisions that were made previously (and must keep polishing that same framework as much as they can)? For example, if they figure how to copy the (presumably more efficient) way that Rosetta 2 handles emulation, could they, if necessary, scrap their current emulation efforts and transparently roll out an entirely new one? My guess is yes, they could, but I have no clue.

    • wmiller

      In reply to JG1170:

      Rosetta 2 isn’t emulation, it’s translation.


      There is little that could be taken from this and applied to such an approach.

      • Jorge Garcia

        In reply to wmiller:


        Thanks, but why couldn't Microsoft develop a similar translation mechanism?

        • SWCetacean

          In reply to JG1170:

          Windows has a similar-ish mechanism called the XTACache service. It caches Arm translations of x86 code so that the next time an application is run, the emulator can recognize that code and load the cached Arm code. It's similar to how Javascript engines in browsers work: interpret code on the first pass, then compile better code for commonly-used code segments. From what I've heard of Rosetta, it seems to do a full translation of each application at install-time. Considering that Windows already does emulation, and caches the output of the emulator, it should be entirely possible for Microsoft to develop a full-program translator. It's probably more difficult and resource-intensive to do full-app translation though; from what I recall from my compiler courses, it's easier to optimize and ensure correctness in a small chunk of code at runtime vs doing the same on an entire program without running it. So then it's a question of how much effort Microsoft wants to put into developing a translator, vs developing native Arm64 developer tools and convincing people to use them.


          I'll ask my wife about the emulator/translator difficulty gap. She's doing her PhD in static code analysis and compiler-based correctness checking.

          • Scsekaran

            In reply to SWCetacean:

            Does Microsoft Office on ARM devices work in this way?

            • SWCetacean

              In reply to Scsekaran:

              Microsoft Office on Arm works using hybrid executables, which are really interesting in their own right. They basically use Arm code that uses the same calling conventions as x86 code. Calling conventions are the ways in which one piece of code uses another piece of code. So x86 code calls what looks like x86 code, but that "x86 code" is really Arm code. That's not the same as emulation.

                • SWCetacean

                  In reply to Scsekaran:

                  So I discussed the difficulties of translation vs emulation with my wife, and we both thought that if you have an emulator, then you should be able to perform a naive translation of an entire program. A smart translator would be able to do optimization on the full program to obtain better performance than you would get with a naive translation. Thus it's just a matter of how much time/effort you want to put into expanding the emulator into a full-blown smart translator. That's possibly one reason for better non-native performance of Mac apps on M1 vs Windows apps on Arm.


                  One potential pitfall of ahead-of-time translation is in the case of differences in memory model. Arm has a more relaxed memory model than x86/x64 (more relaxed meaning that the CPU can re-order more types of memory operations to increase performance) which can cause problems with x86 programs that have implicit assumptions about the order in which memory operations occur. That's why in Windows compatibility settings on WoA, there are additional options for multi-threading that allow the Arm processor to mimic a stricter memory model for specific programs that need it. Ensuring that the program logic's assumptions are maintained is something that is complicated to do statically, and even Apple's Rosetta 2 is not pure translation; it still has an emulator that is used at runtime for code segments that couldn't be translated ahead of time.


                  Finally, we thought that a big reason why Apple could aggressively pursue translation as an option is due to more uniform application install/distribution mechanisms. Most modern Mac apps are cleanly packaged and installation is as simple as dragging the package into the Applications directory. There is a standardized mechanism that macOS uses to install programs and most apps use that mechanism. Rosetta 2 can translate apps installed this way at install-time rather than at first-run. On Windows, there is a long history of 3rd-party custom installers that use custom install logic and put their files every which way on a system. So our guess is that rather than writing new logic to chase down every dependency and library for every application at install time, the Windows engineers decided to use the combination of runtime translation + caching to offload that process of finding dependencies and libraries to the application logic that already exists. That incurs a runtime cost, but simplifies the deployment/management cost.

