MP 7,1 What does Apple's switch to ARM based processors mean for the future of the Mac Pro

[EdT]

I don’t think Apple is (completely) stupid. Their tendency towards policy secrecy has bit them in the recent past, but they have smart people running their departments.

If they are going to kill Mac OS in the near future then the entire MAC Pro project was a colossal waste of time and money for Apple. They could put up with the couple of years of some people bad mouthing them-which people do anyway- and just do minimal upgrades to all lines of Intel computers and put development mostly into the IOS line for the day when 95% of software is subscription and cloud based. At that point you are really just using your IOS as a terminal anyway, the power is in the networked servers, not the device the customer is using.

If that’s true why ditch 68x processors at all? Keep selling Laptops and desktops with the current architecture because after a fairly low point how many cores or what clock speed doesn’t matter. What matters is ether and internet speed and reliability.
And if it takes telecos or satellite networks 7-10 years the monkey is off EVERY computer makers back. It’s on the companies supplying the cloud software and hardware. Why should Apple, or any other company, spend a lot of time and development creating faster computers of any type- ARM or otherwise- when the device that actually runs the code isn’t the local computer?

I don’t really buy into the cloud based system. Right now there are large areas of the country that have terrible connectivity. I don’t think that will improve soon, and I don’t think anyone can accurately say when it will be acceptable performance no matter where you are.

And I think that is the reason everyone is still developing desktops. I don’t know if Apple is betting correctly, but I think they are trying to control a larger chunk of their future by designing chips not dependent on either Intel or AMD. Maybe they think that they are the 800 lb gorilla and what they say will go. Maybe they think that but they’re wrong.

 

[Azrael9]

I don’t think Apple is (completely) stupid.

And I think that is the reason everyone is still developing desktops. I don’t know if Apple is betting correctly, but I think they are trying to control a larger chunk of their future by designing chips not dependent on either Intel or AMD. Maybe they think that they are the 800 lb gorilla and what they say will go. Maybe they think that but they’re wrong.

It's about controlling their own destiny. Control.

And now they have the resources to do that.

£1.5 trillion company. I'd bet that can do whatever the hell they want to.

The Mac is currently inconsequential. Despite Apple Stores and iOS mindshare. (Both of which has probably helped sustain sales of 5 million Macs.)

Bringing them over to Mac ARM is a super nova moment. Having everything on the same A chip. No brainer.

Bigger screens? More cores. Higher clocks.

It's about destiny.

This time they ARE the 800 pound gorilla. Totally different in magnitude re: PPC vs INtel.

Azrael.
[automerge]1591892966[/automerge]

It's about controlling their own destiny. Control.

And now they have the resources to do that.

£1.5 trillion company. I'd bet that can do whatever the hell they want to.

The Mac is currently inconsequential. Despite Apple Stores and iOS mindshare. (Both of which has probably helped sustain sales of 5 million Macs.)

Bringing them over to Mac ARM is a super nova moment. Having everything on the same A chip. No brainer.

Bigger screens? More cores. Higher clocks.

It's about destiny.

This time they ARE the 800 pound gorilla. Totally different in magnitude re: PPC vs INtel.

Azrael.

"Now in its 31st year, WWDC20 will be the biggest WWDC to date, bringing together the global Apple developer community of more than 23 million in an unprecedented, virtual way, from June 22 to 26. Apple today shared the WWDC20 lineup, including keynote and Platforms State of the Union timing, and information on how developers can learn about the future of iOS, iPadOS, macOS, tvOS, and watchOS, and engage with Apple engineers."

They didn't have 'all a dis' for the last port over. The juggernaut's going in one direction and Mac is going to get dragged along with it.

 

[djjeff]

Keep in mind Intel once had a 2 year lead on fabrication technology which has now fallen to a 2 year lag. Essentially 10nm is late by 4 years already, I wouldn't call that temporary and they've already pushed back XEON processors that were meant to launch this year on 10nm to next year.

It's your opinion Intel will never move to 10nm or smaller process technology?
[automerge]1591895559[/automerge]

Yes, because it already is, no ?

