I guess it makes financial sense. Office 365 is where the money is and that is (or will be?) available for ASi Macs. Windows is free and it probably doesn't make sense to spend resources to make it compatible with ASi.
MS could go the route of actually charging for Windows on ASi, but I guess the numbers don't add up so it's not viable right now.
Windows is a $32 billion a year business for MS so it definitely isn't free. OEM's have to pay MS a licensing fee to install Windows on their machines.
The real question is if Microsoft will continue to try and break Apple Silicon compatibility, or will allow its use unofficially.
On one hand, it threatens Microsoft's bread-and-butter Wintel homogeny - on the other hand, there will soon be a fairly vast pool of Apple Silicon devices which will be entirely outside the Windows market if they do.
And let's face it: this may be the fastest ARM Windows will run for some time, despite Microsoft's claim that they're making their own chips (née Qualcomm).
Apple doesn’t now or will ever sell enough Macs to threaten Windows grip on the desktop market.
Grip on the market can be a fleeting thing - and while it can appear that the market is one humongous blob, it's actually composed of a lot of smaller pieces.
The server market is almost certainly going with some ARM or other RISC variant - if for no other reason than to minimize data center energy use. Most servers don't have to use high power CPUs - they just need the ability to run a ton of discrete processes like file shares. You only have to walk behind a server rack to see that there are real cost savings to be had by reducing energy use. Future servers may still be administered by MCSEs, but more than likely they'll not be running on x86 hardware.
The consumer market is much less dependent on Windows legacy software and is more OS agnostic. Most users don't care what OS they're running and can be swayed by price/performance/energy efficiency, especially in the laptop segment of the market. The government may actually move in and mandate energy efficiency as they have in six states already, just as they have done in the past with automobiles.
On the enterprise side, a lot of enterprise software is moving from Windows client/server based to web based solutions due to the high cost of maintaining a whole fleet of Windows client machines at a reasonably homogenous level of software compatibility (and keeping end users from installing their own possibly vulnerable software). A lot of enterprises are replacing Windows client machines with linux based thin clients attached to network file servers for storage - much easier and cheaper to maintain and replace. This also means that end users can't necessarily come home and install pirated copies of the same software they use at work, further reducing the demand for home hardware compatible with what you use at work.
So ... the market isn't some big hulking thing you can hold on to with both hands and maintain a firm grip. The Wintel homogeny came to be for a lot of these factors which are fracturing into smaller categories, each with their own motivation and reason for being.
I remember the days when Macs were not Intel based, but we could run Intel software using an emulator (SoftWindows).
I see no reason why that concept couldn't be resurrected. The Mac Silicon chips are quite fast. While an emulated Intel processor may not be great for gaming, it should be very good for running a lot of existing Windows based business software.
ARM Windows can already run x86 software using a code translator like Rosetta.
Let’s be realistic: Microsoft CANNOT support Windows on M-series CPU Macs, at least not until Apple wants is and actively supports it (maybe even pays it) and Apple doesn’t seem to have any intentions to do so.
Why? Because the M1 chip, while having ARM standard cores, also has proprietary GPU cores and proprietary neutral cores. In short, M$ could only support a fraction of what the CPU has to offer, and would need to rely on costly reverse engineering for the rest, something they’d have to repeat with each iteration of Apple Silicon.
At this point, it’s not even clear if Apple isn’t forking the entire ARM architecture: Apple has introduced proprietary extensions that stand at odds with ARM’s recent extensions to the instruction set.
ARM pushes Ethos (NPU) and Mail (GPU), and as far as I understand Apple’s NPU and GPU cores aren’t superior implementations of the same architecture (as is the case for the CPU part of the M-series chips). So in essence, let’s be clear: Apple’s M-series is only ARM architecture in regards to the ever less relevant CPU part of the SoC, NPU and GPU cores are utterly proprietary, closed, undocumented in need of reverse engineering, designed to be accessed through Apple’s APIs and libraries.
