Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
What are you even talking about? Rosetta is for emulating x86 apps in macOS, not Windows! Windows ARM in a virtual machine will not emulate x86 apps using Rosetta–it does so using Microsoft's own emulation tool.

I think you're misreading that post.

He meant that on the Mac side, we can run x86 apps (via CrossOver) because CrossOver has a 32-bit to 64-bit translation layer, and when it's x86-64, Rosetta 2 does the rest.

In Windows on ARM, many apps are arm32, so they don't run at all with M1 because as far as we know, M1 is only compatible with arm64. Those arm32 apps will crash if you attempt to open them in the VM as Windows does not translate arm32 to arm64.

The "only" way out now is to find the x86 versions of those apps and replace the arm32 versions.
 
Isn't BootCamp not allowed on the M1?

Apple have said that they won't support "direct booting of alternative OSs" (i.e. BootCamp) on Apple Silicon, and

AFAIK booting other OSs is "allowed" (M1 systems can be configured to boot unsigned images) but that is moot unless alternative OSs exist. For an OS to run on M1 hardware it will need to work with the Apple bootloader/firmware and have drivers for video, SSD and all the other hardware that is now provided by Apple's proprietary systems-on-a-chip. That would be hard without active support from Apple, but maybe not impossible, with reverse engineering.

(Intel Macs used a lot of generic PC hardware that worked with built-in or third party Windows drivers, and used almost-but-not-quite standard firmware that could be extended to support Windows... that changed a bit with the T2 but Apple still supported BootCamp).

What Apple are supporting is virtualisation, which always includes a certain amount of hardware emulation and "dummy" drivers for which the hypervisor (Parallels etc.) passes on to the MacOS drivers.

I honestly don't know why Microsoft doesn't work with Apple to make Boot Camp possible with ARM Windows.

Well, for a start, Apple would have to supply specs for how to write video, SSD, etc. drivers for Apple Silicon and that would mean having to "freeze" those specs to some extent. Currently, as the only source for ASi-compatible operating systems, they are free to chop and change specs with every new Mac model and fix them with a point release of MacOS.

Meanwhile, for a large proportion of users who want to be able to run Windows or Linux on their Mac, virtualisation is the best tool for the job - in return for a bit of lost performance you avoid having to reboot to switch OS, avoid having to physically partition the disc (wasteful - with a VM you can have a virtual disc file that starts small and grows on demand), get file sharing - and often cut & paste - between MacOS and Windows (rather than mess around with APFS for Windows and/or NTFS on Mac) plus get other useful features like snapshots, more flexible networking and more protection from nasties infecting the PC side.
 
Can anybody help me understand the use of Windows on ARM with a Mac?? Is there enough app support to justify adding the OS to an M1?
At the risk of stating the obvious, it's to be able to run Windows software on your Mac....via one of two mechanisms:
1) Running Windows apps that are already ARM-native; the number and quality of these will hopefully increase over time
2) Running 32-bit and 64-bit Intel x86 Windows applications via emulation. If this runs acceptably well, it would be sufficient for most Mac users who need to run some Windows applications.
 
A solution to a problem that pretty much does not exist. There's a very limited set of apps that run on Windows ARM and I suspect that they're already available on a Mac. I doubt many would need this.
It's not the Windows-on-ARM apps that matter (for now); it's the ability to run x86_64 Windows apps - i.e. all the existing Windows software....provided it works well enough.
 
Another data point: while those default setting (2-core) benchmark scores might not seem impressive, I just ran the same test under Parallels with default settings on my 2-year-old Intel Mac Mini with a 3.19 GHz Core i7 Coffee Lake—and my scores were quite a bit worse than what your M1 turned in: 939 single-core / 1649 multi-core. Once they get everything out of beta (and assuming Microsoft allows use of Windows in this use case) I'd take that trade-up happily.
 
The reason why some apps crash and don't launch is that some of the Windows Store apps are 32-bit ARM (which won't run on Apple Silicon which hasn't had support for 32-bit ARM instruction sets since A10 Fusion and A10X Fusion) Incidentally, Microsoft will eventually have to update those apps as ARM is set to ditch 32-bit instruction sets in their designs over the next couple of years (so the future of ARM will be as 32-bit-less as Apple Silicon has been for the last few years).