On what data do you base this statement?
[automerge]1591895965[/automerge]

I agree with you that the blind trust in ARM superiority should be taken with a healthy dose of skepticism (which is quite funny, because just few years ago the popular opinion was that ARM would never go beyond washing machines controllers). But I also think that the entire story is much more complicated. First of all, the ISA does matter. ARM design is much leaner than x64, which greatly simplifies the instruction prefetch and decoding logics — this is where the precious chip area can go to something more useful. Second, I feel that ARM made some right choices when designing some of their instructions (e.g. the vector support — x64 is a total mess in that regard).

It appears history is starting to repeat again with the CISC versus RISC debate. PPC, a RISC design, was a good processor. Unfortunately it couldn't scale and ultimately Apple abandoned with in favor of a CISC design (which, many argue, is only CISC from its external view).

IMO CISC versus RISC is largely irrelevant these days.

I think what I am trying to say is that one does not have to be a veteran to be successful in CPU design. Whats more important is money (for R&D) and a good portion of luck (for making the right choices). It seems that Apple has both. Their CPU designs so far were truly impressive and so far unmatched by the rest of the industry. I wouldn't be surprised if they did manage to catch up and overtake stagnating Intel.

Which desktop CPUs are you referring to?

 

[EdT]

It's about controlling their own destiny. Control.

And now they have the resources to do that.

£1.5 trillion company. I'd bet that can do whatever the hell they want to.

The Mac is currently inconsequential. Despite Apple Stores and iOS mindshare. (Both of which has probably helped sustain sales of 5 million Macs.)

Bringing them over to Mac ARM is a super nova moment. Having everything on the same A chip. No brainer.

Bigger screens? More cores. Higher clocks.

It's about destiny.

This time they ARE the 800 pound gorilla. Totally different in magnitude re: PPC vs INtel.

Azrael.
[automerge]1591892966[/automerge]


"Now in its 31st year, WWDC20 will be the biggest WWDC to date, bringing together the global Apple developer community of more than 23 million in an unprecedented, virtual way, from June 22 to 26. Apple today shared the WWDC20 lineup, including keynote and Platforms State of the Union timing, and information on how developers can learn about the future of iOS, iPadOS, macOS, tvOS, and watchOS, and engage with Apple engineers."

They didn't have 'all a dis' for the last port over. The juggernaut's going in one direction and Mac is going to get dragged along with it.

I also think this is about control, But long term (and I don’t know if that’s a few years or a few decades) I do think, eventually, the mostly dumb terminal will be the computer most people are using. The infrastructure is nowhere near where it has to be for that to be true. Big cities, which is where most people are, have ACCESS to better service but people can’t afford to pay what ISP’s charge. And people away from big population areas can’t get it at all. The physical hardware isn’t in place. If I knew when cheap/fast/reliable service would be available coast to coast I would have a better idea how smart Apple is.

For right now, I think Apple wants to be able to control how their computers work so ARM it is. I can’t see them obsoleting Mac OS after spending time effort and advertising for the last few years, and Apple working on ARM powered computers hasn’t been a secret either.

But I could be wrong- maybe 3 years from now the MacBook Pro and Mac Pro computers will be dead as the dodo.

 

[goMac]

Every Mac other than the iMac already has an ARM CPU in it: the T2, which is based on the A10. It obviously doesn't increase the cost of Macs that much, especially when you consider the iPhone SE pricing with the A13. 'ARM Based' could also mean a 'T3' with more system functionality offloaded to it, with the addition of an x86_64 chip as a co-processor (which is essentially already the case today)

XNU can presumably arbitrarily execute in a mixed-architecture environment, especially considering mixed-arch fat binaries are already baked in to the OS. No need for any emulation.

T2 isn't a co-processor in that sense. It acts like an PCIe accelerator, like Afterburner. T2 doesn't allow the OS to execute ARM code.

XNU doesn't have mixed-CPU-hardware-architecture support. It can do things like executing 32 bit x86 on a 64 bit x86 CPU, but it doesn't support bridging two totally different CPU types into the same environment. That's tremendously complicated even from a hardware perspective.

 

[Nugget]

Windows runs on ARM so Virtual Machines can still be a thing. Even Boot Camp if Apple wants to.

This is a facile argument that completely ignores why people use virtualization and dual-boot capabilities. Even Windows users don't seem to want to run Windows for ARM.

People use Boot Camp, Docker, VMware, and Parallels because they need to run Windows applications. They don't use it because they want to run the Windows operating system (not to mention the people who are dual booting and virtualizing non-Windows operating systems).

macOS moving to ARM is the effective death of virtualization and dual booting. I wouldn't be surprised to see Apple decide that they're willing to lose those users and use cases, but let's be realistic about what it means for those users and the applications and systems they are using today by virtue of the x86/amd64 compatibility that macOS currently provides.