Now tell me again how Microsoft could possibly support Windows on M-series hardware, except through an inferior emulation that’s performance limited because it can’t use the hardware to the fullest. There will of course be VMs that will run it, as whatever standard hardware M$ will support, can be emulated; but it will be “unsupported, use at your own peril”.
It’s also not clear if and when Apple will implement the new developments in the ARM architecture, or just roll their own extensions:
Let’s be realistic: Microsoft CANNOT support Windows on M-series CPU Macs, at least not until Apple wants is and actively supports it (maybe even pays it) and Apple doesn’t seem to have any intentions to do so.
Why? Because the M1 chip, while having ARM standard cores, also has proprietary GPU cores and proprietary neutral cores. In short, M$ could only support a fraction of what the CPU has to offer, and would need to rely on costly reverse engineering for the rest, something they’d have to repeat with each iteration of Apple Silicon.
At this point, it’s not even clear if Apple isn’t forking the entire ARM architecture: Apple has introduced proprietary extensions that stand at odds with ARM’s recent extensions to the instruction set.
ARM pushes Ethos (NPU) and Mail (GPU), and as far as I understand Apple’s NPU and GPU cores aren’t superior implementations of the same architecture (as is the case for the CPU part of the M-series chips). So in essence, let’s be clear: Apple’s M-series is only ARM architecture in regards to the ever less relevant CPU part of the SoC, NPU and GPU cores are utterly proprietary, closed, undocumented in need of reverse engineering, designed to be accessed through Apple’s APIs and libraries.
Now tell me again how Microsoft could possibly support Windows on M-series hardware, except through an inferior emulation that’s performance limited because it can’t use the hardware to the fullest. There will of course be VMs that will run it, as whatever standard hardware M$ will support, can be emulated; but it will be “unsupported, use at your own peril”.
It’s also not clear if and when Apple will implement the new developments in the ARM architecture, or just roll their own extensions:
HUH.... That's odd... considering many Windows only games work right now on Apple M1 silicon through Windows for Arm (with decent framerates I might add)... As far as I recall, games require the use of GPUs to function. Are you proposing that these users are using black magic since Microsoft couldn't possibly support these closed, mysterious GPUs with their proprietary special sauce?
Let’s be realistic: Microsoft CANNOT support Windows on M-series CPU Macs, at least not until Apple wants is and actively supports it (maybe even pays it) and Apple doesn’t seem to have any intentions to do so.
This is mostly talking about running Windows on Arm via a virtual machine. While efforts like the Asahi Linux development show that it is possible to run alternate kernels on Apple Silicon Macs, it doesn't appear that Apple is willing to help with the hardware documentation to make that viable for most commercial uses. But from a VM, Apple already made their hypervisor and virtual machine APIs support paravirtualization and high performance access to hardware. Apple still needs to do more work on their APIs as there are still gaps but it sure looks like Apple is working hard on those frameworks. Recently with Monterey they have macOS Monterey guests working with paravirtualization of graphics hardware as a feature. Not exactly something that says that Apple isn't interested in support.
Ultimately, running a Windows 11 guest OS on the macOS hypervisor comes down to Microsoft. Apple has made it possible, now Microsoft has to at least license their OS to make it commercially viable.
Let’s be realistic: Microsoft CANNOT support Windows on M-series CPU Macs, at least not until Apple wants is and actively supports it (maybe even pays it) and Apple doesn’t seem to have any intentions to do so.
Why? Because the M1 chip, while having ARM standard cores, also has proprietary GPU cores and proprietary neutral cores. In short, M$ could only support a fraction of what the CPU has to offer, and would need to rely on costly reverse engineering for the rest, something they’d have to repeat with each iteration of Apple Silicon.
At this point, it’s not even clear if Apple isn’t forking the entire ARM architecture: Apple has introduced proprietary extensions that stand at odds with ARM’s recent extensions to the instruction set.
