Apple to dump Intel CPUs from Macs for Arm – yup, the rumor that just won't die is back

Apple will abandon Intel processor chips for its Macs in favour of homegrown Arm-based chips, according to a securities analyst. Formerly at KGI, Ming-Chi Kuo of KeyBanc Capital Markets, has a strong record of reliable Apple predictions. Kuo issued the prediction in a research note, suggesting the switch would be made in 2020 …

[Richard 12]

Rosetta-a-like is absolutely necessary

Remind me, just how successful was WindowsRT?

The operating system isn't the major hurdle when changing CPU/GPU architecture, it's the applications.

If none of your existing applications will work, why would you buy the new Mac?

Why would a software house spend the resources making it work at all on the new platform, unless their customers are going to buy it again?

Software like Adobe Creative Cloud is fine of course, because their customers are already paying monthly, but anything you already bought will probably have to be bought again.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

Indeed. The author seems to be confusing backwards compatibility where fat binaries are fine, with forwards compatibility which will require a translation layer for old x86 binaries.

[diodesign]

Re: Rosetta-a-like is absolutely necessary

Dunno, dude. Gonna have to disagree. Applications could ship with Arm and x86 inside, with the right one running depending on the underlying hardware. It's not that hard to understand.

The difference now from the Rosetta era (2005) is the Apple Mac App Store (introduced in 2010). Just have the app store push Arm builds (or fat binaries) to Arm Macs. The iOS App Store at least deals in LLVM bytecode submitted by developers, which Apple compiles and optimizes for various different targets. Same could be done for Mac apps.

If you're downloading apps direct from developers, then you'll have to get the Arm rebuild or use Rosetta. If you're building it yourself, then just recompile.

So basically, the App Store, as a central repository, this time around is why the reliance on Rosetta won't be quite so heavy.

C.

[Anonymous Coward]

You're ignoring Windows

A fairly decent share of Macs are used to run Windows, at least occasionally. So there will absolutely need to be dynamic translation to handle that. And while new Mac applications can use fat binaries, not everyone has the luxury of having 100% of their applications still being supported by the vendor. There HAS to be emulation to handle that case. Not sure why there is argument over this - dynamic translation is a solved problem, and Apple has already done it twice. Doing it again in 2020 is not a difficult task at all.

Fortunately going from an architecture with fewer registers to one with more registers a lot easier/faster to emulate than the other way around. Even better, Microsoft is supposedly supporting ARM Windows that does its own translation from x64 to ARM64, if so (i.e. if they still are in 2020, which isn't a sure thing given Microsoft's spotty history in supporting non-x86 Windows) then it shouldn't be too much of a problem since Apple's translation only needs to worry about MacOS.

[Richard Plinston]

Re: You're ignoring Windows

> Microsoft is supposedly supporting ARM Windows that does its own translation from x64 to ARM64,

No, Microsoft is not doing 'x64' (x86-AMD64) to ARM64. It is doing x86-32 to ARM64.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

"It's not that hard to understand"

Oh ok, well you sure told me. Just humour me however and explain how old x86 only binaries will run on arm without a translation layer.

[Cl9]

Re: Rosetta-a-like is absolutely necessary

I think he's saying that the application package itself could contain both code for ARM CPU's AND Intel CPUs, with the OS loading the correct code based on the current hardware.

[Richard 12]

Re: Rosetta-a-like is absolutely necessary

Sorry, but you genuinely have no idea what you're on about, diodesign.

How does that help run any of the thousands of existing applications on this new architecture?

Unless the software developers spend the time to port the x86 application to ARM, there will never, ever be an ARM version.

Rosetta made it possible for existing PowerPC binaries to run on x86, which meant you could buy an x86 Mac and still have all your existing software, even if the developers didn't update.

Fat binaries merely made it easier for developers to ship a single "installer" for both architectures. They don't magically cause x86 or ARM versions of applications to come into existence.

Porting a commercial x86 application to ARM is not just a case of recompiling it, unless it's really quite trivial. It often costs thousands of pounds to port - who's going to pay for that?

[gnasher729]

Re: Rosetta-a-like is absolutely necessary

"Unless the software developers spend the time to port the x86 application to ARM, there will never, ever be an ARM version."

You know how much effort it will be? Open your project in Xcode, select "x86_64" and "arm64" as the processor architectures, select "Archive" and upload to the App Store.

[Anonymous Coward]

"select x86_64 and arm64"

Anyone who sells an application without properly testing it on new platform deserves all the scorn heaped upon them by unhappy users. It might be bug free if you are lucky (and if Apple's ports of the libraries your application uses are perfect and don't introduce new architecture specific issues) but only someone for whom software development is a hobby would ever suggest such a thing as a solution.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

"You know how much effort it will be? Open your project in Xcode, select "x86_64" and "arm64" as the processor architectures, select "Archive" and upload to the App Store."