 

[Nugget]

It's becoming more of a niche to host the VM's on your own machine directly and much more common to use containers now, probably 10 to 1 on containers to VM's on-machine actually and of course containers won't have any issue with the processor switch as all the libraries and frameworks underpinning what most people use containers for will all get updated very quickly.

It's not really that simple, though. For the vast majority of users who are running containers on macOS, this fundamentally requires a VM underneath. Docker Desktop for Mac is relying on a Linux virtual machine running underneath so that users can make use of the current Docker image ecosystem.

I know that in my company we have developers who are building and running docker images locally all day long and who require the ability to run x86 docker containers. This works great in macOS and as a result about 80% of our developers are running Mac desktops or laptops. In a macOS ARM world this will become 0% because none of us will be able to do our jobs if we are only able to run ARM Docker containers locally. We require compatibility with our production systems which are (and will continue to be) amd64. If Apple switches to ARM we will all have to migrate to Windows or Linux.

 

[goMac]

I know that in my company we have developers who are building and running docker images locally all day long and who require the ability to run x86 docker containers. This works great in macOS and as a result about 80% of our developers are running Mac desktops or laptops. In a macOS ARM world this will become 0% because none of us will be able to do our jobs if we are only able to run ARM Docker containers locally. We require compatibility with our production systems which are (and will continue to be) amd64. If Apple switches to ARM we will all have to migrate to Windows or Linux.

Docker can be set up with QEMU so you can run ARM containers on x86 or x86 containers on ARM.

Performance will be slower, which might be a bad thing, but won't know the hit until there are more details about ARM Mac performance.

It looks like there is also work going on around multi-arch containers and cross environment support so by the time ARM is really being pushed into the higher end Macs this problem might be better solved.

I think as ARM picks up in the cloud, and on the Windows side (especially once Apple enters the fray and starts pressuring Wintel with better performance and efficiency) we'll see a lot more work go into solving a mixed ARM and x86 world. We're still in the early days of ARM on the desktop and server.

 

[Azrael9]

I also think this is about control, But long term (and I don’t know if that’s a few years or a few decades) I do think, eventually, the mostly dumb terminal will be the computer most people are using. The infrastructure is nowhere near where it has to be for that to be true. Big cities, which is where most people are, have ACCESS to better service but people can’t afford to pay what ISP’s charge. And people away from big population areas can’t get it at all. The physical hardware isn’t in place. If I knew when cheap/fast/reliable service would be available coast to coast I would have a better idea how smart Apple is.

For right now, I think Apple wants to be able to control how their computers work so ARM it is. I can’t see them obsoleting Mac OS after spending time effort and advertising for the last few years, and Apple working on ARM powered computers hasn’t been a secret either.

But I could be wrong- maybe 3 years from now the MacBook Pro and Mac Pro computers will be dead as the dodo.

The hardware will be legacy aka PPC. It will be 'good' for about 3-5 years in Apple update terms of in terms of performance the user can live with terms...whilst looking at a A16X blow past their legacy machine...

Just how many times have Mac fans said, 'Well, if not for Intel then Apple could...would...but...etc...'

Now Apple will dictate that. No Intel excuses.

Spiritually, isn't this what we all wanted? An Apple CPU in an Apple Mac torching the unholy hoardes of darkness? (M$/Intel etc...)

Back to scorching Intel bunny Men suits...

Azrael.

 

[dspdoc]

The hardware will be legacy aka PPC. It will be 'good' for about 3-5 years in Apple update terms of in terms of performance the user can live with terms...whilst looking at a A16X blow past their legacy machine...

Just how many times have Mac fans said, 'Well, if not for Intel then Apple could...would...but...etc...'

Now Apple will dictate that. No Intel excuses.

Azrael.

You are having quite the field day blasting away here at MP aren't you? You must not own one...

 

[Azrael9]

You are having quite the field day blasting away here at MP aren't you? You must not own one...

It's not me you have to worry about.

It's Apple.

Own one? For 40k? I don't think so. I won't be crying crocodile tears for the passing of INtel...

Maybe if Apple offered a rationally priced one, you know, like they used ta? £2.5k. And that aint 'cheap.'

