Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Craig Federighi: Native Windows on M1 Macs is 'Really up to Microsoft'
- Thread starter MacRumors
- Start date
- Sort by reaction score
That is not a slam dunk for intel because there is always a huge gap between what intel promise on paper and what they offer. This is a big reason why Apple finally had to ditch intel. Intel have been behind ever since Skylake.In 2021 Intel will roll out Adler Lake ( probably gen12 ) CPU SoC with 4-8 big cores and 4-8 smaller cores. If they get the right updates weaved into Windows to do effective load balancing out of the small cores when the workload is low then that should substantively shrink the battery life gap substantially.
Light workloads , the target Adler lake goal can cut a Gen 11 (tigher Lake) consumption almost in half (40-50% range in some cases ). Chop down the video decode/encode and Zoom calls get longer without big primary processor cores being much different.
( Intel could pull it off since a major portion here is software and getting cooperation from Microsoft.... which has Intel SurfaceBooks. It isn't solely based on fab tech that is stuck. ) Are they going to match M1 (and M1-variants coming later in the year)? Probably not but all Intel has to do shrink the gap enough and point to "continuity" of the software stack for end users to retain a large fraction of the users. Most users aren't upgrading ( most users are generally on much longer upgrade cycles now so that also blunts the short term impact M1 Macs will have. )
It isn't a slam dunk for ARM. Pretty likely there will be come Cortex-X derivaties show up, but the C8X solutions now just aren't competitive if needs substantive "horsepower". Yeah get longer battery but somewhat "slow".
The other ARM implementations will have to deliver for ARM to go mainstream. So far that isn't happening. Qualcomm is dabbling at it, but not in with boot feet and full effort. Cortex-X1 will likely be better but still ( and a door that Microsoft could go through directly themselves ... pragmatically get a custom solution from ARM instead of Qualcomm. ) .
By your logic then MS would never have allowed Windows to run in virtualisation on any Mac then, but they do.
Actually it is up to MS because Apple M1 can run Windows, but the issue is the licensing rights which are totally and 100% only to do with MS, sorry!
MS ignoring Apple here would be shooting themselves in the foot. Windows for M1 will be available at some point. That is a promise.
Ok some things here. Microsoft is not going to provide support for M series mac's for you. They will be doing it for their corporate customers. In fact, it might only be available to volume license customers. At least for a while. But make no mistake, Microsoft is investing heavily in ARM. They have ported Windows, and most certainly the entire windows family to arm. They have hired CPU designers and worked with Qualcomm to develop two CPUs. They are porting Edge (that's chrome edge) to arm. And they have begun the work to port the Office apps. Most people don't know it, but they are the largest supplier of development tools. They have added support for native arm apps to Visual Studio and have begun a marketing push to those customers to make their apps native arm. Microsoft is all in on ARM.
One thing that has held them back is a lack of systems that have more of a desktop CPU that is capable of running Windows in the manner that we are used to. The M systems would give them that. In the end Microsoft is a software company. They want to make their software available on the systems that run it best. If that's macs then they will be fine with that.
As I noted earlier, I'd expect that we will hear more from them around the time of the Next Major update to Windows in April 2021 or so.
One thing that has held them back is a lack of systems that have more of a desktop CPU that is capable of running Windows in the manner that we are used to. The M systems would give them that. In the end Microsoft is a software company. They want to make their software available on the systems that run it best. If that's macs then they will be fine with that.
As I noted earlier, I'd expect that we will hear more from them around the time of the Next Major update to Windows in April 2021 or so.
Yes there is a version of Windows 10 for ARM. And yes they would have to make a special version of that for M systems. To handle the boot system and to provide device drivers. That's no problem for them. If the market is right.
I think it comes down to two things: licensing (currently Microsoft licenses the ARM version of Windows to OEMs) and support for the M1 chip. From what I've read in the past the ARM version of Windows is tied to the Qualcom Arm Chip.
That the million dollar question. Does Windows want to spend the engineer effort? Is there a financial benefit to them?
Careful about misleading people. As it stands now, an AS Mac can not run Windows either via dual boot or virtualized. Parallels may be developing a native AS app, but that does not mean it will virtualize x86 Windows. It won’t.
I beg to differ. M1 is a game changer as far as the speed and efficiency. If MS made native windows OS, Macs with M1 chips would blow away other laptops in speed and battery life, not to mention cooling.
The windows code base is already a rats nest of inherited code. Generation after generation. Bastardizations of interfaces and so on. It has improved in UI/UX by vast margins since Windows 10 was released but it’s still just utterly unfinished and unrefined.
That said it’s the jack of all trades and master at none approach. Not the best tool for the job but a singular tool that can address almost any need with enough effort.
Macs are brilliant and more specialized tools. They get derided for a lack of versatility or flexibility but that is exactly what is needed to “SPECIALIZE” a tool.
Obviously these are generalizations. But you get the point.
