Yet Another Wildly Speculative Windows on ARM Thread

curtisspendlove
Nov 13, 2020
37

So…

The Apple M1 Mac event got me thinking through my tech stack again.

And it makes me wonder if MS just postponed their ARM strategy or mostly abandoned it.

Apple was pretty evasive about the capabilities (or significant lack thereof) in regard to hypervisors on the new platform.

They demoed (I think it was Parallels) running some Linux VM in the WWDC presentation and made a pretty big deal that “it’ll work for developers, we love developers”. But they have been very cagey in interviews and such.

Regardless the presentation this week was all about the consumer line of hardware. The devices upgraded were largely consumer level devices (and the lowest end of the MacBook Pro). So I guess it reaches a *bit* into “prosumer” as well.

But there’s a lot of devices to go, including the really beefy systems: iMac Pro and Mac Pro.

I don’t think Apple really cares too much about Boot Camp / Windows. I’m sure they have metrics on it and know how often it is used.

That said, I doubt they’d be too sad to be able to announce “Boot Camp for M-Series” if MS put the effort into it.

I don’t even know if it’s worth it to Microsoft though either. I’m guessing that they *do* have metrics on how many copies of Windows are running on Apple’s CPUs.

One trick with ARM is that it isn’t always quite as simple to support multiple vendors’ chips as it is with x86/64. I don’t know the nitty-gritty details but I know it isn’t a “it just works” scenario and some person-hours are spent for each desired vendor platform.

THAT said, I was thinking through some things wondering if ARM is the bifurcation line for “legacy Windows” and “modern Windows”.

I’m guessing the age-old “legacy” argument still holds true though. Many will suggest people only want Windows so they can run some ancient cross-stitching app to help with their hobby.

Gaming is, of course, a large issue. So I’m not sure what the solution is there.

The huge one is something Paul’s mentioned in the past: drivers. If you can’t print on the new laptop you just bought, that *is* going to be an issue effecting lots of people. (Same applies to xyz peripheral, etc.)

But I think there is a market for it. But I may be severely overestimating the number of people who could do a “clean break” and not worry about the legacy stuff.

But then, again, the majority of those people are probably using a phone or tablet as their primary computer now. And only using their traditional PC for a couple tasks that mobile devices and small screens aren’t optimal for.

:: shrug ::

Conversation 37 comments

dftf

13 November, 2020 - 1:35 pm

Okay, so regarding Windows running on a mac: Virtualization software: I'm sure both VMWare Fusion and Parallels Desktop will both see an Apple ARM release. And I cannot see why "Windows 10 on ARM" would not be a supported guest (seeing as performance would be better, assuming you only need to run ARM-apps inside the Windows session: if you need to run x86 or AMD64 apps, you might be better-off with the AMD64 version of Windows (as I'm sure it'd be faster to let the VM software do the AMD64 -> ARM conversion, not the internal Windows 10 on ARM session). Boot Camp: only Apple knows there.  Supporting the regular 32-bit or 64-bit versions of Windows would be a no-no I'd expect (their new CPU is 64-bit only, so 32-bit Windows is immediately out; and the AMD64 version would require some sort of hardware-level support to make it work), but can't see why the ARM version of Windows 10 could not easily be supported. And looking at just Microsoft and ARM: I'm sure "Windows 10 on ARM" will see an uptick in use once it is able to run classic AMD64 apps, and not just x86 apps. That does rather limit things at the moment, as many companies (such as Adobe and Autodesk) are no-longer releasing new versions of their apps in x86, only AMD64. (Not to mention virtually all new PC games require 64-bit, as they typically need 8-16GB of RAM). Though, I do wonder: security-aside (e.g. ARM not being as-affected by Spectre and Meltdown as Intel was), the only real benefit to end-users of "Windows 10 on ARM" is increased battery-life.  But AMD are now down to, what, 5nm CPUs, which will also bring battery-life increases. So, with that in-mind, does ARM then have any-other advantages (other than making it easier to port Android apps to Windows, perhaps) over ever-increasingly more efficient AMD and Intel chips?