*Proud owner of a 1997 PPC Tower. 604e chip. I also have a G4 Tower.

MP. Over priced. Out of date. Under powered at £6k. A joke. 'Whitcha 'must not own one...'

*rassssssssp.*

Azrael.

 

[b0fh]

It means you need to start thinking about either moving your hack to Windows or purchase a real Macintosh. The Hackintosh movement has about 2-4 years left before it disappears. Ultimately, it depends on how long Apple supports macOS on Intel, but it won’t be more than 5 years and probably less.

Or... it means raspberry pis are going to be a lot more fun to play with...
[automerge]1591907069[/automerge]

Apple's done this transition twice before. Both times they had a built-in emulator to run the previous architecture's applications. A 68K emulator and a PPC one. I would assume Apple will have to do something similar here.

My 2nd thought is virtualization. I use VMWare heavily for previous versions of the macOS and Windows. A move to ARM will mean this kind of usage will switch to emulation, rather than virtualization. No one's done Windows emulation on the Mac in around 15 years. Connectix is long gone. Will VMWare or Parallels get into the emulation business? Will a new company emerge? Will they want to add, and will Apple allow them to emulate Intel versions of macOS?

VMware already killed their Fusion team (the original team was 15-20 people).

VMware axes Fusion and Workstation US devs

The show will go on, says Virtzilla, probably in China. Or how about open source?

www.theregister.com

 

[b0fh]

Has everyone forgotten about Wine? "Wine Is Not an Emulator" has been around for ages, and allows you to run windows application in linux, or macOS, without virtualization, and without emulation.

It translates Windows API calls into POSIX calls on the fly. People have taken to running Wine *IN* Windows as well.

WineOnWindows - The Official Wine Wiki

I'll bet you 10 to 1 that Apple has something similar in the works, where it takes a macOS API calls and translates it into ARM calls on the fly, and 99% (number pulled out of my ass, yo) of the current applications would run without knowing what happened.

Anything running intensive graphics stuff such as Metal and so on probably wouldn't work. Anything that calls the CPU directly probably wouldn't work. Everything else? Easy peasy.

 

[goMac]

VMware already killed their Fusion team (the original team was 15-20 people).

VMware axes Fusion and Workstation US devs

The show will go on, says Virtzilla, probably in China. Or how about open source?

www.theregister.com

VMWare has been working on ARM versions of their offerings. They'll have a virtualization engine for ARM, and Apple's already added an ARM version of their Hypervisor framework to iOS.

If VMWare wants to add in x86 emulation that's on them. There are existing solutions like QEMU. But they'll at least be fine virtualizing ARM OSs like Linux and Windows-on-ARM.
[automerge]1591907793[/automerge]

Has everyone forgotten about Wine? "Wine Is Not an Emulator" has been around for ages, and allows you to run windows application in linux, or macOS, without virtualization, and without emulation.

It translates Windows API calls into POSIX calls on the fly. People have taken to running Wine *IN* Windows as well.

WineOnWindows - The Official Wine Wiki

I'll bet you 10 to 1 that Apple has something similar in the works, where it takes a macOS API calls and translates it into ARM calls on the fly, and 99% (number pulled out of my ass, yo) of the current applications would run without knowing what happened.

Anything running intensive graphics stuff such as Metal and so on probably wouldn't work. Anything that calls the CPU directly probably wouldn't work. Everything else? Easy peasy.

You mean like Rosetta? Yes, probably.

Rosetta (software) - Wikipedia

en.wikipedia.org

 

[dspdoc]

It's not me you have to worry about.

It's Apple.

Own one? For 40k? I don't think so. I won't be crying crocodile tears for the passing of INtel...

Maybe if Apple offered a rationally priced one, you know, like they used ta? £2.5k. And that aint 'cheap.'

*Proud owner of a 1997 PPC Tower. 604e chip. I also have a G4 Tower.

MP. Over priced. Out of date. Under powered at £6k. A joke. 'Whitcha 'must not own one...'

*rassssssssp.*

Azrael.

They don't cost $40k. Nowhere near that price unless you start stacking on things from Apple that you can get at half the cost elsewhere.