And Fabio working in design is going to know about Xcode is he?

The fact that the apple fanboys don't even understand the issues of CPU architectures and binaries tells you all you need to know about these clowns.

[stephanh]

Re: Rosetta-a-like is absolutely necessary

It may be a simple recompile, but often it won't. I expect something like Photoshop to contain an amount of hand-optimized assembly, for instance.

Even after you have the port building, you need to set up arm machines for testing, in parallel with your x86 infrastructure. All doable but it costs money.

Final build needs to be tested and validated on multiple platforms. These are on-going costs; there is a reason software companies try to limit the number of platforms they support.

[BinkyTheMagicPaperclip]

Re: Rosetta-a-like is absolutely necessary

You're assuming Xcode is being used to develop all applications, rather than a cross platform build system - which would be far more sensible. Processor specific optimisations are also used in a number of cases, and it's always possible there is other code that assumes PC architecture (support chipsets on ARM are usually different).

At least ARM can run in little endian mode.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

> How does that help run any of the thousands of existing applications on this new architecture?

By thousands of existing applications surely you mean just three: Adobe Photoshop, Lightroom and Microsoft Orifice Office. That's about 90% of Apple's ecosystem, outside of Apple.

Here's the breaking news: all three of them already run on AArch64 (ARM64).

[itzman]

Re: Rosetta-a-like is absolutely necessary

Actually whilst that is true for the general stuff, apple DESKTOPS sell into niche aseas like desktop publishing where the adobe creative suite at a huge amount per desktop or old fashioned Quark express is de rigeur, and into things like video and music editing which again are de rigeur...

[Richard Plinston]

Re: Rosetta-a-like is absolutely necessary

> Here's the breaking news: all three of them already run on AArch64 (ARM64).

The RT version of Office was severely limited in some features.

[gnasher729]

Re: Rosetta-a-like is absolutely necessary

The idea is that you downloaded your old x86 binaries (which won't run on macOS 10.15 anyway) from the App Store, so you just download it again.

[K]

Re: Rosetta-a-like is absolutely necessary

Sorry, but I just don't buy it. Whilst I don't doubt Apple has the capacity to do this.

the software eco-system was significantly smaller last time they did this and their market share a fraction of what it is today. They would alienate a significant number of there customers and businesses who develop for their platform.

Saying that, given Apples recent actions on things such as right-to-repair, they might just have that level of arrogance to try it, and the temptation to finally have a walled in eco-system which they get a 30% cut from might be too much to resist.

[stephanh]

Re: Rosetta-a-like is absolutely necessary

Please note that there seems to be a common misperception that LLVM bitcode is portable across architectures - it is not. There is "ARM bitcode" and "x86 bitcode".

Applications will need to be compiled from scratch to work on ARM. Of course Apple has "fat binary" support which will make it possible to ship a single executable for x86 and ARM.

[big_D]

Re: Rosetta-a-like is absolutely necessary

@diodesign and all the LoB software that doesn't come through the Store and is currently Intel only?

If you are lucky, the developer is still around and can develop a new version that will work on ARM as well. If you are very lucky they won't charge you for the privelege.

Likewise, what about the key software you need that is no longer maintained / developed and doesn't have a "new", maintained equivalent?

I'm thinking especially about comms software here. At a previous employer, they used a VOIP system that had a special soft client with the call management facilities (who is currently talking, voicemail, chat, call logging, video conference etc.), their Mac software came out 2 years after it was promised, you could use a standad SIP client, but only to make calls, all of the other functionality was missing.

Them getting their fingers out and releasing an ARM version in a short timeframe would be close to zero, I suspect.

So you either have to wait 2 - 3 years after the switch to ARM, before upgrading your hardware or you need a software emulaiton layer a la Rosetta, WoA / WoS, so that legacy software can run, until a new, ARM based version becomes available.

The Store can't just deliver ARM versions, the author has to change their code and recompile for ARM, at the very least. It doesn't happen "automagically".

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

Fat binaries work for new applications but completely fail for existing ones. That's why there was Rosetta, and why there would need to be something like that for any new transition.

Anyone running VMs on their Mac is also going to find this interesting: I don't even want to think what the performance of an x86 VM sitting on top of something dynamically translating to ARM is going to be like: absolutely crap I expect.

I don't put it beyond Apple to just make the decision anyway.

[JDX]

Re: Rosetta-a-like is absolutely necessary

What about legacy apps? Are we suggesting nobody uses them on Mac?

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

@diodesign:

"Dunno, dude. Gonna have to disagree. Applications could ship with Arm and x86 inside, with the right one running depending on the underlying hardware. It's not that hard to understand."

When Apple shifted Macs from the 68k CPU line to PowerPC, and then from PowerPC to Intel, that is exactly the approach it used in both cases.