That said it’s the jack of all trades and master at none approach. Not the best tool for the job but a singular tool that can address almost any need with enough effort.
Macs are brilliant and more specialized tools. They get derided for a lack of versatility or flexibility but that is exactly what is needed to “SPECIALIZE” a tool.
Obviously these are generalizations. But you get the point.
Doesn't matter. Every day buyers who want a PC by and large will not consider Apple. And they probably shouldn't.
The M1 isn't really the highlight here. Yes, it's an amazing chip, but it's the OS and Rosetta that make the M1 special. Windows just isn't going to do it on M1.
I'm thinking that Microsoft releases an "app" version of Windows 10; you won't need VM software in a sense....you just pay for this Windows app from the Mac App Store, download it, answer a few questions the first time you run it, and you have a full fledged Win 10 installation running on top of MacOS.
Making Windows 10 work on the Apple M1 would likely require a joint effort between Microsoft and Apple. Microsoft would need to agree to licensing, and would need to fix any software compatibility issues in the kernel.
Windows on ARM64 expects a UEFI boot environment and ARM ACPI as the interface to the firmware. It would be up to Apple to make any firmware changes required for this. Apple would also have to put Microsoft root certificates in the firmware to allow firmware to validate the OS boot code. Microsoft also expects hardware vendors to develop device drivers for unique hardware. Microsoft pretty much only supplies drivers for "standard" hardware like NVMe/AHCI storage and XHCI USB controllers.
The graphics display controller on the M1 is a Apple design, so would require Apple to write a decent performing driver for Windows. This is likely a couple man-years of work to get all the acceleration working optimally, Windows would boot with a dumb frame buffer, although graphics performance might be pretty impressive. On Intel based machines, this graphics driver is written by Intel/AMD/NVidia, not Microsoft or Apple. I'm guessing game users may be a significant reason people would want a dual booted M1, so graphics performance might be important. Just for reference, the Lenovo Windows ARM64 (Qualcomm processor) laptop has I believe more than 50 Qualcomm specific hardware drivers, some pretty simple some really complex. I could easily imagine it was $10-$20M for Qualcomm in driver development to support Windows.
There is the question of would people find Windows in a virtual machine an attractive solution. For some users (corporate) perhaps yes, for some users (game) perhaps not as the graphics performance would likely be degraded over bare metal Windows boot. Apple does have a current job posting for a virtualization engineer, so it seems like a good bet Apple is expecting to support VMs in the firmware/OS.
I've spent the last 2.5 years at a ARM64 Server CPU vendor (not Qualcomm) making ARM64 Windows Server work correctly on the ARM64 architecture, so know what it takes to run Windows on ARM64 processors. Windows Server and Windows desktop are mostly the same code base at Microsoft (with features enabled/disabled to match the OS usage). Intel Windows and ARM64 Windows are also mostly the same code base.
As both Apple and Microsoft are for profit companies, the real question comes down to what is the estimated revenue benefits of having Windows run well on the M1. Based on this estimated revenue delta, Apple and Microsoft would be in a position to make decisions if the engineering effort would be cost justified or not. Apple and Microsoft being very large corporations, I'd guess they would not view it worth their trouble unless it meant a revenue delta above the accounting noise, which might be hundreds of millions of dollars.
I read at statistica.com that Apple sells about $25B/year of Macs (a surprisingly large number), and lets say running Windows is important for 5% of those sales, so if in 2 years, Mac's no longer can run Windows (due to the shift to ARM64) then Apple may lose $1.25B in Mac sales. Off the top of my head, making Windows work on Apple's ARM64 platform would cost a lot less than $1.25B, so the engineering required seems potentially justified for Apple. From Microsoft's point of view, $1.25B in systems sales sounds like maybe $50M in OS license sales, which is less clearly cost justified, especially since some of those sales may take away from Microsoft hardware sales. I think the question for Microsoft is working on Windows for M1 a good use of their not unlimited engineering resources (especially engineering staff), or would those same engineers be better used for something like Azure cloud engineering. If Apple funded some of the Microsoft NRE costs, it might make Windows on M1 easier for management at Microsoft to cost justify. Just a disclaimer, I don't have any inside info about Apple or Microsoft, so these back of envelope calculations may be way off.
Windows on ARM64 expects a UEFI boot environment and ARM ACPI as the interface to the firmware. It would be up to Apple to make any firmware changes required for this. Apple would also have to put Microsoft root certificates in the firmware to allow firmware to validate the OS boot code. Microsoft also expects hardware vendors to develop device drivers for unique hardware. Microsoft pretty much only supplies drivers for "standard" hardware like NVMe/AHCI storage and XHCI USB controllers.
