Mac With Apple-Designed Arm Processor Coming in First Half of 2021

[Rafterman]

Maybe the ARM units will be produced, in addition to continued Intel units, as a lower cost alternative. No reason to assume the ARMs are replacing the Intels outright.

 

[bsolar]

The thing for x86 emulation is Intel licensing for their proprietary instruction set architecture, which they decide to make Apple's life difficult, it will suck.

I don't think the instruction set itself is protected, but some implementation patents at least in amd64 likely still are. I'm not sure they apply to an emulation or translation layer though. For sure examples of such solutions already exist:

• Windows on ARM has an x86 to ARM translation layer which allows x86 applications to run on ARM without the need of being recompiled.
• QEMU supports emulation for a lot of different architectures, including x86-32 and amd64.

 

[cmaier]

Maybe the ARM units will be produced, in addition to continued Intel units, as a lower cost alternative. No reason to assume the ARMs are replacing the Intels outright.

In the end Apple will have to discontinue Intel, in order to make sure that the software industry targets ARM. As long as intel still exists, software companies have less reason to make sure theIr stuff works with ARM

 

[Stephen.R]

A small percentage of these are sold to people running windows and Linux.

I think you're greatly underestimating how popular Macs are in the development community.

 

[Rednow]

Guys are we getting 10th Gen Intel CPUs for an updated 16” mbp in 2020? What is the probability we will?

 

[oneMadRssn]

I've just read some comments on the "Apple switching to Intel" thread from 2005. Yours will also appear in the history books as we look back on this thread in 15 years.

History is nice to know. But this time is materially different and looking back to 2005 is of little value. The PPC-x86 switch was a switch from a unpopular minority architecture to the popular majority architecture. The switch actually increased the amount of software available to Macs literally overnight. Almost every PPC app already had an x86 alternative before Apple even made the switch.

A switch from x86/x64 to ARM is not like the last time because it is not a switch to a popular majority architecture. Looking at the desktop/laptop market, ARM is the unpopular minority architecture. The switch will decrease the amount of software available to Macs in the short term. There are a lot of x86/x64 apps that do not have an good alternative for ARM today.

Another reason this time is different and thus analogies to the past aren't of much value is the broader market is different. Back during the PPC-x86 switch, there was no huge mobile market. Today whole x86/x64 is the popular majority architecture on desktops/laptops, ARM is the popular majority architecture for mobile, tablets, and IoT. Will the upcoming rumored switch to ARM benefit from that mobile market? Nobody knows, because there is no historical equivalent which we can consider.

 

[Emanuel Rodriguez]

I don’t think so as anyone NEEDING to use an old Intel app can just keep their current system. No translation required.

Right. You try explaining to users why their old Intel apps can't run on their shiny new MacBook Pro. What if their current laptop is 10 years old and they desperately need an upgrade? Sorry, can't do it. Your apps won't run. Buy a PC. Ridiculous.

Apple has always used and probably always will use translation layers when porting between architectures. They did something similar with the 68k to PowerPC and PowerPC to Intel transitions, and they will do it from Intel to ARM.

 

[MikeZTM]

AFAIK Apple can't make an x86 chip without a license from Intel, which probably ain't gonna happen. AMD have a license for historical reasons - back at the birth of the IBM PC, IBM required Intel to licence a couple of alternative manufacturers as a "second source" to ensure their supply.

Considering that Apple use ARM for both iOS devices and in the T2 chips it is by far the most likely choice.



Yeah - hackers had Windows running on Intel Macs before bootcamp came out. Bootcamp has always been, mainly, a point-and-click tool for automating something a masochist could do manually, plus a couple of drivers for things like trackpads.



...last time I looked, you couldn't buy ARM for Windows as a product - you had to get a MS-approved device.

I rather doubt many Bootcamp users want Windows for ARM anyway - if you need to run obscure Windows-only software or do testing you probably want x86... Windows is hamstrung by the need for backwards compatibility - MacOS can be a bit more agile in switching platforms (which its done several times already).

You can get Windows for ARM from retail -- Windows for ARM is not a special version. It's just a CPU arch SKU just like x86 and x64. Your retail Pro/Home key will activate Windows 10 fro ARM.

 

[bsolar]

A switch from x86/x64 to ARM is not like the last time because it is not a switch to a popular majority architecture.

It depends on how you segment the market. If you separate desktop/laptop from mobile, sure, but if you include them together, on mobile ARM is basically the standard.

Furthermore, nowadays desktop applications are often made in frameworks which would make a migration easier. Think Xcode/Swift, think Electron applications etc...