Furthermore, (a) this is a technical preview of Parallels for Apple Silicon (i.e. pre-Beta [i.e. bugs and glitches are going to be common]) and (b) it's running an Insider Preview build (also Beta) version of Windows 10 for ARM64 that never was supposed to have driver support for the VM it was installed on. All things considered, the fact that we're even able to get THIS far without Microsoft or Apple's intervention is incredibly impressive. I would still caution anyone needing serious Windows use to either get an x86-64 PC made in the last seven years (including current stuff) to use alongside your M1 Mac or to buy one of the last rounds of Intel Macs as that will be much more stable and compatible with what you need Windows on a Mac for to begin with.
 
Can anybody help me understand the use of Windows on ARM with a Mac?? Is there enough app support to justify adding the OS to an M1?

Some people need to run apps that don't exist (and won't ever exist) for Mac. Doesn't matter whether your Mac is Apple Silicon or Intel based. The need will continue to exist for a while.

Wouldn’t that make Macs the best Windows machines? If M1 is already beating official ARM Microsoft-branded PCs via virtualization, imagine the native performance benchmarks if you could install ARM Windows via Boot Camp.

ARM64 Windows machines won't be the best Windows machines until emulated performance is on par with what Apple has for Rosetta 2. Until then, it's a niche substitute for x64 Windows 10 which makes an Intel Mac a better buy. But shift both platforms over to ARM64 to the point where we keep seeing this degree of performance (including apps), and then you'll have the best Windows machines. For the time being, they're only the best ARM-based Windows machines.

ARM Windows also runs Intel apps (though the x64 emulator is still in beta). This would thus allow us to run a ton of Windows apps, including games, in an environment potentially much more reliable than CrossOver.

I honestly don't know why Microsoft doesn't work with Apple to make Boot Camp possible with ARM Windows. Perhaps in 1-2 years? Anyway, the first Intel Macs didn't have Boot Camp from day one either, so we might still see something in the future.

It's a serious win-win for both of them to make running the ARM64 version of Windows 10 run on Apple Silicon Macs. I'd bet that it's in the works. It'd be stupid for both of them for it not to be. Though, I don't believe that it'll be Boot Camp as we know it. I think virtualization is a for sure. I do believe that Apple will also reverse course on direct booting for Windows 10 for ARM64 on Apple Silicon Macs, but this will take some time as it won't be a simple matter of cooking drivers for the industry standard PC components that Apple used in Intel Macs and then enabling CSM support as was done on 2006-2014 Intel Macs. Apple and Microsoft would need to engineer a whole new bootloader and Apple would need to create drivers for Apple Silicon Mac SoCs and components for Windows 10 for ARM64. Not impossible, but it still requires much more work than was needed for Intel Mac Boot Camp.

Isn't BootCamp not allowed on the M1?

It's not. This is virtualization, not Boot Camp.

Just another reason I don't do Windows! :eek:

Because you don't like pre-release software that isn't fully baked? Then what are you doing using macOS where at least every other annual release is a long-term public beta? 🤣

Apple have said that they won't support "direct booting of alternative OSs" (i.e. BootCamp) on Apple Silicon, and

They didn't support it on Intel Macs until Intel Macs had been shipping for three months already. Not saying that they'll reverse course on it this soon, but certainly, it's not outside of the realm of reason to believe that they may do so again.

AFAIK booting other OSs is "allowed" (M1 systems can be configured to boot unsigned images) but that is moot unless alternative OSs exist. For an OS to run on M1 hardware it will need to work with the Apple bootloader/firmware and have drivers for video, SSD and all the other hardware that is now provided by Apple's proprietary systems-on-a-chip. That would be hard without active support from Apple, but maybe not impossible, with reverse engineering.

Yes, and Microsoft and Apple both stand A LOT to gain by collaborating to rectify this. Also, it's not Apple's bootloader that needs to be changed, because it's not Apple's OS (the bootloader is an OS component, not a hardware component). Apple and Microsoft both need to collaborate on the Bootloader side of things. And Apple would need to write drivers (not just use off-the-shelf drivers that already exist for pre-existing components, like you said).

What Apple are supporting is virtualisation, which always includes a certain amount of hardware emulation and "dummy" drivers for which the hypervisor (Parallels etc.) passes on to the MacOS drivers.

It's Apple's Hypervisor framework that Parallels and VMware will be leveraging on Apple Silicon Macs. Technically, Apple's involvement will still be required for drivers. It just won't be as crazy as designing drivers for the native hardware. But it's still technically Apple's hypervisor that needs drivers.