Also, in both cases, Apple provided CPU emulation so that you could continue to use (most) old applications on the new CPU regardless - at least for a while.

IIRC, Apple only dropped the ability to use 68k applications when it shifted Macs to Intel, largely because that side of things was handled by an emulator which ran inside the Classic environment (MacOS 9.2.2-ish magically running inside/alongside MacOS X, don't ask me about the details) and re-writing all that code so it'd work on Intel/MacOS X didn't strike Apple as worth the effort and who can blame them?

So, based on what Apple's done in the past, I'd guess that if Macs are moved from Intel to ARM CPUs, then we will again be given a combination of an Intel emulator running on ARM so old applications will continue to run, and also applications shipping with executable code for both CPUs.

[mark l 2]

Re: Rosetta-a-like is absolutely necessary

Your forgetting this is Apple land, where people queue up on the street to buy new devices because it's got poo emoji's and a slightly better camera than the previous model.

If the battery life on Arm power Macs can be significantly improved over Intel powered Macs that alone will make everyone go out and spend money to get the latest shiny shiny.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

I'm kinda, sorta with your view on this although my inner-engineer questions exactly how much of the total energy budget for a Mac goes to just the CPU of the total. Stepping up and down cores are, perhaps, enough of a significant difference but don't cite me on that! Otherwise, meh!

OT: I still miss NeXTSTEP. Met it after I returned to the university and it was good enough to give my mega-modded $10K+ Amiga 2000 a challenge.

[CrazyOldCatMan]

Re: Rosetta-a-like is absolutely necessary

how much of the total energy budget for a Mac goes to just the CPU of the total

Judging by my MacBook Pro, not much. Biggest energy use (by far) is the display, followed by wifi/bluetooth et. al.

[itzman]

Re: Rosetta-a-like is absolutely necessary

"but anything you already bought will probably have to be bought again."

which is of course the point as it was when OSX came out and then when OSX on Intel first came out.

The only up to date code that runs on a power PC Mac these days is Linux...

The Apple way is to ensure you buy and buy again and again.

[Anonymous Coward]

Re: Rosetta-a-like is absolutely necessary

That is false. When OSX on intel came out there was Rosetta, which means that PPC binaries would continue to run. That's, you know, why the thread has 'Rosetta' in its subject.

[The Pi Man]

Re: Rosetta-a-like is absolutely necessary

Apple don't want backwards compatibility - they want to sell you new stuff again and again.

[Ian Joyner]

Re: Rosetta-a-like is absolutely necessary

"Apple don't want backwards compatibility - they want to sell you new stuff again and again."

Another uninformed and wrong comment. Apple are supporting many generations of previous equipment with no forcing people to buy newer versions.

[Dave 126]

Re: Rosetta-a-like is absolutely necessary

> how much of the total energy budget for a Mac goes to just the CPU ?

How long is a piece of string? : ) A lot of the time the CPU is just ticking over but on occasion (rendering an image, compiling some code, running a simulation) the CPU might be operating up to its thermal limit. of course some of these tasks can be done by a GPU (on board, connected by Thunderbolt, over the network to a render farm, or on some rented cloud) but not all. And of course some tasks can be done on on both CPUs and GPUs and the software will fully utilise any computing resources it can find.

[big_D]

Re: Rosetta-a-like is absolutely necessary

Apple forced me to dump my Intel iMac in 2014, because they had stopped supporting it (2007 iMac 24"), it is still running today, with security updates, from Microsoft on the BootCamp side.

[Rattus]

Re: Rosetta-a-like is absolutely necessary

"If none of your existing applications will work, why would you buy the new Mac?"

Hello?

This is apple we are talking about here. If Apple tell the Fanboz (and Girlz) to buy new software AND buy new hardware they will (and be happy to pay a premium for the plasure of it!)

[Anonymous Coward]

Stop Press!

Apple has announced an event in NYC

https://forums.macrumors.com/threads/apple-launches-special-event-page-for-october-event-with-dynamic-set-of-apple-logos.2148397/

We won't have long to wait to find out what they are going to do with the Mac line.

IMHO, that will be nothing apart from retiring it in 2020. EOL!

Sad really. MacOS is still a lot better than that shite that MS calls Windows 10.

Oh well, Linux here I come.

[WootyB]

Re: Stop Press!

Apple is too money hungry to care about me as a customer, qnd Microsoft doesn't care about its tech savvy users.

If Linux were a viable option, I would switch our entire system over, even if all fee's were the same.

[Anonymous Coward]

Re: Stop Press!

For me, Linux, with a windows VM in it, became a viable option about ten years ago.

I feel for you.

[TVU]

Re: Stop Press!

"We won't have long to wait to find out what they are going to do with the Mac line"

Hopefully, Apple will turn them back into what they once were - a range of very productive laptops with a variety of useful ports instead of the sliver-thin lifestyle toys fit only for making Facebook updates and watching cat videos that they currently are.