Log in to Reply

angusmatheson

15 November, 2020 - 10:52 am

<blockquote><a href="#592736">In reply to dftf:</a></blockquote>All this talk about wanting to run Windows on ARM macs. But no one wants to run ARM windows computers at all. No one in this thread has said they use an ARM windows PC at all, to say the least for any development or more professional use. For me, that is telling. So far windows has failed with a transition to ARM. Windows 8’s Windows RT and now Windows on ARM with the surface X – seemed to have bombed. Why can’t Microsoft pull this off? And why do they attempt, have problems and then abandon it. Why not find a solution and evolve it? This iteration of MacOS on arm won’t be perfect, but I bet next years will be better. And the year after better still. I think that is why I’m puzzled by all the talk about the M1 chip. ARMchips are fine. They work great in lots of Android and iOS tablets. The question here is can MacOS on ARM work well? 

Log in to Reply
sevenacids

16 November, 2020 - 5:40 am

<blockquote><a href="#592736">In reply to dftf:</a></blockquote> "…does ARM then have any-other advantages (other than making it easier to port Android apps to Windows, perhaps) over ever-increasingly more efficient AMD and Intel chips?" It all depends on: How much advantage does ARM have when it comes to general purpose, multi-threaded, high performance applications? The focus is on general purpose. Because, for example, if you look at the Apple SoC, it seems to be architected and optimized for consumer aspects such as video processing and certain machine learning tasks. But what about compiling a huge code base, for example? What about the graphics cores and computing-intensive applications, such as OpenCL kernels? Apple silicon doesn't seem to be made for such tasks. Then the problem with high performance is always: It's based on computing power. Computing power requires energy and as with all electronics, this will inevitable lead to heat dissipation. A modern x86/AMD64 CPU has a energy density per area that generates more heat than a fuel rod in a nuclear power plant. So how much of a difference will ARM have here over x86/AMD64? Especially when it comes to AMD chips because they already have a better multi-threading performance compared to Intel ones. And then there are compilers and how good they are at generating highly optimized code. IMHO, a lot about ARM is just hype and x86/x64 is in many aspects not as bad or ancient as we are made to believe from many sides.

Log in to Reply

curtisspendlove

16 November, 2020 - 3:50 pm

<blockquote><a href="#593283">In reply to sevenacids:</a></blockquote>And then there are compilers and how good they are at generating highly optimized code. IMHO, a lot about ARM is just hype and x86/x64 is in many aspects not as bad or ancient as we are made to believe from many sides. Agreed. I’m an Intel fanboy. So I have no desire or plan to throw my Intel systems into the nearest lake or anything. (In fact I have a new, very expensive, gaming / video editing rig planned for next year.) But I love technology as a whole and this type of thing fascinates me. The fact that the fastest supercomputer in the world is currently ARM is interesting. I’m sure if Intel or AMD wanted to beat it they could. But then the ARM team could just come right back and beat it again … ad-infinitum. 

Log in to Reply

bkkcanuck

13 November, 2020 - 5:32 pm

The Macbook Pro released is not prosumer but slightly higher end home consumer mostly (there is a very big difference between that lowest end Macbook Pro and the rest of the Macbook Pro line. As far as I have figured out, the ARM Instruction Set ISA for a particular version of it – is standard and all licensees MUST implement it fully. There has been some discussion of allowing custom instruction extensions to the core processor but that change has not been made. With Apple (and probably others) there are more specialized processors integrated onto the SoC package (i.e. the Neural Engine and GPU for example) that are not part of the main processor and thus differ, but then you have that now in a certain way where you have different driver for different motherboard components. There won't be bootcamp again, but I could see Microsoft releasing effectively a hypervisor version of it that sits either in another VM or potentially their own container of sorts (only a version that runs on ARM without legacy support). I think more likely though that it will be done through CloudPC. I think the best early approach for Microsoft is to work towards removing legacy from the main operating system but having compatibility containers as necessary. If they don't figure out how to make a break — sooner or later they will legacy themselves into completely legacy use only.