Well, for a start, Apple would have to supply specs for how to write video, SSD, etc. drivers for Apple Silicon and that would mean having to "freeze" those specs to some extent. Currently, as the only source for ASi-compatible operating systems, they are free to chop and change specs with every new Mac model and fix them with a point release of MacOS.

Apple doesn't need to supply specs to anyone. They just need to write the drivers. If anything, it's Microsoft that would need to supply specs for writing drivers for ARM64 Windows 10. But I'm pretty sure those exist already for OEMs.

Meanwhile, for a large proportion of users who want to be able to run Windows or Linux on their Mac, virtualisation is the best tool for the job - in return for a bit of lost performance you avoid having to reboot to switch OS, avoid having to physically partition the disc (wasteful - with a VM you can have a virtual disc file that starts small and grows on demand), get file sharing - and often cut & paste - between MacOS and Windows (rather than mess around with APFS for Windows and/or NTFS on Mac) plus get other useful features like snapshots, more flexible networking and more protection from nasties infecting the PC side.
APFS and dynamic disk partitioning renders your concerns here moot. Direct booting will always be superior in terms of performance (hell, that much is evident in the benchmark scores of these VMs versus macOS bare metal installations). I'm not saying it won't suffice for most, but I wouldn't go so far as saying that it's "the best" solution. I'll happily agree that later era Boot Camp has become a cludgy mess that is a far cry from how neat and streamlined it was when it launched originally. But a lot of that is due to how Apple managed the T2 integration and driver support within Windows. That and their on-logic-board SSDs. But, with Apple Silicon, they have the freedom to make it different as just about everything else under the hood with these new Macs is different. Apple and Microsoft just need to work together (and the incentives for them to do so are already there; both stand to gain big by doing so).
 
Was the last time you used an windows machine during Vista or something?

This reads like fanboy garbage.
I work as an IT, he is pretty right even if he was joking. No freeze, but so many bugs, issues to lose your mind in 2020. Windows 10 installing sponsored content and games with pro and enterprise license is absolutely garbage.
 
I work as an IT, he is pretty right even if he was joking. No freeze, but so many bugs, issues to lose your mind in 2020. Windows 10 installing sponsored content and games with pro and enterprise license is absolutely garbage.
Seriously. I hate having to uninstall a bunch of junk every time I setup a fresh windows installation.

No really, windows 10 has been good so far. Fast and doesn't need a reinstall every 1-2 years.
I keep an up-to-date installation in a vm and maintain multiple machines for my family, so I’m pretty familiar with it. My biggest gripes have more to do with design than performance.
 
For those of you wondering why people would want to run Windows, my particular app which does not exist on Mac is Metatrader 5. It’s a 64 bit x86 app, which I hope will run okay on Windows ARM.

The only other solution for Windows on an M1, which I’m currently using, is to remote to a Windows VM. My client of choice is Jump Desktop. It’s not ideal because of network lag and the fact that you have to be connected to use it.

The rest of the app ecosystem is far better on Mac. I absolutely love OmniFocus and GarageBand. To my surprise, at some point in the last couple of years, Microsoft even brought in support for VBA in their Mac Office suite, which used to be a reason to run Windows Office
 
Two comments about the very informative test.

firstly, how do I get a copy of Windows ARM? I signed up to the program, but it seems like the only way I can get it is via a conventional Windows instance, which I don’t have. Chicken and egg.

secondly, you seem to dole out only 3GB RAM to the VM. Unless Windows ARM is considerably more frugal with RAM usage, that is way too low and would greatly impact performance. I am lucky to have a 16GB MBP M1, and I would allocate 8GB to the VM. On an 8GB machine, I’d make it 4-5 GB.
 
In Windows on ARM, many apps are arm32, so they don't run at all with M1 because as far as we know, M1 is only compatible with arm64. Those arm32 apps will crash if you attempt to open them in the VM as Windows does not translate arm32 to arm64.

Do we know for a fact that M1 can't run ARM32, or is that conjecture?
 
Some people need to run apps that don't exist (and won't ever exist) for Mac. Doesn't matter whether your Mac is Apple Silicon or Intel based. The need will continue to exist for a while.