[DJV]

@TVU

"sliver-thin lifestyle toys fit only for making Facebook updates and watching cat videos"

...and keyboards that snuff it upon encountering a microscopic speck of dust!

[Timmy B]

Re: Stop Press!

Sorry, TVU.

Had to downvote you as you have totally unrealistic ideas or it's too early in the morning for my sacasmometer to reach that high. Either way it's not yet the weekend and I'm grumpy...

[Michael Strorm]

Re: Stop Press!

@Timmy B; Oh, a *sarcasmometer*? That sounds like a *really* good idea, I'm sure.

(Sarcasmometer explodes).

[vtcodger]

Re: Stop Press!

"Hopefully, Apple will turn them back into what they once were ..."

I was thinking just before reading this that extrapolating current trends, the ARM based Apple will have no x86 compatibility. It will come entombed in a single synthetic diamond and will not be openable, much less repairable. It will have no ports whatsoever and will be charged by a proprietary $400 electron beam charger. There will be no keyboard at all although a magnificent, lifelike, but totally inoperative image of a keyboard can be displayed by the proper combination of finger motions or by appealing to Siri's better nature.

It'll be a huge seller and will cost significantly less than a low end Tesla Model 3.

[Anonymous Coward]

Re: Stop Press!

Having used all three I can confirm that Linux is better supported and more rational than OS/X..

OS/X is OK if you don't mind doing things exactly the way Apple want you to. And paying through the nose.

But woe betide you if you want to do it differently.

[Ian Michael Gumby]

@AC...Re: Stop Press!

Apple OSX is derived from Mach (micro kernel) that run NeXT-Step.

In terms of developer platform. I'd take a mac over a linux box but would still run the code on a cluster of linux boxes. There are a lot of things that the Mac gets right.

With respect to ARM vs Intel... The Arm chips could make sense if you're running something like a macbook air where all you need is enough power to run a web browser and some other stuff.

Where Intel comes in is when you want to run a lot of things at the same time or need a bit of CPU/GPU horse power and memory that you can't get via an ARM chip.

So for Macbook Pros, still need the intel chip.

That said... Intel needs the competition to continue to improve their chips.

[Daniel von Asmuth]

Re: @AC...Stop Press!

"There are a lot of things that the Mac gets right."

Let's guess: CISC processors, lots of RAM, SCSI discs, Apple Desktop Bus, Ethertalk port, VGA display, Postscript Laserjet, NuBus. Apple got a few things right (including firing Jobs), but SUN did them better.

[Claptrap314]

Re: Stop Press!

Apple: Don't Do Different.

[Daniel von Asmuth]

Re: Stop Press!

I'm half expecting Apple to drop the MacIntosh family of microcomputers. Switching from Intel to ARM may be a good idea today,, but by 2020, Intel's new 7 nm EUV process will make Chipzilla great again.

[Charles 9]

Have to wonder, though, what happens when you have to run an unported x86 MacOS app that happens to be performance-intensive, like say a high-end MacOS game (which DO exist--since Steam has a MacOS version).

[Dave 126]

Just make sure you have the right hardware, I guess. Lots of games are hardly worth playing without a discreet GPU, so why not games that won't run unless the machine has an x86 CPU?

In any case, different games place different demands on CPUs and GPUs, with many not requiring that much CPU power - so possibly more tolerant of x86 emulation on ARM. Possibly of note is that many MacOS Steam games are built with Metal, the same graphics API used in iOS on Apple's own silicon, includibg the usual big names:

https://en.m.wikipedia.org/wiki/Metal_(API)

[WootyB]

Worth the RISC

Since macOS is just a cousin of unix, porting current applications to ARM will be a breeze. Bootcamp may or may not be removed however, as it relies on ARM ports of Windows for the Apple-branded CPU.

Recently, IBM open-sourced their new POWER9 CPU's (think PowerPC), and they already had 90% of all Debian Linux packages ported ofer in a few weeks/months.

I believe in 10 years, architecture (ie. x86, PowerPC, ARM) will be more of a preffered brand, except in very specialized applications.

[alain williams]

Re: Worth the RISC

I believe in 10 years, architecture (ie. x86, PowerPC, ARM) will be more of a preffered brand

Having spent the last 35 years working with Unix/Linux - that has always been my view. A new architecture is just a 'make' away for decently written programs.

20+ years ago the common use of different architectures was much greater than today.

[theblackhand]

Re: Worth the RISC

"I believe in 10 years, architecture (ie. x86, PowerPC, ARM) will be more of a preffered brand, except in very specialized applications."

I'd be very surprised if PowerPC is a viable architecture in 10 years time. SPARC has been open sourced for a while and it is still in decline, with MIPS head start in the open source CPU designs appearing to give it the advantage in the also ran stakes.