 

[deconstruct60]

Every Mac other than the iMac already has an ARM CPU in it: the T2, which is based on the A10. It obviously doesn't increase the cost of Macs that much, especially when you consider the iPhone SE pricing with the A13. 'ARM Based' could also mean a 'T3' with more system functionality offloaded to it, with the addition of an x86_64 chip as a co-processor (which is essentially already the case today)

If there is an update to T3 it will probably be along the line of providing better security and better AI ( especially when that is interwined with biometrics ). It wouldn't be to run more user level apps or "ease porting" of iOS ( or iPadOS ) apps.

The current T2 is hooked to (in addition to the primary drive ):
i. microphone , speakers and camera, (touch ID sensor were present)
ii. system management controller chip (SMIC) .
iii. fans.
iv if present the touch bar.

for the first group..... If want a "smarter" Siri that does more processing locally (i.e., privately ) then adding the tensor units of the current A13 to the T3 would make sense. Likewise the image processing fixed function units; bringing those up to date would make things more uniform across wide rage of products with similar cameras. Don't really need that much of the general CPU cores or quite as much of the A13. Can shrink those a bit to save some costs. If Apple wanted to attach "FaceID" to some Mac then those two SoC subunits would help.

System management and cooling control. Apple could be weaving more of that into a T3 and perhaps make the custom Mac SMIC go away.

The GPU needed to drive the touch bar ( which is basically just copying the frame buffer and pushing it over

The supposed "huge gap" that the new A13-A14-A14x have on the Intel processors could be relatively easily leveled out with simply just an update to the T-series security process.

And frankly even if Apple moves to ARM as the main host processor having a discrete T-series processor to manage security would still be a good idea from a security perspective. If going to have DMA access ports ( Thunderbolt ) , lots more random devices plugged , an open enclosure where folks can insert random drive , etc. Then the security running on the same CPU and RAM is probably not quite as good as "air gapping' . Yes the A13 has a small security processor built in but it also doesn't have anywhere near the depth and breath of attack vectors either.
[ Intel is the only one having problems with side channel.
https://www.tomshardware.com/news/side-channel-attack-hits-arm-cpu ]

Side channel vector doesn't matter if don't let anybody untrusted run code there. The T2 is on the edge of doing "too much". it really doesn't need anything more broad added to its workload.

XNU can presumably arbitrarily execute in a mixed-architecture environment, especially considering mixed-arch fat binaries are already baked in to the OS. No need for any emulation.

Might be able to attach some QEMU sidecar to run virtualize instances that think they are looking at a different architecture ( i.e, presented with a virtual machine to run against). But the kernel juggle multiple cores with different instruction sets at the same time and assigning arbitary proceses to resources ? No .

The kernel scheduler needs to get in and out fast. There shouldn't be time to look at processes and say "are you one of this kind and maybe I can assign you to that core". That is just too much junky overhead. That's substantially different from "low" versus "high" resource consuming threads/processes that can be assigned to "big" or "efficiency/small" cores. Doing 'fair scheduling' is something the scheduler has to do so has to do some measuring for 'resource hogs' anyway.

 

[theluggage]

Has everyone forgotten about Wine? "Wine Is Not an Emulator" has been around for ages, and allows you to run windows application in linux, or macOS, without virtualization, and without emulation.

...but WINE still can't run x86 code on an ARM (at least, not without going through an emulator such as QEMU and taking a performance hit). The best bet for Windows on an ARM Mac would be to run ARM Windows and use the x86 emulation/translation feature in that.

 

[theluggage]

It's taken them YEARS and after all that time what they released on iOS/ARM isn't even half of what Photoshop is on Mac/Intel.

That's not necessarily relevant.

Porting a MacOS app to iPadOS/iOS requires a major design and re-write of the user-interface design and code to make it work well on a touch screen, other significant changes to adapt to the strict limitations on what Apps can do, not to mention the (relatively) limited storage and RAM on an iPad (most Adobe CS users would laugh at the idea of running CS in 6GB RAM...)

Porting from MacOS x86 to MacOS ARM is potentially far simpler: the vast majority of the code will be written in C, C++ or ObjC which simply needs re-compiling (past changes have required switching from big-endian PPC to little-endian x86, 32 to 64 bits etc. which can potentially break source code - there's none of that with a switch to ARM 64). All that needs to change is any hand-optimised x86 code or direct use of x86 SIMD instructions etc. - and frankly that shouldn't be going on in modern software anyway. Nobody can really say how big the job is unless they're a programmer at Adobe and know what sort of mess the code is in, but whatever challenges do exist you can double them once you add iOS to the problem.