Log in to Reply
longhorn

13 November, 2020 - 8:15 pm

I don't think ARM matters for Windows and Linux. The only reason ARM matters for macOS is because Apple makes their own SoCs. And Apple SoCs seem to be good. Let's say you gain some energy efficiency with a Qualcomm SoC, but then AMD will likely run circles around that chip. It's more powerful and more compatible. ARM Macs are not for Windows and virtualization. Virtualization might be possible, but with a performance penalty because of instruction translation. ARM Macs moved closer to iPads and iPhones and away from Windows and Linux, simply because of architecture. There is no point running Windows on ARM if you need Windows since most likely you need a Windows x64 program with full compatibility and performance. The good news is that macOS should be an adequate Web dev platform so Linux isn't really needed. Bottom line: You don't buy an ARM Mac if you need Windows or Linux. Then you buy a PC. 

Log in to Reply

curtisspendlove

13 November, 2020 - 11:14 pm

<blockquote><a href="#592856">In reply to longhorn:</a></blockquote>I think we will see a Linux story at WWDC 2021. They’ll always lean on their own toolset more heavily, of course, but I think they are aware that they have to compete with WSL. Windows, at the moment, is a far more effective web dev platform than it ever has been. I agree you don’t buy a Mac for Windows (you probably never really should have, but it was a best of one world and a foot in the water of the other world situation). But like the new Microsoft, Apple knows that the cloud runs on Linux. VMs, hypervisors, or containers…it has to be there for the cloud dev crowd (it is more than just “web devs” now). 

Log in to Reply

longhorn

14 November, 2020 - 2:33 pm

<blockquote><a href="#592889">In reply to curtisspendlove:</a></blockquote>I don't think Apple has to compete with WSL. That's not the focus for Apple SoCs. These are more consumer focused, bringing iOS apps to Mac devices. It's the Mac, macOS and iOS that are important. When you move away from industry standard (x86) in desktop (more than 90 %) and cloud (almost 90 %), then focus will be different. It will be less about common workloads and more focus on Apple unless your work is remote. Linus Thorvalds said that x86 will dominate the cloud for the foreseeable future, because devs don't feel comfortable putting their x86 workloads in an ARM-based cloud. So Web/Cloud devs are prepared to pay more for x86 cloud hosting. While I may be too negative about these ARM-based Macs, it's important to see what they are: They are a bet on consumers and digital artists and a move away from industry standard workloads. They are new "PowerPC" Macs and the Intel era is over. But for the target audience these ARM Macs may beat anything in the x86 world just as Apple presenters said. 

Log in to Reply
wright_is
Premium Member
16 November, 2020 - 4:51 am

<blockquote><a href="#592889">In reply to curtisspendlove:</a></blockquote>Microsoft only needs WSL to provide the *NIX tools for Windows developers. Apple has the advantage that it is a full *NIX to start with, so there is less need to have a Linux subsystem.I would see it more as Linux under Parallels/VMware or cloud instances moving forward.

Log in to Reply

curtisspendlove

16 November, 2020 - 3:35 pm

<blockquote><a href="#593276">In reply to wright_is:</a></blockquote>I would see it more as Linux under Parallels/VMware or cloud instances moving forward. Yup. I was thinking through this. And in the context of a post above (I think it was “longhorn”‘s). I hadn’t considered that to run on bare-metal (aka, Boot Camp or such) Apple would have to provide drivers / APIs / whatever for the SoC now. Not just x86/64 drivers for the trackpad and other integrated devices. A *far* bigger effort. It would make *way* more sense to provide an emulation layer or Hypervisor layer to support VMs and/or containers. And even more likely to just rely on their partners (as you mention Parallels/VMware/VirtalBox) to implement those layers. As time marches on, I think VMs and containers are far more advantageous than bare-metal anyway. I, personally, like bare metal for most things. But for dev workloads, testing, builds, etc containers are perfect. This is why Docker has become so popular. It’s nearly perfect for most use cases. 

Log in to Reply

curtisspendlove

13 November, 2020 - 11:12 pm

So one interesting thing I read is that apparently M1 firmware will not boot unsigned kernels, *BUT* they will have a developer tool that allows you to essentially sign a boot package…so in *theory* one might be able to sign a Linux or WoA image and make it bootable. The ball *may* be in MS’s court for that. 