ARM64 Windows machines won't be the best Windows machines until emulated performance is on par with what Apple has for Rosetta 2. Until then, it's a niche substitute for x64 Windows 10 which makes an Intel Mac a better buy. But shift both platforms over to ARM64 to the point where we keep seeing this degree of performance (including apps), and then you'll have the best Windows machines. For the time being, they're only the best ARM-based Windows machines.



It's a serious win-win for both of them to make running the ARM64 version of Windows 10 run on Apple Silicon Macs. I'd bet that it's in the works. It'd be stupid for both of them for it not to be. Though, I don't believe that it'll be Boot Camp as we know it. I think virtualization is a for sure. I do believe that Apple will also reverse course on direct booting for Windows 10 for ARM64 on Apple Silicon Macs, but this will take some time as it won't be a simple matter of cooking drivers for the industry standard PC components that Apple used in Intel Macs and then enabling CSM support as was done on 2006-2014 Intel Macs. Apple and Microsoft would need to engineer a whole new bootloader and Apple would need to create drivers for Apple Silicon Mac SoCs and components for Windows 10 for ARM64. Not impossible, but it still requires much more work than was needed for Intel Mac Boot Camp.



It's not. This is virtualization, not Boot Camp.



Because you don't like pre-release software that isn't fully baked? Then what are you doing using macOS where at least every other annual release is a long-term public beta? 🤣



They didn't support it on Intel Macs until Intel Macs had been shipping for three months already. Not saying that they'll reverse course on it this soon, but certainly, it's not outside of the realm of reason to believe that they may do so again.



Yes, and Microsoft and Apple both stand A LOT to gain by collaborating to rectify this. Also, it's not Apple's bootloader that needs to be changed, because it's not Apple's OS (the bootloader is an OS component, not a hardware component). Apple and Microsoft both need to collaborate on the Bootloader side of things. And Apple would need to write drivers (not just use off-the-shelf drivers that already exist for pre-existing components, like you said).



It's Apple's Hypervisor framework that Parallels and VMware will be leveraging on Apple Silicon Macs. Technically, Apple's involvement will still be required for drivers. It just won't be as crazy as designing drivers for the native hardware. But it's still technically Apple's hypervisor that needs drivers.




Apple doesn't need to supply specs to anyone. They just need to write the drivers. If anything, it's Microsoft that would need to supply specs for writing drivers for ARM64 Windows 10. But I'm pretty sure those exist already for OEMs.


APFS and dynamic disk partitioning renders your concerns here moot. Direct booting will always be superior in terms of performance (hell, that much is evident in the benchmark scores of these VMs versus macOS bare metal installations). I'm not saying it won't suffice for most, but I wouldn't go so far as saying that it's "the best" solution. I'll happily agree that later era Boot Camp has become a cludgy mess that is a far cry from how neat and streamlined it was when it launched originally. But a lot of that is due to how Apple managed the T2 integration and driver support within Windows. That and their on-logic-board SSDs. But, with Apple Silicon, they have the freedom to make it different as just about everything else under the hood with these new Macs is different. Apple and Microsoft just need to work together (and the incentives for them to do so are already there; both stand to gain big by doing so).
Not now, but the performance of Apple's M1s and what is to come from them will influence the industry towards ARM processors and the development of operating systems to work on them.

ARM chips are proving themselves a technological iteration in power draw and overall efficiency and it's a marvel. In Dec 2020, Windows on ARM doesn't look like much. By December 2022 or 2023 however...

ARM Windows also runs Intel apps (though the x64 emulator is still in beta). This would thus allow us to run a ton of Windows apps, including games, in an environment potentially much more reliable than CrossOver.

I honestly don't know why Microsoft doesn't work with Apple to make Boot Camp possible with ARM Windows. Perhaps in 1-2 years? Anyway, the first Intel Macs didn't have Boot Camp from day one either, so we might still see something in the future.
As a only MacOS user, I often get limited on my software options mainly because the ability to run windows on Mac and the fact that Apple did a poor job in promoting Metal vs CUDA implementation by nvidia, has allowed and even incentives developers not to offer Mac versions of their apps.
The ARM macs performance and the lack of bootcamp could make Mac versions more appealing... Though I’m no expert, on the other hand, as I understand Microsoft is the only company that could offer a real competitive product to Apple’s silicon, assuming Intel or more likely AMD are able to quickly provide the customized SoC for it, as it has the capability and is the only company which already provides a OS coupled hardware solution (surface) and this is a prerequisite for manage and take full advantage of the unified architecture responsible for the amazing performance of the ARM Mac silicon..so maybe Microsoft might have reasons not to get Windows running on Macs right now, as they can render themselves as the only real contenders to ARM Macs. I’m afraid other PC makers will be left with the upgradability argument as the solely reason to opt for them.
I’m conflicted by my desire of Apple not supporting Bootcamp, as this will limit options for customers, but right now with head lead on performance, this could be incentive enough for developers to adopt Metal.
 