 

[goMac]

That's not necessarily relevant.

Porting a MacOS app to iPadOS/iOS requires a major design and re-write of the user-interface design and code to make it work well on a touch screen, other significant changes to adapt to the strict limitations on what Apps can do, not to mention the (relatively) limited storage and RAM on an iPad (most Adobe CS users would laugh at the idea of running CS in 6GB RAM...)

Porting from MacOS x86 to MacOS ARM is potentially far simpler: the vast majority of the code will be written in C, C++ or ObjC which simply needs re-compiling (past changes have required switching from big-endian PPC to little-endian x86, 32 to 64 bits etc. which can potentially break source code - there's none of that with a switch to ARM 64). All that needs to change is any hand-optimised x86 code or direct use of x86 SIMD instructions etc. - and frankly that shouldn't be going on in modern software anyway. Nobody can really say how big the job is unless they're a programmer at Adobe and know what sort of mess the code is in, but whatever challenges do exist you can double them once you add iOS to the problem.

Right. The holdup with porting Photoshop to iOS is the UI, not the CPU. iOS and Mac don't use the same application framework. It's like if they were doing a fresh port from Windows to Mac or something.

 

[Azrael9]

They don't cost $40k. Nowhere near that price unless you start stacking on things from Apple that you can get at half the cost elsewhere.

Ye-eah...we * know * that.

It's *only* £6k for a rough spec with highlights such as? ...an 8 core with a farcical low end gpu.

Not like I couldn't get a PC tower for 1/5 of the cost that couldn't out perform it...

Azrael.

 

[oneMadRssn]

That's not necessarily relevant.

Porting a MacOS app to iPadOS/iOS requires a major design and re-write of the user-interface design and code to make it work well on a touch screen, other significant changes to adapt to the strict limitations on what Apps can do, not to mention the (relatively) limited storage and RAM on an iPad (most Adobe CS users would laugh at the idea of running CS in 6GB RAM...)

If that were true, then why is Photoshop for iPad still missing a lot of functionality? It's certainly not they couldn't find a place to fit the buttons or sliders. And while RAM may be a limiting factor, competitors to Photoshop on iPad show us that all the functionality is doable. So the only reason they haven't gotten there yet must be that porting the actual coded functions is not as simple as you make it sound.

Porting from MacOS x86 to MacOS ARM is potentially far simpler: the vast majority of the code will be written in C, C++ or ObjC which simply needs re-compiling (past changes have required switching from big-endian PPC to little-endian x86, 32 to 64 bits etc. which can potentially break source code - there's none of that with a switch to ARM 64). All that needs to change is any hand-optimised x86 code or direct use of x86 SIMD instructions etc. - and frankly that shouldn't be going on in modern software anyway. Nobody can really say how big the job is unless they're a programmer at Adobe and know what sort of mess the code is in, but whatever challenges do exist you can double them once you add iOS to the problem.

Again, read this thread: https://threadreaderapp.com/thread/1270465602570686464.html

Its written by someone who knows a lot more about this than you or me, and indicates that the switch for software developers will be a lot more burdensome than merely recompiling.

 

[deconstruct60]

Read this: https://threadreaderapp.com/thread/1270465602570686464.html

There is A LOT more to it than "just recompile."

That is a hugely overly complicated thread that basically ignores some obvious things and lots of arm flapping to rehash justifying the choices that Windows ARM made. Mostly the latter. It is doubtful Apple is going to

The thread largely invents to goofy premise that macOS on ARM wouldn't largely be exactly the same set of API that are already there. ( minus OpenGL I think that this a probably really , really , really wrong). Apple has zero reasons to completely throw away a huge chunk of the macOS API mostly because Apple incrementally tossed "old" APIs over time on a regular basis. It 'doesn't have the hugely overly constipated problem that Windows has which won't throw away anything because have to keep "everything". Apple regularly on 3-6 intervals tells folks "interface XYZ is deprecated and we're in the process of dumping it. Get ready to move on"

Windows is the opposite. It suppose to be great because can still call you 1997 era Win32 API call from days gone by . So yes. Windows needed an huge excuse to do a "regime change" on the APIs and used ARM as a wedge to push that out.

