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.
Parallels Desktop for Arm's Mac?
- Thread starter ginofresh
- Start date
- Sort by reaction score
Microsoft already has an ARM-based Hardware product in its portfolio, the Surface Pro X. And Microsoft is said to be developing an x64-emulator for ARM, with tests slated to start in late 2020 (which nicely coincides with Apple being said to release the first ARM Macs in that time frame). Here is a link to a German news article about that topic.
I think if the goal is to support running Windows 10 on a Silicon-based Mac, it needs to be the current flavor of Windows 10 that the world uses - many professionals need to test apps in this environment, or they are using legacy apps that will never move to the Mac so they certainly won't move to a smaller market like Windows for ARM. So imho Windows ARM isn't likely to be a viable solution - we need commercial Windows 10.
My theory, since we are conjecturing, is one of these three:
1. Parallels will create emulation capability to run Windows 10. They are working closely with Apple already, keep in mind Apple Silicon is not a CPU but a SoC/System on a chip - if Apple values the need for virtual Windows by their pro customers, they could certainly enhance the experience through SoC design and potentially other processors to make emulation run acceptably. For example, just putting some/most/all of the foundational emulation-necessary instructions in silicon would speed things up a lot.
2. Apple has already shown it knows how to build and use co-processor chips for security, graphics, etc. Apple Silicon, just like the SoC on their phones, is a collection of well-integrated processors and electronics, not just a CPU chip. For high end pro machines, what about an intel co-processor chip as part of the SoC, and processing is shuttled off to that chip like is done with graphics, security, etc.? Then virtualization is very viable if that design works and is not too costly. While Apple wants to get away from Intel CPUs because they can make their own better and quicker (they'll have 5nm chips before Intel has 7nm chips, for example), a co-processor for emulation or virtualization does not have to be the best in industry cpu - a less expensive alternative would work fine if virtualization is being done in hardware on a true x86 chip.
3. There are $100 sticks that run Windows 10 via USB, they are small form factors and inexpensive; even as is, they are that size and cost in part due to the need to add memory, storage space/SSD, and other hardware as part of the stick so it's a small bootable 'computer'. Couldn't this design be replicated as part of the SoC architecture, using the Mac's native disk, memory, etc., so that there is a low cost capability to support Windows 10?
I totally acknowledge these ideas are 15% reality and 85% myth, as I am not a computer hardware engineer. And yet, I ponder if there is enough truth in these theories that the true idea is out there and we will be pleasantly surprised that Apple is not going to abandon its customers that need to run Windows apps on MacOS. I am about 55% confident. ?
My theory, since we are conjecturing, is one of these three:
1. Parallels will create emulation capability to run Windows 10. They are working closely with Apple already, keep in mind Apple Silicon is not a CPU but a SoC/System on a chip - if Apple values the need for virtual Windows by their pro customers, they could certainly enhance the experience through SoC design and potentially other processors to make emulation run acceptably. For example, just putting some/most/all of the foundational emulation-necessary instructions in silicon would speed things up a lot.
2. Apple has already shown it knows how to build and use co-processor chips for security, graphics, etc. Apple Silicon, just like the SoC on their phones, is a collection of well-integrated processors and electronics, not just a CPU chip. For high end pro machines, what about an intel co-processor chip as part of the SoC, and processing is shuttled off to that chip like is done with graphics, security, etc.? Then virtualization is very viable if that design works and is not too costly. While Apple wants to get away from Intel CPUs because they can make their own better and quicker (they'll have 5nm chips before Intel has 7nm chips, for example), a co-processor for emulation or virtualization does not have to be the best in industry cpu - a less expensive alternative would work fine if virtualization is being done in hardware on a true x86 chip.
3. There are $100 sticks that run Windows 10 via USB, they are small form factors and inexpensive; even as is, they are that size and cost in part due to the need to add memory, storage space/SSD, and other hardware as part of the stick so it's a small bootable 'computer'. Couldn't this design be replicated as part of the SoC architecture, using the Mac's native disk, memory, etc., so that there is a low cost capability to support Windows 10?
I totally acknowledge these ideas are 15% reality and 85% myth, as I am not a computer hardware engineer. And yet, I ponder if there is enough truth in these theories that the true idea is out there and we will be pleasantly surprised that Apple is not going to abandon its customers that need to run Windows apps on MacOS. I am about 55% confident. ?
An x64 emulator isn't really important at the moment. Windows for ARM already has an emulator for 32-bit Windows applications, and almost all Windows apps are still available as 32-bit versions. (After all, you can still buy a 32-bit version of Windows for Intel chips, and that version does not run anything but 32-bit apps, even on 64-bit hardware.)