Log in to Reply

longhorn

14 November, 2020 - 11:36 am

<blockquote><a href="#592888">In reply to curtisspendlove:</a></blockquote> I don't think Apple will ever provide drivers to let Windows and Linux run natively. Craig Federighi (interviewed by Marques Brownlee) more or less said that Boot Camp/other native OS option is a thing of the past. On the Linux side this is already a fact on T2 Macs. Since no T2 driver is available, internal storage can't be seen by the Linux kernel. Also, Microsoft and Linux kernel devs wouldn't know how to produce code running on Apple SoCs unless Apple gave them some specifics. While it's still ARM, getting your code to boot probably isn't trivial. It would be like producing a different OS image for the iPhone. Running OS images in the cloud (different PC) is probably the preferred solution. Running local VMs on a different architecture seems weird, but Parallels can probably pull it off if they want to, but I doubt as a viable product. Both Microsoft and Parallels likely saw this coming so now there is an option for Windows VMs on certain Chromebooks. In essence, Apple is becoming more Apple. 

Log in to Reply

matsan

14 November, 2020 - 3:04 am

My take on this is that Apple will succeed because they have made it "easy" for developers to tag along for the ride, be it 32/64 bit transitions or architecture. Just look at the transition to 64-bit that was completed with macOS 10.14.The Cocoa API has been the standard for 10+ years and when they got rid of Carbon API the transition was already done by the developers. I understand why since Apple burns all the bridges on their way forward so developers have to stay alert. The Microsoft way seems to be to slowly plow forward and leave zombie APIs behind in their tracks (UWA, WinForms, WinFX anyone?) . I cannot understand this strategy and have myself been burnt in projects based on .Net and the mess with versions it was in the 3.0-4.5 era.Looking at my own work as front/back-end PHP/JavaScript developer I'd say my next machine is very likely to be Apple Silicon if they can up the RAM to 32GB and there are no high emulation penalties for PHPstorm (JAVA-based). I'm currently on an Intel Mac MBP 16" with Docker, Chrome and PHPstorm IDE as my apps. Almost all of my other tools are installed using homebrew and some of my other tools are already getting updates to Universal App (BBEdit and DefaultFolder for example). The PC stuff I need I run through Remote Desktop to VMs hosted by AWS in the cloud. Boy, I'm ready! :-)

Log in to Reply

curtisspendlove

14 November, 2020 - 4:06 am

<blockquote><a href="#592926">In reply to matsan:</a></blockquote>Pretty close mirror to where I’m at. I’m curious to learn how well the 16GB models can handle a webdev workload. I know macOS handles things far differently than iOS, but I can’t help but hope RAM is handled more cleanly in the M1 SoC. I don’t run a huge amount of VMs locally, mostly just Docker containers nowadays. But I’m used to 32GB RAM *minimum* so I’m afraid these machines might not be for me yet either. That said, I expect the mid-level MacBooks and maybe the iMac line in the first half of 2021. So we will see where that goes. I’m also hoping for a beefed-up Mac Mini to come in that phase (I’m guessing the mid-level chip should support more I/O lanes and a larger (or separate) memory package. Ive never liked all-in-ones. Regardless, I don’t need Windows for much in my productivity life anymore. My windows workstation is basically a gaming rig nowadays. Or it gets booted into Ubuntu for some Linux lurve (I have a dream of one day submitting an accepted PR to the Ubuntu project). If I absolutely have to hit it from my Mac I can just RDP from anywhere as I have all my main systems on a nebula overlay (basically uses VPN tech and makes a flat network out of all your physical / cloud systems so they all behave as if they were part of the same LAN). But I’m *very* curious to see what the “vNext” is for Windows. They have to do *something* … or I guess they just let it ride with all the legacy baggage as long as it is still viable. :: shrug :: Honestly, I’m curious to see if they can manage to do anything on par with Apple’s silicon team. Technically, I can’t see how anything would stop them. So it just has to matter. And right now, I don’t think it does. But that makes me wonder why did they even try on the first place?

Log in to Reply

matsan

14 November, 2020 - 9:13 am

<blockquote><a href="#592928">In reply to curtisspendlove:</a></blockquote>My MBP 16" is 16 GB and it can handle all I throw at it when it comes to webdev, memory management in macOS is stellar. The few issues I have is with PHPstorm and I guess due to the JAVA environment. Buying a new machine in the close future (~ 4-6 months) I will most likely make it 32 GB.