Apple told folks to get off 32 bits when on Intel. Apple has been telling for for the last 2-3 years to get on board with Metal and in WWDC 2019 told folks that OPenGL/OpenCL deprecated. (end of the line coming. you need to port to Metal). So is it really going to be a surprise that at WWDC 2020 they will possibly crank that up higher? Nope. Pretty good chance macOS 10.17 OpenGL is axed concurrent with the arrival of Apple's GPU (which never had a full OpenGL stack; just the 'ES' subset).

Apple moved to Mac OS X with Carbon and Cocoa. Catalyst really isn't the "new" API he is trying to make it out to be. It is far more so aimed at closing the 'fork' that developed in Cocoa between the "CocoaTouch" and 'classic' Cocoa'. Some foundational classes drifted apart. The UI code drifted apart a bit. And Apple is mainly doing a parital 'join' / merge from that drifting fork. For any decently substantive app you still have to write/compose a "touch" and "mouse pointer" separate GUI part of the application but the common model stuff doesn't have to have somewhat quirky gaps.

It doesn't makes about sense to go Catalyst "only" because SwiftUI (and chucking OpenGL ) is the bigger change. They basically will have a "close to the old way" ( before Carbon now Catalyst ) and "new way " ( before Cocoa and now "SwiftUI + Metal + updated foundation" ). Plus maybe a emulator like "classic" ( to run Mac OS 9 ... whereas this time something with OpenGL apps run again a weak virtual GPU. ). But the "Classic" sandbox will probably be on both the Intel and ARM version. [ roughly similar to the sandbox Windows has for running DOS. ]

In short, for the top level apps it pretty likely to be a recompile because Apple has been setting the stage for the folks who got on board with Metal and went 64-bit clean. The folks who have been saying "Screw Metal" for the last 2-3 years... yeah they are in trouble..... but they have pragmatically already been in trouble (because not listening or looking at the neon signs Apple has up about their commitment to Metal. )

The lower level driver stuff? That part is not going to be a recompile and likely will be a big disruption. But again Apple announced that loudly at WWDC 2019. The warning went out over a year ago. If the hammer drops in 2021 that's a two year heads up.

 

[Adult80HD]

VMware already killed their Fusion team (the original team was 15-20 people).

VMware axes Fusion and Workstation US devs

The show will go on, says Virtzilla, probably in China. Or how about open source?

www.theregister.com

Not accurate. They moved the dev offshore. That article is four and half years old on top of that. FAKE NEWS.

 

[dspdoc]

What in the world does this mean?!?! I have never seen so many new high end machines show up this quickly on the refurb page!

 

[Adult80HD]

SNIP

In short, for the top level apps it pretty likely to be a recompile because Apple has been setting the stage for the folks who got on board with Metal and went 64-bit clean. The folks who have been saying "Screw Metal" for the last 2-3 years... yeah they are in trouble..... but they have pragmatically already been in trouble (because not listening or looking at the neon signs Apple has up about their commitment to Metal. )

The lower level driver stuff? That part is not going to be a recompile and likely will be a big disruption. But again Apple announced that loudly at WWDC 2019. The warning went out over a year ago. If the hammer drops in 2021 that's a two year heads up.

The problem here is that there are a LOT of major apps that aren't fully moved onto Metal, etc. Apple's trying to force their hand, but they're just not there. The ones that are tend to be smaller, newer companies. Some of those apps are pretty good, but it's not easy to get people to switch wholesale. There's momentum behind the big names like Adobe, so you won't find a ton of people easily switching to say, Affinity, even if it's become very capable and leverages Metal very well. I mean Apple used Affinity as an example of a "Pro" app at their Mac Pro launch...but I honestly don't know any professional users that have switched to Affinity from Photoshop. I'm sure there's some...but it's not the majority.

I think Apple even knows this and they're playing a longer game. This isn't the PPC-->Intel switch, they *don't* have to make it all or nothing. Just introduce an ARM MacOS that runs all of their Apple applications beautifully and also easily runs anything written for iPadOS only faster. I wouldn't be surprised if iPadOS just becomes ARM MacOS. They'll have an emulator to let most x86 MacOS apps run fine, but if you want them to work really well, then Apple will happily sell you a "Pro" machine--a MacBook Pro, an iMac Pro, a Mac Pro. Over time as companies like Adobe slowly drag their way to a full-blown ARM version then Intel x86 MacOS may or may not go away, depending on the economics for Apple and what Intel does. But I think a new line of entry-level ARM MacOS devices will be here to stay.