                • Scsekaran

                  In reply to SWCetacean:

                  That's great info. Thanks again

        • bkkcanuck

          In reply to JG1170:

          Probably depends on whether the Windows for ARM has exactly the same APIs that are available in Windows 10.

        • jfgordon

          In reply to JG1170:

          My guess is: For the same reasons that we do not have tabs in File Explorer, or that icons in Windows are a mixture of styles going back 25 years – i.e., lack of a compelling business case for doing it.

          • Jorge Garcia

            In reply to jfgordon:

            I think the incentive will come form HP and Dell crying about losing an ever-increasing percentage of the consumer desktop/laptop/non-pro video editor pie to Apple. If someone has $700 to spend on a computer, and they don't care about AAA gaming, there is almost no compelling reason to stop them from buying a Mac Mini with M1.

          • winner

            In reply to jfgordon:

            Lack of architectural focus, of quality focus. A huge corporation with no visionary, and no compelling push to finish and polish the job.

            • angusmatheson

              In reply to Winner:

              Apple is reaping the benefits of slow incremental change to allow this to happen. 64 bit ARM, then abandoning 32 bit apps on ARM then on x86. This felt small and user and developer unfriendly at the time. But set up this uniformity that allowed this transition to be so seamless. Also swift, catalyst also moved things toward unity for years. Windows instead has jumped from one ARM platform - Windows phone 7, Windows phone 8 and Windows RT, now Windows on ARM - and maybe now there are working on Windows 10 X on ARM. Progress takes time. Slow plodding work to make something better. Each reinvention of Microsoft’s ARM strategy prevents any evolutionary progress to be made on the old one that could make it better. Yes Windows 10 core allowed one common core to be run anywhere. That is in fact unity. But I don’t see how Windows is reaping the benefits of that work. (Maybe in Xbox and PC gaming where unified x86 make games easier to share. But PS4/5 seems to be doing just as well without that. And PC gaming is still quite small).

    • Aaron44126

      In reply to JG1170:

      They could swap out one method of emulation for another if they so chose. But they would have to be very careful doing so, as they could cause regressions — apps that worked with the old method that are now broken because of something that isn't quite right with the new one.


      I am hoping that Apple's work with Rosetta 2 gives Microsoft a sort of kick in the pants to improve their x86/x64 emulation on ARM systems. Apple's works by basically decoding the application code of the x64 binary and generating equivalent ARM instructions to run natively — this is "dynamic recompilation", not really straight interpreter-based emulation, and is generally regarded as the best way to handle emulation if you are shooting for best performance (it is also more complicated to implement).

  21. christophercollins

    Until Qualcomm makes a chip for Microsoft that has the same type of optimizations as the M1, the results on this will be horrid. This has to be done in the hardware level to be both fast and power efficient.


    Windows is my primary OS, but I do have an M1 Macbook and once it runs through Rosetta the first time, it is every bit as fast as the same app on an Intel Mac, at least all my apps are.


    I'd imagine Qualcomm with take some lessons from Apple here and make a much better chip next time around.


    • peterc

      In reply to ChristopherCollins:

      >>I'd imagine Qualcomm with take some lessons from Apple here and make a much better chip next time around.


      I reckon if they had a guarantee of enough chip orders they would, but can they get that size order? If all windows OEM and MS place a combined order for a say 5 year supply deal with 5 chip enhancements they might.

  22. ecumenical

    Huh, cool. Better than I expected. Hopefully Qualcomm can get cracking on some performance improvements (or MS can do some heavier semi-custom/custom stuff that better fits the PC space).

  23. straker135

    Thanks Paul, this is interesting and I am tempted to try on my Surface Pro X, which I love from a hardware design perspective. But I am cautious as I have experienced flakey typing cover connection issues with Insider Builds that became so frustrating I had to Reset.