 

[oneMadRssn]

Legacy software is the problem.

Open source software is the least problem - it just needs one person, anywhere, with some development experience to grab the software and re-build it for ARM and submit any patches they make - in many cases is going to be just a case of re-compiling. On top of that, a lot of open-source software was written for Unix/Linux which has a long tradition of architecture-independence since Linux is already supported on ARM32, ARM64, PPC, x86 etc. Most of it already runs on Linux for ARM64 (go look at the current state of the main Linux distros for ARM, such as Raspbian) so the only problem cases are applications which have x86 dependencies in their Mac-specific code which should be fairly rare.

The only serious recent past attempts at a mass-market ARM PC (unless you go back to the 80s/90s and the original Acorn systems) have been Windows RT and the Raspberry Pi.

Windows has a huge issue with "legacy" software of the "our Fourtune 100 company relies on binaries last compiled in 1996 with an obsolete COBOL compiler" variety (yeah, 1990s code still runs under the 32 bit version of Windows 10) - beyond the worst nightmares of Mac.

Windows RT was a knobbled version of Windows restricted to "modern" software in Microsoft's App store - alongside which Apples Mac App Store (let alone the iOS/iPadOS store) is a veritable cornucopia. The latest attempt to do Windows on ARM is still early days...

Microsoft are hamstrung by their monopolistic past - as soon as they compromise backwards compatibility, they're competing on a level playing field with Mac, iOS, Android and Linux... and everybody hates Microsoft. Apple have far more flexibility without the corporate albatross around their neck which they've already used to switch architecture 2-3 times as well as switch from Classic MacOS to the completely different NextStep OS X...

The Raspberry Pi has been something of a success, and many of the usual Linux Apps have been "ported" (in many cases, minor tweaks then re-compile) with problems more likely down to the Pi's limited resources and flakey I/O rather than ARM.

If it was made for Linux then, chances are, it's already been ported to ARM (e.g. k3b is available on the Raspberry Pi - although the Pi is cursed with no SATA and a shonky USB implementation so you probably wouldn't want to burn DVDs on it, but that's not because ARM).

Apple could lock down MacOS on x86 tomorrow if they were so inclined, more so once they've replaced their last remaining pre-T2-chip system, the iMac. Moving to ARM would be an excuse for doing that, not a reason.

Most ChromeBooks are x86, and a lot of the code they run is processor-independent. Google have dumped the idea of "ChromeOS Apps" in favour of a 3-way split between (platform-independent) web apps, Android compatibility (mostly CPU-independent bytecode) and sandboxed Linux binaries (with many Linux Apps already buildable for ARM).

Thanks for the detailed reply, I appreciate it.

My knowledge of software development and architectures is enough to be dangerous, but shallower than the experts. So I didn't appreciate that about the open source software until you said it.

I think should watch to see what Microsoft does with the Surface Pro X and Windows 10 on ARM, and how the market reacts to those product lines. Not just this year, but over the course of a few years. Microsoft has been ahead of its time in the past, and then gotten it right later. (In particular, I'm reminded of Windows XP Tablet PC Edition before and Microsoft Surface Pro today). Maybe this time will be similar. Either way, I think it will be a good indicator of the issues or successes Apple will see in their transition to ARM.

 

[hans1972]

Except it's not going to run most of the apps that people use. Not as you know them. Perhaps ipad lite versions? You may as well use an ipad..

Of course it can. Either by the developer doing a recompile or by emulation.

 

[Stephen.R]

think Electron applications etc

There is zero chance Apple's "answer" to the issue of applications that won't run on Arm is "just use Electron".

 

[bsolar]

There is zero chance Apple's "answer" to the issue of applications that won't run on Arm is "just use Electron".

No, you missed the point. There are already applications using Electron. For those applications a migration to ARM would be a non-issue.

 

[MikeZTM]

FYI most x86 and AMD64 parents expired decades ago, Early 2000s to be exact, AMD actually licensed it's Zen platform to THATIC who sells these in the Chinese domestic market, but due politics this agreement comes to an abrupt end.

Not just AMD manufactures x86 CPU, via technology also does (wo need to license Intel or AMD), Zhaoxin is another manufacturer of x86-amd64 compatible CPUs who never licensed a single patent.

Zhaoxin get it from VIA.
And VIA got it from Cyrix.

 

[hans1972]

ever tried to port a complex application across cpus? Especially from a cisc to risc platform? Its rarely cost effective to do port from windows to Mac even on intel. Across cpu families, it's likely no cost effective At all. It'd also be the end of virtualization of windows, so you can count parallels and fusion out too.

if It's a dual path, fine. A complete replacement would, yet again, reflect how little Apple understands the real world. And unless they come out with a very clear, explicit dual path statement, even the release will create a lot of doubts. The corporate bean counters wouldn't take the risk on buying assets and building infrastructure that may go away in 2-3 years.