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 4:19 am

<blockquote><a href="#592926">In reply to matsan:</a></blockquote>The problem isn't that Apple has made it so easy for developers to go to the new platform, it is the whip forcing them to move to the new platform as they burn all the bridges back to the old system.Microsoft offers its developers similar tools to move forward, but they don't burn those bridges, because their corporate customers have legacy software, that is worth more than a fleet of PCs, that has to keep running.If they abandoned those customers, they would stick with the old Windows they have, until they could move to a new platform (E.g. Linux). At that is a majority of Microsoft's customers. They can't afford to alienate, probably, over three quarters of their income source – if they abandon them on the PC, they'd probably take their server and cloud business elsewhere as well.

Log in to Reply

jimchamplin
Premium Member
14 November, 2020 - 10:11 am

Linux isn’t needed when Macports exists. It’s superior to Homebrew in almost every regard, including speed, security, and selection of packages. It also builds packages at install time, ensuring they’re optimized for the system you’re installing on. One can transform the terminal into a full BSD experience with Macports.

Log in to Reply

F4IL

15 November, 2020 - 12:57 pm

<blockquote><a href="#592985">In reply to jimchamplin:</a></blockquote>They need linux (i.e the kernel) for developers. They need to be able to distribute and deploy (docker) containers.

Log in to Reply

curtisspendlove

16 November, 2020 - 4:46 am

<blockquote><a href="#593172">In reply to F4IL:</a></blockquote>They need linux (i.e the kernel) for developers. They need to be able to distribute and deploy (docker) containers. This is one of my concerns. I haven’t had an M1 to test anything on. But I’ve read that Docker is currently non-functional on M1 Macs. This is unwise. There may be a workaround by now but many web developers rely on Docker and VMs are good for dev / testing as well. 

Log in to Reply
jimchamplin
Premium Member
16 November, 2020 - 9:58 pm

<blockquote><a href="#593172">In reply to F4IL:</a></blockquote>Aaaahh… I see what you mean. Yeah, that wouldn't work then :)

Log in to Reply

curtisspendlove

16 November, 2020 - 4:44 am

<blockquote><a href="#592985">In reply to jimchamplin:</a></blockquote>Linux isn’t needed when Macports exists. It’s superior to Homebrew in almost every regard, including speed, security, and selection of packages. It also builds packages at install time, ensuring they’re optimized for the system you’re installing on. Interesting. I haven’t checked out MacPorts for a long time. I didn’t realize it built things as native BSD binaries. Home brew has become more and more brittle over the years. 

Log in to Reply

jimchamplin
Premium Member
16 November, 2020 - 10:05 pm

<blockquote><a href="#593274">In reply to curtisspendlove:</a></blockquote>My understanding is it's how they dealt with the PPC > Intel transition. Since Tiger and Leopard both ran on PPC and IA32, just compiling at install time. Turns out it's great to make sure the bins are optimized for different Intel CPUs. Fair to assume that MacPorts won't be too bothered by the move to Apple M-series CPUs. And yeah, Homebrew I found a little too… un-UNIXy for my taste. Both in its design, and the sort of dev community. They view it as a Mac app and treat it as such, constantly ripping features out in the name of streamlining. MacPorts is simply a *nix-style package manager for Darwin.

Log in to Reply

codymesh

14 November, 2020 - 12:13 pm

Apple is able to move to ARM because their processor designs have enabled them to scale up performance. But for the rest of the industry…we have Qualcomm, and performance still remains pretty far behind Intel. As long as this remains the case, ARM for PCs will never be as good as Apple's.

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 4:14 am

<blockquote><a href="#593034">In reply to codymesh:</a></blockquote>Qualcomm could also scale up, if they wanted… But it is a niche that wouldn't bring any ROI for them, at least not in the short term. They would need to be in the hundreds of millions of units range to make it worthwhile to really invest in making a competitive processor. They are just tweaking their existing designs to minimize costs.Fujitsu has already shown what you can do in the Super Computer arena and there are already several suppliers that make datacentre server chipsets. The versatility is there, but Windows on ARM is too niche to warrant a proper investment from Qualcomm.