    Thinking about the issue of x64 more broadly in my use case I have not found anything that I really need that I am missing so far in day to day use on the WOA ProX. The one exception is the Canon printer apps for my multifunction printer, having said which the Microsoft defaults are good enough. We even have Team in native ARM64 now. Not a lot I guess, but I don't use Adobe products unless I have to, I usually do light image manipulation on Paint.Net, even creating icons for fun.


    You've listed Photoshop but what other specific apps do you use in your workflow that need x64?


    All the best

    • Paul Thurrott

      I'm not sure I've ever made a list per se. The graphics app is the big one, and I very much prefer Affinity over Elements (and very much prefer both over Paint.NET, which I had been using previously on WOA). The nice thing about normal Windows is you install it and it just works. The problem with WOA is that things are working OK, compatibility-wise, and then one day something won't even install and a regular user wouldn't understand.
  24. briantlewis

    Having received an M1 mini to replace my i5 version you're not going to be disappointed. Chrome is so fast on this thing it's silly. Translated apps feel faster than they were native on the i5. Only anomaly I've noticed is when translating the creative cloud apps launched twice during that process. Super impressed thus far.

  25. cseafous

    Sounds like they are on the right track. I am glad to hear that since I have been wanting the Surface Pro X for some time. However, it sounds like the issues you covered won't be completely worked out until the third for fourth generation of the Surface Pro X. My Surface Pro (5) is still going strong with no problems (other than decreasing battery life), so I am in no hurry.

  26. angusmatheson

    I find it fascinating that all the M1 Mac reviewers are M1 Mac vs Intel Mac. I haven’t seen M1 Mac vs Windows on ARM PC (like e surface pro X). I think this says something about how siloed users and reviews are - and how unpopular WOA computers are. I think people who buy Macs, buy Macs and people who buy PCs buy PCs. So if you are going to review a M1 Mac it is unlikely that you have a WOA or even an intel PC lying around to test. And if you have those lying around, you probably didn’t rush out to get an M1 Mac. This is a little about maintaining your workflow and purchased software, but I think it is more fundamentally tribalism. Admittedly I live in a Web app, but it seems to me using anything Windows, MacOS, iOS, and all flavors of Linux - are all basically the same (admittedly I have neither an M1 Mac nor a WOA to see how different they are - and like most of us I’m unlikely to need to buy either). That tribalism (and the conservatism of workplace procurement departments) I think will keep most users and business using Windows for years to come.

    • wright_is

      In reply to Angusmatheson:

      Given that the WoA devices are woefully underpowered and can't compare with an Intel powered laptop and the M1 is in many instances faster than Intel processors, it isn't really surprising they aren't being compared head to head.

      I agree, a majority of Mac users will buy new Macs and a majority of Windows users with buy new Windows PCs. But it is still a huge shake up of the processor market, even if the processors are limited to Apple devices.

      With more and more software working on both platforms and a lot of it being subscription based, the investment in software isn't the problem it used to be, at least for mainstream packages. The photo cataloging and editng software I use is available for Windows and macOS and I can switch back and forth at will, I don't need to buy a new license - as long as I only use it on one device at a time. The same for MS Office.

      That said, at work a majority of the software we use is Windows only, so there isn't an easy path to switch - other than booting into macOS and then using RDP to log onto a terminal server to run the Windows software - although some of the stuff, like CTI software needs to run locally and is Windows only.

      So, I think, for creative professionals and for many home users, it is easier to swap back and forth. For business users, especially for those working at companies that have a long history of software and automation, the move is much harder. Most of the ERP platforms I've dealt with in the last 20 years are Windows only, at least on the front end, for example.

  27. ponsaelius

    Windows on Arm (WOA) is really Windows on Qualcomm (WOQ). Whereas the M1 is Apple Silicon or MACOS on Apple.


    The issue is that Apple are saying this is our primary chip now. Intel is not dead but developers need to get onboard. This is it. We have a chip set for the 2020s. Long battery life, good performance, and very soon anything you run on a Mac will just work. We even have "universal apps" that work on IOS and MacOS.


    In fact, they have delivered on the promise of Windows 8, Windows 10, integration with mobile.