When you program a Mac application, you do not program for the CPU. All that is hidden by the operating system. You are making a macOS application, not an Intel CPU application.

So you are looking for a macOS -> macOS transition which in many cases will just be a recompile. And I believe that for all the applications in the Mac App Store, Apple will compile it without the developer doing anything.

 

[bsolar]

And VIA got it from Cyrix.

But Cyrix originally didn't get any license: they reverse-engineered x86 and defeated Intel in court over it. Then later they sued Intel over the first Pentiums violating their patents. The latter case is what led Cyrix and Intel to settle through cross-licensing.

 

[Stephen.R]

Either by the developer doing a recompile or by emulation.

Have you ever used something doing emulation of another CPU architecture? It's not pleasant.

Ever wondered why the iOS Simulator doesn't simulate an ARM CPU, it simulates iOS on x86?


Not to mention any number of tools that rely on hardware features, and then those that rely on hardware acceleration of features.

One of the huge benefits of the Intel shift, is that Windows users can make the jump to a Mac, but still have options for specialised/legacy Windows apps they need to run: either via Bootcamp, a hypervisor like Parallels or VMWare, or in some cases even via tools like Wine.

It's arguably easier to migrate from Windows on x86 to a Mac, than it is to migrate to Windows on ARM.

 

[theluggage]

This is the only option. Millions of macs are sold for developers needing access to Windows, Docker, Linux, etc.

Although an overnight change would be a problem, over a 3-5 year timescale it could be a diminishing issue.

- web/webapp development - and using web technologies (HTML5/Javascript) to write 'desktop' apps - is increasingly important, and predominantly platform independent - most of the main Linux distros have ARM versions, Apache, Node.js, Ngenix, Mysql, PostgreSQL, Mongo, PHP, Python and, yes, Docker etc. are all ported.

- Linux/Unix application development is traditionally architecture independent and Linux on non-x86 processors is increasingly important. Lots of interest in Linux in the server world (because power consumption = $$$) - so building and testing your Docker images for both ARM and x86 is likely to become a thing. Meanwhile, why try and set up an approximation to your target environment on a Mac, when you can spin up an exact clone of your target in the cloud?

Meanwhile, Ding! Dong! Internet Explorer is dead, "original" Edge never really lived and MS is now using Chromium. Testing stuff on IE was a major reason for wanting Windows on Mac which is now going away. (You still need to test on as many platforms as possible, of course, but not the sort of continuous testing you needed to deal with IE quirks - last time I tested a web app in Parallels it turned into a wild goose chase because of a bug in Parallels - and there's no way of testing stuff on a Windows touch screen),

- modern Windows 'native' development is predominantly using .net (CPU-independent bytecode) and since MS are having another try with Windows on ARM (which could theoretically run on an ARM Mac) if that succeeds you'll need a way of testing that, anyway.

- the legacy software issue could be diminishing, too, as concern over security makes it less acceptable to be running decades-old 'abandonware'.

- also mobile development is more and more important and that means targeting and testing Apps on ARM and websites on iOS/Android browsers. I wouldn't place so much weight on this one in light of the above points - mobile is largely architecture-independent - but it is there.

Yes, Apple could stuff it up by prematurely discontinuing x86 Macs (or letting them get hopelessly out of date) and forcing people to ARM before they are ready... but seriously, folks, CPU-dependent code in Apps is so 2005 and just as obsolete as USB-A, OpenGL, 32 bit and CUDA (...and, yes, that was a double-edged statement).

 

[MikeZTM]

What do Vmware Fusion have to say about this? I run a Windows VM all day long for work. Windows on one display, MacOS on another. I'd hate to give up that way of working.

Hypervisor | Apple Developer Documentation

Build virtualization solutions on top of a lightweight hypervisor, without third-party kernel extensions.

developer.apple.com

Parallels MacAppStore version is already using this.
This is Apple's "KVM".

This feature was added in 10.10 so 5 years ago and I do not think it will go away on ARM Mac.

VMware Fusion can just use this feature to implement their VM just like how Parallels did today and they still need to build their drivers for VM client (Including GPU drivers).

 

[incoherent_1]

I know nothing about hardware design. Nothing. I’m a software guy.