AMD has got the low power/high volume CPU market covered. With Intels miss-steps, it looks like a combination of Intel and AMD will keep the mid-to-high performance/high volume CPU market covered and Power is the last real survivor in the very high performance bracket with Itanium having no uarch changes since 2012 and SPARC likely to be one final iteration away from being purely niche.

Can ARM challenge Intel/AMD in the mainstream server market? Possibly, but I'm unsure how it will match performance without increasing cache sizes/pipeline length and moving more functions onto the chip which hurts performance and cost.

Intels current failings at 10nm may completely alter that assumption as it would keep ARM/AMD on equivalent process tech. Intel now have a massive hole in their production line assuming they are sticking with 10nm - a surprise change to 7nm in Fab 28 would address a lot of this but I'm unsure if it is even possible and if not Intel will end up with two failed 10nm Fabs that will need re-built as x nm...

[Anonymous Coward]

Re: Worth the RISC

@WootyB

"Since macOS is just a cousin of unix, porting current applications to ARM will be a breeze."

You'd've thought so, but I've recently contacted the developers of two Mac applications I'm rather fond of (both of them first came out on 68k Macs). They're currently only available in a 32 bit version.

According to the developers, the only way they can convert them to 64 bit is to completely re-write them in Apple's Swift language which they are having to learn. And that's going from Intel to Intel...

On the other hand, I'd guess that if you've got a fully 64 bit application written using Swift and XCode, then Apple will arrange things so that converting an Intel application to ARM will be low-effort, assuming that Macs are switched to ARM CPUs.

[gnasher729]

Re: Worth the RISC

"You'd've thought so, but I've recently contacted the developers of two Mac applications I'm rather fond of (both of them first came out on 68k Macs). They're currently only available in a 32 bit version.

According to the developers, the only way they can convert them to 64 bit is to completely re-write them in Apple's Swift language which they are having to learn. And that's going from Intel to Intel...

On the other hand, I'd guess that if you've got a fully 64 bit application written using Swift and XCode, then Apple will arrange things so that converting an Intel application to ARM will be low-effort, assuming that Macs are switched to ARM CPUs."

Going from 32 bit to 64 bit can produce all kinds of problems, especially if the code was written at a time where nobody that 64 bit might ever exist. As you said, built for 68k Macs initially. That would have been before 1990.

What these developers probably have is _ancient_ code ported to MacOS X using the Carbon framework, which was never ported to 64 bits. That would have happened around 2002 or so when MacOS X was introduced, and stopped being supported about six years ago. There is no requirement to switch to Swift, Objective-C would do just fine. But 32 bit won't run on MacOS 10.15 anyway (that's the next version after the just released one).

Intel 64-bit vs Arm 64-bit using modern frameworks is no problem. The capabilities of both processors are the same, so any code not using Intel assembler code will compile and run identically.

[Daniel von Asmuth]

Re: Worth the RISC

Porcessor architecture stopped mattering around 1995. The future is in application-specific chip designs and coprocessors.

[Michael Jarve]

Same trick, new pony

As the article contends, this is not Apple’s first rodeo switching architectures. But, as capable as ARM processors are now, I think there is still arguments to be made as to X86’s superiority, especially if Apple wishes to keep its foothold in the creative/pro markets. Apple’s intimate knowledge of ARM not withstanding (they were early champions of ARM), many third-party developers are x86 houses for power apps, with a mere toehold in ARM. The Unix underpinnings of Mac OS are quite portable (compared to classic Mac OS), but if they are not abandoning the workstation market (as they seemingly have) they will need something more than “good enough” ARM. Unless they have a quite powerful ARM chip in mind, without cooperation from nVidia and AMD, any such adventure is doomed to fail. When Apple went from PowerPC to Intel, they were becoming [i]more[/i] compatible with market/ industry; this would be going back the other way. As innovative as PowerPC was at the time, ARM is still the economic/ecological solution to CPU’s and will remain so untill they have a real competitive desktop/workstation offering that competes with x86 on a performance and economy basis. In addition, it would screw anyone invested in a MacInSoft workstation and do away with any developer cred, absent some really awesome emulation mojo. I, for one, don’t look forward to a “PC” compatibility card of ancient times ( if, ever again an expandable Mac were to exist).

[Wade Burchette]

Re: Same trick, new pony

AMD has already shown that they will help with with a semi-custom chip. Look at the consoles, the PS4, XBox, and the new Chinese one are all semi-custom AMD parts and Sony, at least, is working with AMD for the PS5. AMD is even helping Intel with a semi-custom CPU by putting an AMD GPU on an Intel CPU. What Apple may end up doing is going ARM and licensing x86 from AMD in a semi-custom package.

Regardless, this will be one of those rumors I will believe when I see. I personally think more than likely there will be a Macbook/iPad hybrid, like a Chromebook.