The graphics display controller on the M1 is a Apple design, so would require Apple to write a decent performing driver for Windows. This is likely a couple man-years of work to get all the acceleration working optimally, Windows would boot with a dumb frame buffer, although graphics performance might be pretty impressive. On Intel based machines, this graphics driver is written by Intel/AMD/NVidia, not Microsoft or Apple. I'm guessing game users may be a significant reason people would want a dual booted M1, so graphics performance might be important. Just for reference, the Lenovo Windows ARM64 (Qualcomm processor) laptop has I believe more than 50 Qualcomm specific hardware drivers, some pretty simple some really complex. I could easily imagine it was $10-$20M for Qualcomm in driver development to support Windows.
There is the question of would people find Windows in a virtual machine an attractive solution. For some users (corporate) perhaps yes, for some users (game) perhaps not as the graphics performance would likely be degraded over bare metal Windows boot. Apple does have a current job posting for a virtualization engineer, so it seems like a good bet Apple is expecting to support VMs in the firmware/OS.
I've spent the last 2.5 years at a ARM64 Server CPU vendor (not Qualcomm) making ARM64 Windows Server work correctly on the ARM64 architecture, so know what it takes to run Windows on ARM64 processors. Windows Server and Windows desktop are mostly the same code base at Microsoft (with features enabled/disabled to match the OS usage). Intel Windows and ARM64 Windows are also mostly the same code base.
As both Apple and Microsoft are for profit companies, the real question comes down to what is the estimated revenue benefits of having Windows run well on the M1. Based on this estimated revenue delta, Apple and Microsoft would be in a position to make decisions if the engineering effort would be cost justified or not. Apple and Microsoft being very large corporations, I'd guess they would not view it worth their trouble unless it meant a revenue delta above the accounting noise, which might be hundreds of millions of dollars.
I read at statistica.com that Apple sells about $25B/year of Macs (a surprisingly large number), and lets say running Windows is important for 5% of those sales, so if in 2 years, Mac's no longer can run Windows (due to the shift to ARM64) then Apple may lose $1.25B in Mac sales. Off the top of my head, making Windows work on Apple's ARM64 platform would cost a lot less than $1.25B, so the engineering required seems potentially justified for Apple. From Microsoft's point of view, $1.25B in systems sales sounds like maybe $50M in OS license sales, which is less clearly cost justified, especially since some of those sales may take away from Microsoft hardware sales. I think the question for Microsoft is working on Windows for M1 a good use of their not unlimited engineering resources (especially engineering staff), or would those same engineers be better used for something like Azure cloud engineering. If Apple funded some of the Microsoft NRE costs, it might make Windows on M1 easier for management at Microsoft to cost justify. Just a disclaimer, I don't have any inside info about Apple or Microsoft, so these back of envelope calculations may be way off.
Let's suppose MS were willing to license an ARM version of Windows, and it ran well on an AS Mac. It's still (at least for a while) not going to be as clean as running Windows on an Intel Mac, because most who need to run Windows-only applications are running legacy x86 applications for work that haven't been, and likely won't be, ported to Windows-for-ARM. Thus these will need to be run though the emulation layer MS had included in Windows-for-ARM.
For a taste of some of the challenges accompanying this, see: https://www.theverge.com/2019/11/5/...view-arm-windows-10-apps-features-specs-price
[Note also that the only apps that can run in this way are 32-bit; the emulation layer doesn't work for 64-bit x86 apps. I know the AS Macs won't run 32-bit apps natively, but I'm guessing that this should be possible with the emulation layer.]
Where things might get interesting is if PC manufacturers move towards ARM, such that a large ecosystem of native ARM-for-Windows apps is created. PC users might then want to buy Macs purely to run Windows, if Apple's chips, at that point in the future, offer a significant performance advantage over what's available in PC's.
The conventional wisdom (which I generally agree with) has been that it didn't make sense to buy a Mac unless you wanted to run MacOS: Similar performance (with a few exceptions) can be obtained with PC's, and for less money. But, with AS, that may change—you'll still pay more but, assuming my hypothetical is realized, you'll get performance not available from a PC.
For a taste of some of the challenges accompanying this, see: https://www.theverge.com/2019/11/5/...view-arm-windows-10-apps-features-specs-price
[Note also that the only apps that can run in this way are 32-bit; the emulation layer doesn't work for 64-bit x86 apps. I know the AS Macs won't run 32-bit apps natively, but I'm guessing that this should be possible with the emulation layer.]
Where things might get interesting is if PC manufacturers move towards ARM, such that a large ecosystem of native ARM-for-Windows apps is created. PC users might then want to buy Macs purely to run Windows, if Apple's chips, at that point in the future, offer a significant performance advantage over what's available in PC's.
The conventional wisdom (which I generally agree with) has been that it didn't make sense to buy a Mac unless you wanted to run MacOS: Similar performance (with a few exceptions) can be obtained with PC's, and for less money. But, with AS, that may change—you'll still pay more but, assuming my hypothetical is realized, you'll get performance not available from a PC.
Last edited:
Ubuntu will come out with a version. they had versions for G4 and G5 so why not the M1.