Log in to Reply
curtisspendlove

16 November, 2020 - 4:40 am

<blockquote><a href="#593034">In reply to codymesh:</a></blockquote>Yes. But that is part of my point. If Microsoft cared (or if it was a strong enough business case) they could throw money at that problem like Apple did. MS is not a poor company. They have plenty of revenue to throw around. This would be an awesome way to lead from the front of the pack. Which is kinda what Surface is all about. 

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 7:35 am

<blockquote><a href="#593273">In reply to curtisspendlove:</a></blockquote>The problem is, the WoA project was a wake-up call to Intel, it wasn't an all-in serious attempt to move. It was more, "get your arse in gear or else we have other alternatives."Now Apple is showing that ARM can actually be a real alternative, not a weak threat.

Log in to Reply

curtisspendlove

16 November, 2020 - 3:29 pm

<blockquote><a href="#593294">In reply to wright_is:</a></blockquote>The problem is, the WoA project was a wake-up call to Intel, it wasn't an all-in serious attempt to move. Hmmm. This makes sense. They were perturbed, maybe, but not outright burning-fire angry like Apple seems to have been. Interesting that they do have the tech. I wonder if Apple’s moves here re-ignited any fires in MS. I’m guessing not any serious ones. But the whole thing still intrigues me. The problem seems to be the same as usual: momentum and sunk-cost. 

Log in to Reply

angusmatheson

15 November, 2020 - 10:25 am

There has been all this talk about benchmarks and complaints about the Apple M1 event. But I don’t understand why none is the tech media analysis isn’t looking at windows on ARM. Windows on ARM has some very real problems as I i dersrand. Low power so it was slow. Emulation was a problem only 32 bit apps and power hungry so that distorted expected extended battery life. Drivers were a significant problem. iPads are fast enough for web browsing and light productivity tasks – and Apple still sells intel 13 inch MacBook pros. Macs only run 64 bit apps now – so the emulation story is at least there for all apps unlike windows how slow it is, and how power hungry is Rosetta II remains a question. I bet it won’t be as good as apple advertised. But unlike windows on ARM – all iPad and iPhone apps run natively. They is a huge pool of software. Now is it going to satisfy every pro workflow? No. But for regular uses I suspect the app questions will be fine. Drivers were a huge problem on windows on arm. Apple has all those drivers for Mac. I know windows has to wait foe their partners to update the drivers. I wonder if apple has a better sense of Mac drivers so that it can transform them all and since it is clear all Mac are going ARM manufacturers are willing to do the work, unlike windows which really does seem to have developed windows on ARM and now abandoned it. All this talk about M1 benchmarks seems crazy to me. Apple never gives those numbers. The question is can apple transition to ARM with a product that avoids Microsoft’s pitfalls and can (with this these products) apple satisfy basic users needs. That we will not know until they are in the hands of users. As my grandmother used to say, “The proof is the pudding is in the tasting.”

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 3:39 am

<blockquote><a href="#593154">In reply to Angusmatheson:</a></blockquote>Part of the problem is that it is a "side project" for Qualcomm, with limited sales, so they won't invest big money in making a proper ARM laptop or desktop chip, so Microsoft is limited to "lightly tweaked" SoCs.Apple, on the other hand is "all-in" on ARM.

Log in to Reply

curtisspendlove

16 November, 2020 - 4:34 am

<blockquote><a href="#593267">In reply to wright_is:</a></blockquote>Apple, on the other hand is "all-in" on ARM. This. And they have years of experience with this stuff from their A-series chips and all the work they have already put into iOS. For instance, they’ve said stuff like the image signal processing and ML that they bragged about to make the crappy webcam less crappy for Zoom calls was basically lifted from stuff they already had in iPhone. I’m expecting these new Macs to live up to the hype, at least mostly. I agree and I’m waiting to see though. But they have a *lot* to lose here. If they’re lying they’ll be called out on it. And they know that. The other thing is that you can look back in retrospect and see the chess moves of the past few years slowing inching toward this. As you said, stuff like obliterating 32 but apps. Submitting in intermediate layer bitcode to the App Stores that can be deterministicallly downloaded to an exact, specific processor architecture, etc. Apple also has a great developer story (for those developers they haven’t completely alienated due to their business practices). They sent out DTKs to many of the essential OSS projects (like home brew) and provided them technical contacts. They provided a low-cost option to get a DTK for smaller devs so they could get practice in and hopefully have their apps Universal on day 1. I feel Microsoft could do a lot of this. But one of their big problems is confidence. Will they actually follow through. Or will the developers have to worry this is just one more new tech that will be abandoned or superseded in a couple years. 