ARM pushes Ethos (NPU) and Mail (GPU), and as far as I understand Apple’s NPU and GPU cores aren’t superior implementations of the same architecture (as is the case for the CPU part of the M-series chips). So in essence, let’s be clear: Apple’s M-series is only ARM architecture in regards to the ever less relevant CPU part of the SoC, NPU and GPU cores are utterly proprietary, closed, undocumented in need of reverse engineering, designed to be accessed through Apple’s APIs and libraries.
Now tell me again how Microsoft could possibly support Windows on M-series hardware, except through an inferior emulation that’s performance limited because it can’t use the hardware to the fullest. There will of course be VMs that will run it, as whatever standard hardware M$ will support, can be emulated; but it will be “unsupported, use at your own peril”.
It’s also not clear if and when Apple will implement the new developments in the ARM architecture, or just roll their own extensions:
However, I've noticed Apple doesn't document much early on. The M1 SoC is in it's initial development phase, with more to come. My guess is Apple doesn't know where it's going with Apple Silicon, so don't want to be bound to support what would be prematurely documented.
Of course, it could be laziness!
The last thing us programmer types like to do is document our code. We're just happy to get the stuff working, then we're off to the next deployment.
Let’s be realistic: Microsoft CANNOT support Windows on M-series CPU Macs, at least not until Apple wants is and actively supports it (maybe even pays it) and Apple doesn’t seem to have any intentions to do so.
[…]
Ultimately, running a Windows 11 guest OS on the macOS hypervisor comes down to Microsoft. Apple has made it possible, now Microsoft has to at least license their OS to make it commercially viable.
I think a lot of us might be missing the point. It’ll force those us who need to run Windows into a Windows 365 licence. To be honest, I can see a lot of benefits for my use in being able to run my W11 installation from any browser on any device, rather than limit it to within Parallels, which has always felt a little bit clunky.
I remember the days when Macs were not Intel based, but we could run Intel software using an emulator (SoftWindows).
I see no reason why that concept couldn't be resurrected.
Emulation is always possible. For example, anyone should be able to run QEMU to run an x86 PC emulator on anything else. This can easily run Windows software (with a performance penalty). it should also be able to emulate the more generic ARM environment used by ARM builds of Windows.
Whether someone decides to package it (or another similar emulator) as a commercial product remains to be seen, but there's no technical reason why it couldn't be done.
That is not entirely correct. It's true that Apple's custom ARM SoC's are faster and less power hunger than ARM but that doesn't automatically that every other implementation of ARM SoC's are on the level of what Apple has come out with.
The ARM architecture covers a huge product space. It covers small microcontrollers that have only simple memory controllers (no paging or protection) all the way up to server-class CPUs with dozens of cores (and can theoretically support hundreds).
Apple has developed a really great SoC for general purpose computing which outperforms the SoCs used on comparable products from other manufacturers, but given the wide variety of ARM-based CPUs/MCUs/SoCs that exist, it really isn't a fair question to compare "ARM vs. x86", without getting more specific about which ARM platforms you're comparing to which x86 platforms.
The server market is almost certainly going with some ARM or other RISC variant - if for no other reason than to minimize data center energy use.
The server market, for the most part, has been running some variety of Unix for a very long time. Today it is dominated by x86 hardware running Linux. Not too long ago, it was running other Unix operating systems on hardware from Sun, HP, SGI, IBM, MIPS, DEC and many others, most of which have been RISC processors of some kind.
A move to a new high performance generation of ARM processors wouldn't surprise me in the least. Some data centers are already starting to integrate ARM-based servers. More will follow in the future as more and better products become available.
Will x86 disappear from that space? Probably not. They may become less popular in the future, but tech has a habit of never completely disappearing. Just like other RISC architectures have not disappeared completely (and some, like IBM's POWER architecture systems, are quite alive and well, even if they aren't dominating the market).