[Dave 126]

Re: Same trick, new pony

The Reg reporting of this story has perhaps more emphasis on Apple ditching Intel than other reports which don't rule out Intel being retained for some models.

Professionals with legacy software and/or greater power demands will pay for 'Pro' models with Intel and maybe AMD, consumers will be content with ARM and Apple GPU.

Feature complete Photoshop is coming to iOS in 2019, so there'll be a couple of years to demonstrate ARM productivity applications can work well.

[Naselus]

Re: Same trick, new pony

"Feature complete Photoshop is coming to iOS in 2019, so there'll be a couple of years to demonstrate ARM productivity applications can work well."

Relying on Adobe to produce evidence of software 'working well' is a recipe for disaster tbh...

[Martin an gof]

Re: Same trick, new pony

[Apple] were early champions of ARM

You could argue that they were the first, not merely an early champion.

They were one of the three founding partners of Advanced Risc Machines alongside Acorn and VLSI, who had worked together to create the Acorn Risc Machine - ARMs 1, 2 and 3. Apple came on board (IIRC) as the ARM6 was being developed (there was no 4 or 5) though - again from memory - they were a minor partner, something like 10%.

Disclosure: I had a RiscPC with an ARM 610 :-)

M.

[Daniel von Asmuth]

Re: Same trick, new pony

The IBM 801 is supposed to have been the first RISC processor, back in 1980.

https://en.wikipedia.org/wiki/IBM_801

[alain williams]

What else can a move to ARM bring ?

Most of the comments here look at the problems of getting existing programs to work on a new architecture, be that via recompile or emulation or something.

An Apple designed CPU could bring a whole range of new instructions, maybe doing some of the things that are today offloaded to GPUs. This might help with more AI (whatever that means) and other needs where new silicon could give great advantage. Apple will not share its designs and will probably patent what its new silicon does to stop others following suite.

[Anonymous Coward]

Re: What else can a move to ARM bring ?

Possibly. But in the Arm world a so-called AI cpu is already a thing with vendors other than Apple. As it is basically a very fast but low precision parallel FPU unit, the new instructions apply to that, not the CPU.

Intel is not in that market at all, and this could be the tipping point. Since Apple have already done this with the A12, the basics are there. A coprocessor for something from Intel will have all the negatives of coprocessors - like power consumption. So it makes sense to bring all the silicon on one die, which may have been the plan all along.

Meanwhile Huawei have been doing exactly the same thing, though where they are going with it is hard to guess. Other than that it's fairly clear that Windows is unlikely to be the goal given that commercial, if not shooting, war between China and the US looks like the future.

[Richard 12]

Re: What else can a move to ARM bring ?

x86/amd64 already have them in the form of AVX et al, and ARM has NEON. While vaguely similar they're not the same though.

The reason for offloading to the GPU is because a GPU is a massively-parallel array of floating-point processors, and most of this "AI" stuff is simply massively-parallel low-precision computation...

It often makes it slower to bring it into the CPU, because shared address space means you have to keep the cache coherent. The "copy input data to coprocessor memory, run, copy result data back" approach is much faster as there generally isn't very much input or output data compared with the number of intermediate values.

[Anonymous Coward]

Re: What else can a move to ARM bring ?

If Apple did ass extra instructions to an ARM architecture it would fundamentally invalidate every ARM compatible compiler.

That is not really what you need for rapid portage of existing apps.

[stephanh]

Re: What else can a move to ARM bring ?

Compilers already deal with tons of variants on ARM. Note that the Raspberry Pi has its own peculiar flavor of instruction set which means RPI binaries generally don't work on other ARMs.

ARM is trying to clean up this mess a bit in the 64 bit instruction set, but if Apple wants to add custom instructions then there is little ARM can do. It is not if you can run macos binaries on other oses anyway so who will care?

[Anonymous Coward]

Re: What else can a move to ARM bring ?

Everyone's ARMs would have to keep up with everyone else's ARMs.

It'd be an ARMs race.

[timrowledge]

Re: What else can a move to ARM bring ?

Where on earth do you get that idea from? Are you perhaps confused by the way the the very earliest Raspbian releases used a slightly uncommon FP configuration.

It’s all normal now. That’s why quite a large range of OSs run happily on any recent Pi.

[gnasher729]

Re: What else can a move to ARM bring ?

"If Apple did ass extra instructions to an ARM architecture it would fundamentally invalidate every ARM compatible compiler."

Existing compilers would just not use the new instructions. And Apple was always the driving force behind Clang, so you can be sure that any new additions would be supported by Clang - which is what people use to compile MacOS X code. Not that I expect any additions.

[Anonymous Coward]

I hope this happens. I hope other manufacturers will also look into using alternative architectures like RISC-V (tho I doubt that will come to fruition)

I've been wanting an up-to-date Management Engine / PSP free laptop for a while now, and I really don't want to use the decade old Lenovo laptops that can still use Libreboot.