    How all this will work out in terms of sales I don't know. Some people who are now very much mobile workers, who need Office applications might look at Apple PCs. Microsoft is now a business service company so running Office 365 from a Mac is still getting them a monthly subscription.


    Business is slower moving so businesses may note the Apple M1 but will have no plans to change. DevOps in those businesses may not need Windows. So, they may decide the advantage of performance and battery life is good news.


    I think that the clunky Windows relationship with low performance Snapdragon chips will be unfavourably compared to Apple M1. In the future, M2 or M3. The battle between running "real Windows" and "arm Windows" is going to make compatibility and application viability a sticky issue. Without software support then SnapDragon Windows will be forever in emulation, containers or virtualization. Other Microsoft projects have failed due to lack of software support. I am looking at Windows RT, Windows 10S and Windows 10 Mobile.


    I think that if Microsoft want Windows on Qualcomm to happen the SOCs have to be better and Microsoft have to decide the future of Windows is on ARM. Otherwise, at best it's a niche purchase and at worst it's another interesting, failed project.



    • jackwagon

      In reply to ponsaelius:

      I've heard of a concept called "Microsoft's Law": basically that Microsoft can be relied upon to show a vision of the future, but that the "realization" of their vision is generally made by someone else. A lot of this, I'd imagine, is that they have too much invested in the older way of doing things to make any kind of meaningful break from legacy. It's great for their quarterly earnings, but who's to say how it will affect their future?

  28. winner

    Thanks for the report, Paul.

    Perhaps Apple's apparent home run will light a fire under Microsoft and we may see progress faster than the glacial pace since Windows RT. Of course we still have to deal with the massively higher-performing Apple CPUs.

    • wright_is

      In reply to Winner:

      The problem isn't so much Microsoft as Qualcomm and users. Microsoft's biggest market is the enterprise and there is little incentive to move to ARM on Windows, when all of the software and peripherals are x64 only and the suppliers won't re-write their code until there is a groundswell of people moving to ARM.

      Qualcomm are looking at the market and seeing sales in the sub-million category, I would think, at the moment. That isn't a market that is worth investing much time and money in, it is a distraction.

      Apple, on the other hand, has a more captive market and has always pushed things forward and the developers and the users either move with Apple or they can clear off, Apple aren't interested in them. And enough of the Apple users are willing to keep moving along with Apple that the developers will move with them. There aren't as many entrenched enterprise users stuck on old software that can't move to the new platform and there isn't much software that doesn't move.

      Windows is a totally different kettle of fish. A lot of LoB software is sold in low numbers and there just isn't the incentive to re-write it for a new platform and if they do, you will have to completely re-purchase the whole kit and kaboodle and extortionate prices.

      If moving a single machine from Windows 7 to Windows 10 is going to cost you $100,000 in new licenses, you are going to think twice about moving to Windows 10 on that device. The same will be the case, when moving from Intel to ARM. If it still runs on the Intel processor, why would you invest over $100,000 to move that one PC to ARM?

      And if you have a fleet of thousands of PCs, you want them all running the same base image of Windows, so that if a PC breaks, you can just swap it out. If you suddenly have a mix of ARM and Intel machines, that makes support that much more difficult. So as long as a significant number of machines are stuck on Intel, all PCs in the organisation will be stuck on Intel, unless there is a significant cost saving in moving to ARM.

    • shark47

      In reply to Winner:

      Which should be fine if they get close. Right now, it's nowhere close. The Qualcomm 8cx chipset is the biggest hurdle.

  29. davidl

    What have you done with Paul?

  30. codymesh

    Apple added specific x86 to ARM translation abilities to the M1 at the silicon level. There is no way Microsoft can match Apple's performance until Microsoft+Qualcomm massively deepen their collaboration - and even then, it will be legally complicated and the results will likely take years to bear fruit. I think Microsoft's best bet is to make compiling native ARM apps as easy as possible for developers.

Leave a Reply