  • Like
Reactions: psychicist
Do we know for a fact that M1 can't run ARM32, or is that conjecture?

There's no way to check since the only OS that natively runs on M1 currently is MacOS, and it only runs ARM64 code, nothing else (well, unless you also count Rosetta 2 and x86-64).

This has been the case since iOS 11 as well. Apple has only allowed ARM64 apps since iOS 11, so by now, there's almost nothing that runs ARM32 even on the iOS side.

If we can natively boot a Linux OS, perhaps we'll know for sure if Apple's M1 chip is compatible with ARM32. For now, I'd guess the answer is "no", but it's not definitive.
 
Can anybody help me understand the use of Windows on ARM with a Mac?? Is there enough app support to justify adding the OS to an M1?
Oh yes,
specialised accounting and information apps
Dragon naturally speaking
and I tested parallels and windows 10 arm for this - and it worked like a charm.
 
  • Like
Reactions: NetMage
I can tell you one thing
Big Sur is just a BIG BLOATED disaster
Catalina was better
We keep going in reverse on every OS update
12 GB of BLOAT and painfully slow boots
and your updates for all your installed apps fail to show up

Windows 10 has a much better track record for fast boots and reliability.
 
There's no way to check since the only OS that natively runs on M1 currently is MacOS, and it only runs ARM64 code, nothing else (well, unless you also count Rosetta 2 and x86-64).

This has been the case since iOS 11 as well. Apple has only allowed ARM64 apps since iOS 11, so by now, there's almost nothing that runs ARM32 even on the iOS side.

If we can natively boot a Linux OS, perhaps we'll know for sure if Apple's M1 chip is compatible with ARM32. For now, I'd guess the answer is "no", but it's not definitive.
Right. I guess I was expecting Apple to only remove their frameworks from macOS, not actual CPU instructions from their SoC.
 
Can anybody help me understand the use of Windows on ARM with a Mac?? Is there enough app support to justify adding the OS to an M1?

There is for me. All my x86 apps run very well in the compatibility mode which comes with ARM and its been very stable, so its back to being able to do everything I need to do on my one machine.

The M1 runs this ARM version of Windows much, much more smoothly than the 2020 MBA did, so I'm delighted this has been made available so quickly.
 
Microsoft will have a good idea of how may mac users are using windows on their intel based macs. It will be these people that will upgrade their mac machines to the new ARM mac's and thus want to have windows running on these machines also. There has to be enough mac window users for Microsoft to justify the cost of ARM developement for the mac. If the figures do not work in Microsoft's favor, they will not develop windows ARM for macs and will leave it for the VM companies to do it.
 
18.85% is nowhere near to the 200% you would expect with a doubling of cores. I think that was the point.
Doubling the cores would, in the best case, bring a 100% increase.

Now, obviously, 19% is a lot less than that. Two reasons compound to this:

  • the M1 only has eight cores, and the host OS needs to be able to do some work as well. Offering all those cores to the VM probably requires that context switching overhead. I therefore wouldn't have been surprised if running the VM with eight cores is actually slower than with four (and I can imagine it still is, in some scenarios; would love to see some detailed Geekbench or similar results — for example, perhaps effective memory throughput is severely decreased due to latency?).
  • of those eight cores, four are "efficiency" cores, meaning they're much slower than the other four. If you limit the VM to four cores, the controller will mostly give it the performance ones, but once you give it more, it doesn't really have a choice but to allocate the efficiency cores as well, which are a fair bit slower.
On top of that, of course, Amdahl's law. Doubling cores is never going to get you twice the performance, in practice. More like a 80-90% increase. And that's basically what we're seeing with the change from two to four cores: the four-core setup is 82% faster than the two-core setup. That's well within expectations.
 
  • Like
Reactions: rgbrock1
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.