Additionally, many Windows apps (most of those made for Microsoft's .NET platform) are platform-independent and will run natively on ARM, even if the developer never created a special ARM version.
So you would ideally use the native 64-bit ARM version of your Windows software, and if that's not available, just use the 32-bit Intel version.
In the future, there may be more apps that are no longer available in a 32-bit version, and just in case Windows on ARM doesn't take off and developers don't create 64-bit ARM versions of their apps as well, a x64 emulator will be a nice thing to have, of course.
You can use your legacy Windows apps on Windows for ARM as well, because Windows will transparently emulate an Intel CPU. And it's a much better and faster solution than some Parallels emulator that runs the Intel version of Windows. After all, a virtualized Windows for ARM runs natively and with full performance on the new Macs. Only the legacy app itself will have to be emulated. With a hypothetical Parallels emulator, not only the legacy app, but the entire Windows OS including drivers, background services etc. needs to be emulated. This would be much slower and it would needlessly drain the battery.
(There might indeed be some people who need to test complex apps or drivers on Windows 10 for Intel CPUs, but I really doubt Parallels would go through the trouble of creating an emulator just for them.)
Last edited:
Apple is going to force people that have resource hungry x86 Windows apps to switch their workflow to the macOS or buy a spare PC.
Has anyone seen anything about virtualizing old versions of OS X? It doesn't seem like you'd be able to run a virtual version of El Capitan. What if I want to get nostalgic and play games that rely on rosetta 1?
Yes, there are some howto's for running Tiger and newer in hypervisors if you want to play around with PPC Rosetta.
Back in the 80s you could run a software only (though there was also a hardware addon option) emulation of x86 PC/MS-DOS on an Atari ST with at least 1MB of RAM. The ST was using the 68000 @ 8Mhz and could emulate a x86 8088 @2Mhz (If I recall the numbers right).
ANYWAY, my point is if the 68K could through emulation alone do it, I am sure whatever 2020 chip will be able too, even at a cost of performance.
I am not supporting the move, just saying that my guess is it will be doable from a technically point, though that doesn't mean it will be permissible.
ANYWAY, my point is if the 68K could through emulation alone do it, I am sure whatever 2020 chip will be able too, even at a cost of performance.
I am not supporting the move, just saying that my guess is it will be doable from a technically point, though that doesn't mean it will be permissible.
I hope there will be a X86 emulation that can run my existing Virtual Machines (Windows XP, Windows 10, macOS 10.6, macOS 10.13), because in each of this I have some software that won't run directly in host (macOS 10.15.7) or in the others VM's. I guess that M1x CUPs will be enough powerful to have acceptable performance in emulating X86 machines.
Buying a PC or holding my old MacBook Pro to run this software is not a solution when I'm away from home or office... I won't travel with 2 computers!
I am confident that there will be at least some solution from open software community, but I suppose there may also be a market for commercial products: if there was with Virtual PC when Macs were niche, I don't see why not now.
Buying a PC or holding my old MacBook Pro to run this software is not a solution when I'm away from home or office... I won't travel with 2 computers!
I am confident that there will be at least some solution from open software community, but I suppose there may also be a market for commercial products: if there was with Virtual PC when Macs were niche, I don't see why not now.
My speculation: could Apple put in decoding hardware in Apple Silicon to run emulation at much better performance? They will have the transistors available if TSMC keeps shrinking their processes.
Could? Probably. Would? Not likely.
I’d rather see them build components to accelerate other things than backwards compatibility, to be honest. Force software companies to move to an evergreen development process keeping up with up-to-date software libraries and more portable code instead - that’s better for everybody interested in more than the shortest of short-term profit and it necessitates keeping technical debt at a reasonable level.
Apple benefits from people moving off x86 faster - and pushing Windows only software vendors to do AS ports. The benefits of better x86 emulation, though, are the customers. There is already x86 emulation hardware in the chips. I was just wondering if they could do more of that.
I don’t believe there is, actually. I know there were some claims of that here on this forum a while ago, but I’ve seen nothing to indicate this is the case. Rosetta 2 is a very clever piece of software, but nothing indicates it’s dependent on hardware trickery.
As far as anyone knows, the only hardware in the M1 dedicated to x86_64 (AMD64) translation is a memory ordering mode that emulates Intel's memory model for more strict ordering called Total Store Ordering (TSO). By contrast, the normal Arm mode is a less strict ordering mode. Without this mode, the translation from x86_64 would have to create barriers in the Arm instructions to force ordering to match what x86_64 programs expect. Doing it in hardware should speed up execution by forcing this memory ordering automatically in hardware.