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 2:57 am

The problems are manifold, but there are a couple of key ones that need to be addressed:<ol><li>Drivers (and application), as you say. But more specifically, industrial drivers. If you have a peripheral that costs 6 – 7 figures (industrial plant management, CNC machines, laboratory equipment etc.), you aren't going to abandon that stuff, just because the controlling PC is now running ARM. Often the manufacturers do provide newer drivers for newer operating systems, but they are often tied to new hardware – one reason why a lot of laboratories and factories are still overflowing with Windows XP kit, the hardware lasts a couple of decades and you aren't going to invest 7 figures on new hardware, just because the new PC can't install the driver; you will find a PC that can run it.</li><li>As you say, ARM is not ARM. The instruction set is well defined, but the SoCs are each individual. That means the basics of a program are the same (i.e. calculations moving stuff in memory etc.) but actually controlling memory, the video interface, I/O interfaces etc. are all custom to each SoC and that isn't a "driver" level change, that is a fundamental change in how the code is implemented and run – one of the reasons why Android is such a mess.</li></ol>Movign WoQ (Windows on Qualcomm) to WoApple would be a major undertaking and would need a significant proportion of the code to be re-written, I would guess. That would be a mammoth undertaking and would probably need sales in the high 2 to low 3 million range to make it worthwhile.

Log in to Reply

curtisspendlove

16 November, 2020 - 4:20 am

<blockquote><a href="#593264">In reply to wright_is:</a></blockquote>Holy balls, Batman. I didn’t realize that some companies charge that much just for a driver. I mean it makes sense…the engineering cost has to go somewhere, but I assumed they’d be factored into hardware and not software. Or that those things would utilize their own custom peripheral stack. But drivers for an old version of Windows makes sense. It does make me recall stuff I’ve read before (some of it was probably here on Paul’s site and probably from you) 😉 Where organizations will run huge vertical market apps on an “unsupported” OS that is essentially air-gapped and LAN only so it isn’t immediately overwhelmed my malware. Very interesting. I know it would require extra engineering effort to have two versions of Windows, but would “legacy” and “modern” address that? Consumers, small and medium business, and parts of an enterprise (front office staff whose workloads have been converted to web-apps) could use “modern”. And enterprise compatibility stuff, including these crazy custom driver installations could run on “legacy”. I’m sure those companies would pay a heftier OS license cost for good support. (As you mentioned, cheaper to pay MS a $500 license fee for Windows Industry Edition and be done instead of having to figure out how to run legacy stuff on new hardware or find / build a custom “old hardware” PC control plane… :: shrug :: This is kinda fascinating to me. I love shoes like “How it’s Made”. Giant arrays of machines working together to produce stuff is insanely satisfying to me. P.S. The “interface via ARM” is interesting to me as well. I’m guessing these huge CNC machines and what not can’t just plug into a thunderbolt 4 port. 😉 I’m guessing they’d need older standards like Serial or Parallel ports. And probably a bunch of custom stuff layered up on top of that. Would it be semi-feasible to pick some sort of OSS based controller plane? Like the Raspberry Pi (assuming it’s powerful enough) or an OpenRISC SoC? Might be beneficial to those manufacturers that build and sell the machinery to build out a custom solution? Or would that much proprietary stuff scare off potential customers?