[a_yank_lurker]

Basis??

Other than an 'analyst' who has decent track record what is the basis for the switch from Intel to ARM? I can see advantages and disadvantages for both staying and ditching. And other than Intel's bungling (which helps AMD and ARM) is there an excellent reason to switch, I am doubtful. The previous switches (from 68000 to PowerPC then to Intel) where made because of serious problems with the chips, availability of competitive designs, and designs. It sounds as if the switch might be from Intel to AMD rather than ARM.

[Anonymous Coward]

Re: Basis??

I think something else.

I think that Apple has three or maybe four comoputung markets.

1/. Smartphones

2/. Fondleslabs

3/. Laptops

4/. Workstation desktops.

By and large the first three are consumer grade markets, and punters will accept (it seems) trashing last years shiny thing for this years shiny thing and punters don't BUY software. They use what is there, or they effectively 'rent' from the apps store.

I suspect therefore that ARM which is already on phones, and slabs, merely has to give the slab architecture a keyboard and mouse pad and the laptop is ARM there anyway.

The real $64,000 questions is whether the workstation market is in any way a market that apple wants to stay in. Because that is where people make huge investments in third party software. All of which would need to be ported.

And there is a question of whether its worth it or whether in fact MOST of that software, which runs on windows as well, shouldn't actually dictate that the customer move there instead?

[Spazturtle]

Re: Basis??

"Other than an 'analyst' who has decent track record what is the basis for the switch from Intel to ARM?"

The basis is that leaks suggest that Apple wants to ditch Intel and everyone is assuming that means they will go to ARM, for all we know they might be planning on switching to AMD CPUs.

[Dave 126]

Re: Basis??

How does AMDs performance per Watt consistent to Intel's these days? Or, more to the point, how will it compare in a couple of years time? Or, back to the topic:

https://www.anandtech.com/show/13392/the-iphone-xs-xs-max-review-unveiling-the-silicon-secrets/4

[dnicholas]

Re: Basis??

Very competitive, besting a Intel in perf/watt at higher core counts. AMD will improve too, right now the immediate future of Intel chips is (re)spin spin spin

[Anonymous Coward]

Re: Basis??

"he previous switches (from 68000 to PowerPC then to Intel) where made because of serious problems with the chips,"

On a historical note, the big reason Apple switched from PPC to Intel is that IBM wanted Apple to pay the development costs for the next generation of PPC CPUs and Intel CPUs were due to catch up with PPC performance - in fact, had already outstripped PPC performance in many areas when the first Intel Macs came out. I seem to recall it took five years from the start of the transition before Intel CPUs were up to the job of replacing the G5s in the high end tower Macs (whatever the blasted things were called at the time, I forget even though I had one). I think it was vector processing and streaming performance which the G5s were particularly good at.

[Anonymous Coward]

Re: Basis??

I was at IBM when this happened, and on the GPUL / G5 development team. IBM got so angry with Motorolla at the end of the joint deal that they blew $2B just to force them out of the PowerPC business. They sold Apple a roadmap that was a ridiculous joke, and Jobs bought into it. Supposedly, the GPUL, which was a GP chip with a smaller cache & with vector instructions bolted on, was going to service both their high-end servers and their low-end laptops. Those high end servers came out with 9 fans, by the way.

A year later, IBM had paid penalties on the contract every quarter for missed deliverables. Jobs was furious. Showed a couple of IBM VPs a Mac running on Intel & told them that it was their job to make it not happen. When Apple canceled a few months later, the internal announcement at IBM hilariously made no mention of winning the business back.

But Moto was out of the PPC market, so, mission accomplished.

[Anonymous Coward]

Please remember...

Whatever else you may think of Jobs, he was interested in tech, and had a general idea of where he was going.

Cook is a bookkeeper. He doesn't care about tech, innovation, or poor people. He cares only for 'value to investors'.

If they are considering switching to a new architecture it is only because he thinks more money can be made that way, either by larger margins or marketing.

There is nothing else behind this.

[Anonymous Coward]

Re: Please remember...

"Cook is a bookkeeper. He doesn't care about tech, innovation, or poor people. He cares only for 'value to investors'"

I don't think you understand what a bookkeeper does. They ensure your financials are honest.

Cook is a supply chain expert. His job is to maximise the profit from whatever the company does in the way of R&D, sales and marketing.

I don't think he is the problem. I suspect it lies in the design and engineering departments, and not wanting to be involved in a failed project. And that will come from the shareholders.

Apple is now so big and so many people depend on it that it has become too big to be allowed to fail. Remember Jobs ran a much smaller company.

[bussdriver]

groundwork is there already

Apple already is moving to partially compiled binaries for their app store; which allows Apple to finish compiling your binaries to other hardware. When I read about it I thought of it as a better way than what Java did... or something between the two would be ideal.