But I’ve heard some very intelligent people (micro-architecture specialists) say that x86/64 is much more sophisticated, such that even a seemingly-more advanced ARM processor can’t replicate the multitasking, parallel computing required by a full desktop operating system.

Meanwhile, I hear equally smart people saying they’re wrong and ARM is the future.

Which is it?

 

[Unregistered 4U]

They dropped 32bit apps like a cancerous tumor, and basically said, “too bad,” to those that continued to rely on a certain app that won’t be upgraded.

And, if anything, this gives a strong indication of how they feel about this transition. If you are an average Mac user and don’t know anything about emulation or virtualization (the vast majority don’t), then your new Mac won’t run your old software and Apple’s not going to help you.

Yes pesky unreliable chip partner Intel. That must be why it took 4 years to update the Mac Mini...

Nah, that was because the Mac Mini is unimportant to Apple.

You gain some speed but lose freedom; is that worth it?

When you consider that the vast majority of folks buying mobile computing devices today are buying iPads, the vast majority thinks it IS worth it.

"If you really need Windows...buy a Windows machine".

Oh No! I just bought a 7,1 Mac Pro... what will happen with bootcamp or compatibility with windows ?

Fortunately, releasing new systems with new processors won’t affect your Intel processors. They’ll continue to be able to do what they do for the useful lifetime of the device. They may stop receiving OS updates, but folks using Cheesegrater Mac Pro’s today will tell you it’s definitely not the end of the world!

Catalina, the latest version of the OS, is widely derided right now. Catalyst, the system for getting iPad apps on the Mac, has also not worked out especially well so far (to put it mildly).

Catalina is doing the work of culling the herd ahead of an ARM transition. Microsoft doesn’t have that luxury as they have to be everything to everyone. Apple just has to run macOS on ARM as fast or faster than Intel.

This is the only option. Millions of macs are sold for developers needing access to Windows, Docker, Linux, etc.

And those millions of Macs will continue to be used. Folks hold on to their systems for a long time.

I think you're greatly underestimating how popular Macs are in the development community.

I think it’s more that the “development community” is mightily dwarfed by “everyone else using Macs”.

You try explaining to users why their old Intel apps can't run on their shiny new MacBook Pro.

That is happening RIGHT NOW. New Catalina loaded laptops won’t run their old Intel apps. Does Apple provide a translation layer? NO. They’ll just have to return that new system to the store and look for a gently used model that has an older OS installed. AND they’ll get a better price!

What if their current laptop is 10 years old and they desperately need an upgrade?

If their current laptop is 10 years old, then that means their needs stopped growing 10 years ago. The ONLY reason for them to upgrade is because their old system breaks and, if so, I’m sure that there will be systems available on the used market that will suit them.

 

[Marekul]

Of course it can. Either by the developer doing a recompile or by emulation.

Emulation introduces vast amount of overhead. Also try to emulate hardware drivers. Good luck.
Then try recompile some code for another architecture... it is not at all trivial even for simple applications.
On top of that with the increasingly closed nature of apples platform and rising prices for their hardware - open source software will dry out pretty quick on macs.

My bet is ARM macs will be locked down exactly like iOS.

 

[Stephen.R]

There are already applications using Electron.

And their performance is ****ing abysmal. They are evidence of why web technology is not a good choice for desktop applications, not evidence that some great swath of Mac applications is going to be available on Day 1.

Regardless of that, in a hypothetical "Apple adopts ARM for macOS" world, an Electron App vs an app built using any of the languages Xcode supports is worse off.

Xcode is maintained by Apple, who would obviously have a version available to compile macOS on Arm targets way ahead of time.

With Electron, you have to wait for the Chromium project to support macOS on Arm. And then you have to wait for the NodeJS project to pick up those changes and ship for macOS on Arm. And then you have to wait fore the Electron project to pick up those changes and ship for macOS on Arm.

 

[smulji]

Microsoft are hamstrung by their monopolistic past - as soon as they compromise backwards compatibility, they're competing on a level playing field with Mac, iOS, Android and Linux... and everybody hates Microsoft. Apple have far more flexibility without the corporate albatross around their neck which they've already used to switch architecture 2-3 times as well as switch from Classic MacOS to the completely different NextStep OS X...

Not true. Ever since Satya become CEO, MS' popularity has seen a serious resurgence.

 

[firewood]

Legacy software is the problem.

Apple has already removed most of the legacy problem by removing support for Cocoa Carbon, and for 32-bit apps (as they did in the past by removing support for 24-bit 68k and for PowerPC).

Other legacy software has no such problem. I run a lot of the same server code on my Mac and my Raspberry Pi (stuff that use to run on a DEC Alpha many moons ago).