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 7:20 am

<blockquote><a href="#593271">In reply to curtisspendlove:</a></blockquote>Price? Not for the drivers, per se, but that the new drivers are bound to the refreshed hardware, so if you want new drivers, you need new hardware.Legacy and Modern? Yes and no. The problem is, if you are buying, say, 2,000 or 10,000 PCs, you want them all to be running the same image, it makes support and maintenance easier.2 versions of Windows? You need to train the staff on both versions. The Industry version might work.If the general chip makers can make the same sort of performance spring that Apple seems to have achieved, then I can see Microsoft having to move – and ARM is a lot more pleasant to code for than Intel, when you have to go low-level. Intel was always an abomination of an assembler language, M68K, Alpha, SPARC and MIPS were much better, but Intel managed to get in early with Microsoft and, as they say, the rest is history.Yes, they usually use serial ports, some use I/O ports (literally a card that can detect on/off plugged into a spare PCI (not Express) slot. A few use Ethernet, but still need custom software and drivers.Something other than Windows would be easily possible, in theory. But the developers know Windows and the corporates use Windows… All those devs who have been writing system level code in C/C++ on Windows will need to learn a whole new toolset and hardware. That takes time, then they need to re-develop the code for the new platform. That is man years of work that needs to be done, before you can start to sell anything and make money from it – and you need those staff to support the old system as well. I looked at it for a previous employer, "just" and ERP system, moving it over from an ancient Linux installation and a POSIX client under Windows, to a more modern system that would be web-based. We estimated it would take 90% of the development staff at least 3 years to complete, 10% would have to be held back for support and there would be no more customization or user change requests for those 3 years (about 70% of the work was customization on the old system, so 70% revenue gone, for 3 years), before you could start trying to sell the customers an update to a system that works as they want and to customize the new system to match the functionality of the old system…

Log in to Reply

Shehzad Joss Yousaf

17 November, 2020 - 1:57 pm

<blockquote><a href="#593292">In reply to wright_is:</a></blockquote>Not directly responding to your comments, but I just wanted to say thanks for posting with such a breadth of understanding, it's fascinating to read about the history and technical detail.

Log in to Reply

wright_is
Premium Member
16 November, 2020 - 5:01 am

<blockquote><a href="#593203">In reply to bkkcanuck:</a></blockquote>The problem is, Windows on Qualcomm is too niche for Qualcomm to really invest in it. They pimp up their existing mobile processors, but actually redesign them, like Apple has done, to optimize them for the higher TDP and higher performance of a laptop? It isn't worth the investment for the "pifflingly small" number of processors they could sell.They would need to sell hundreds of millions of processors to recoup the R&D costs.

Log in to Reply
curtisspendlove

16 November, 2020 - 3:44 pm

<blockquote><a href="#593278">In reply to bkkcanuck:</a></blockquote>I still don't think WSL has caught up to what I have on my Mac — getting better but still eh… Agreed for the dev-centric workload. But WSL is very solid for what it is meant to be … a bridge … or glue … maybe … that does just the right amount of what is needed to cleanly support Linux on Azure. I know there’s a lot of buzz on the whole “WSL is there so Microsoft can use it as the base kernel and write the Windows API on top of POSIX” thing. But I just don’t see that happening. Too much work for little or no gain. So to me it still seems like it’s just meant to be “enough for lockstep with Azure”. It still suffers from the “death by a thousand papercuts” issue for me. But the home brew toolchain has a lot more papercuts nowadays than in the past decade as well. 

Log in to Reply
winner

16 November, 2020 - 10:07 pm

Of course Apple has twice before made very successful shifts of processor architecture – first 680×0 to Power PC, then Power PC to Intel. They know what they're doing.Another trait I've noticed with Microsoft is that they often throw products out they're before they are really ready. Witness Zune (welcome to the social), Kin, Kinect, Windows RT. Apple OTOH, is normally much more polished before a product is announced, and they dive in with full effort. So I expect Apple to be very successful, and Microsoft not so much. This is going to be a really interesting few upcoming years in the marketplace.

Log in to Reply

Yet Another Wildly Speculative Windows on ARM Thread

Conversation 37 comments

My Surface Journey

Does this APU make sense? A possible PS5 to RTX 4060TI for laptop (NPU = 60 TOPS)

Web Apps vs. Store Apps

[CLOSED] Ask Paul for Friday, April 26

Windows Intelligence In Your Inbox

Sections

About Thurrott

Contact

Our Other Sites

Subscribe