I guess it makes financial sense. Office 365 is where the money is and that is (or will be?) available for ASi Macs. Windows is free and it probably doesn't make sense to spend resources to make it compatible with ASi.
MS could go the route of actually charging for Windows on ASi, but I guess the numbers don't add up so it's not viable right now.
Windows is a $32 billion a year business for MS so it definitely isn't free. OEM's have to pay MS a licensing fee to install Windows on their machines.
Comments
The server market is almost certainly going with some ARM or other RISC variant - if for no other reason than to minimize data center energy use. Most servers don't have to use high power CPUs - they just need the ability to run a ton of discrete processes like file shares. You only have to walk behind a server rack to see that there are real cost savings to be had by reducing energy use. Future servers may still be administered by MCSEs, but more than likely they'll not be running on x86 hardware.
The consumer market is much less dependent on Windows legacy software and is more OS agnostic. Most users don't care what OS they're running and can be swayed by price/performance/energy efficiency, especially in the laptop segment of the market. The government may actually move in and mandate energy efficiency as they have in six states already, just as they have done in the past with automobiles.
On the enterprise side, a lot of enterprise software is moving from Windows client/server based to web based solutions due to the high cost of maintaining a whole fleet of Windows client machines at a reasonably homogenous level of software compatibility (and keeping end users from installing their own possibly vulnerable software). A lot of enterprises are replacing Windows client machines with linux based thin clients attached to network file servers for storage - much easier and cheaper to maintain and replace. This also means that end users can't necessarily come home and install pirated copies of the same software they use at work, further reducing the demand for home hardware compatible with what you use at work.
So ... the market isn't some big hulking thing you can hold on to with both hands and maintain a firm grip. The Wintel homogeny came to be for a lot of these factors which are fracturing into smaller categories, each with their own motivation and reason for being.
Why? Because the M1 chip, while having ARM standard cores, also has proprietary GPU cores and proprietary neutral cores. In short, M$ could only support a fraction of what the CPU has to offer, and would need to rely on costly reverse engineering for the rest, something they’d have to repeat with each iteration of Apple Silicon.
At this point, it’s not even clear if Apple isn’t forking the entire ARM architecture: Apple has introduced proprietary extensions that stand at odds with ARM’s recent extensions to the instruction set.
ARM pushes Ethos (NPU) and Mail (GPU), and as far as I understand Apple’s NPU and GPU cores aren’t superior implementations of the same architecture (as is the case for the CPU part of the M-series chips). So in essence, let’s be clear: Apple’s M-series is only ARM architecture in regards to the ever less relevant CPU part of the SoC, NPU and GPU cores are utterly proprietary, closed, undocumented in need of reverse engineering, designed to be accessed through Apple’s APIs and libraries.
Now tell me again how Microsoft could possibly support Windows on M-series hardware, except through an inferior emulation that’s performance limited because it can’t use the hardware to the fullest. There will of course be VMs that will run it, as whatever standard hardware M$ will support, can be emulated; but it will be “unsupported, use at your own peril”.
It’s also not clear if and when Apple will implement the new developments in the ARM architecture, or just roll their own extensions:
https://www.arm.com/why-arm/architecture/cpu
Ultimately, running a Windows 11 guest OS on the macOS hypervisor comes down to Microsoft. Apple has made it possible, now Microsoft has to at least license their OS to make it commercially viable.
However, I've noticed Apple doesn't document much early on. The M1 SoC is in it's initial development phase, with more to come. My guess is Apple doesn't know where it's going with Apple Silicon, so don't want to be bound to support what would be prematurely documented.
Of course, it could be laziness!
The last thing us programmer types like to do is document our code. We're just happy to get the stuff working, then we're off to the next deployment.
I think a lot of us might be missing the point. It’ll force those us who need to run Windows into a Windows 365 licence. To be honest, I can see a lot of benefits for my use in being able to run my W11 installation from any browser on any device, rather than limit it to within Parallels, which has always felt a little bit clunky.