Apple already has iOS on ARM and now they are getting it perfected for porting to Intel so existing MacOS can run iOS apps without a ton of trouble. I've been wondering why they are killing 32bit apps when the OS was so well designed... they must have a reason to cut off support which should be really easy to provide. Could it be because they plan on hardware accelerated translation for ARM? Or they don't want to emulate the mess that is x86 too?

[45RPM]

Business as Usual

To be honest, I’m surprised that they haven’t already. Intel has had a good innings with Apple but, up to now, Apple seems to switch every ten years or so (Intel is now at 13 I think).

The Mac switched smoothly from 68k to PowerPC, and from PowerPC to Intel without a hitch (and that’s ignoring all of the other architectures that it passed through in prototype form, or only incidentally (Sparc, PA-RISC and so forth)

There’s always a small speed bump, but no one seems to notice or care - and ARM is definitely big enough and fast enough (at least, in Apple’s versions) to take the strain. Just imagine what Apple could do with ARM of the space and power constraints of having to run in a phone or tablet were removed…

[Charlie Clark]

Re: Business as Usual

To be honest, I’m surprised that they haven’t already.

So am I but there are good reasons for the wait. First of all, single process performance on Intel is still better than ARM. This is particularly important for things like browsers which parse the DOM in a single pass.. Secondly, Apple is only just getting the necessary expertise to manage chip development to same degree as its mainly supplier (Intel) does. Thirdly, the compilation process has got to work for all applications at the flick of a switch. This means replacing all x86 specific optimisation calls with their own. Fourthly, some kind of emulation, particularly with hardware acceleration will still be required for legacy applications and things like x86 VMs. It would be stupid to release an OS with shitty VM performance. Going with ARM would, of course, allow a dedicated unit to do the thunking.

But the main reason why Apple hasn't yet moved is: why bother? Switching requires a lot of resources and includes compatibility risks. It has a good arrangement with Intel for early access to new chips and obviously gets a good deal on the volumes.

Moving to ARM should eventually simplify the MacOS/IOS world with the aim presumably of a single OS at some point. But anyone with experience will tell you that, not only do such plans often hit roadbumps, other changes often mean the plan no longer makes sense.

[Gordon 10]

Re: Business as Usual

My first thought was is it Apple-Intel contract renewal time again?

I struggle to believe the savings/scale from owning the whole chip supply chain on Arm exceeds the scale that Intel get on x86 chips from a whole industry. This is an attempt to knock a little bit more off Intels margins. Even though many of Apples intel chips are special SKU's just for them, they are mostly just limited to enabling high end features lower down the range (ie hyperthreading).

The only way Apple *might* get enough synergies is if the iPad and Mac use the same sku's - and I doubt that will happen for practical reasons - ie a CPU for a full fat PC is still different enough from a Tablet PC CPU running iOS to make it impractical.

[Charlie Clark]

Re: Business as Usual

I struggle to believe the savings/scale from owning the whole chip supply chain on Arm exceeds the scale that Intel get on x86 chips from a whole industry.

You provide your own answer further down: the scale obviously already works for the I-Things. ARM chips are already all custom designs but the biggest savings come from process technology and using less silicon.

[stephanh]

Re: Business as Usual

I imagine the ARM chips would only be used for MacBook Air-like models and the high-end would continue to use x86.

[naive]

It would be another world changing innovation of Apple

After Apple changed the world with its smart phones in 2007, this would be another great step. It is weird Intel has a monopoly on CPU's, although the true cause of this is the fact that many mainstream software vendors sell their software based on "per cpu license", thus killing off all initiatives for less powerful but innovative designs, since they would cost more to run applications.

A viable non-Intel CPU based computer would be a great example, showing the world there is more in CPU land than just Intel, we all can profit from competition and innovation.

[HolySchmoley]

Re: It would be another world changing innovation of Apple

"After Apple changed the world with its smart phones in 2007"

The only world that changed was the world of the aimless slurping their Starbucks. 99.99999% of 'the world' has little to do with staring and dabbing at a small slab of glass for hours every day.

[gnasher729]

Re: It would be another world changing innovation of Apple

"The only world that changed was the world of the aimless slurping their Starbucks. 99.99999% of 'the world' has little to do with staring and dabbing at a small slab of glass for hours every day."

Wait a second... Are you telling us there are only 700 iPhone users in the world?

[Anonymous Coward]

Apples customers

I'm sure Apple knows its customers better than I do, but it will depend on how the market share breaks down between consumer and professional.

I can't see professional users who, in my experience, replace equipment as soon as a new model is available, would be prepared to use a product that was significantly slower. Apple have been there before with PowerPC and it cost them a lot of previously happy customers. And these customers know about the performance of emulation software....

If, on the other hand, Apple plan to release a Chromebook-type device that benefits from the existing iOS infrastructure while providing more CPU/GPU/screen size, I can see that possibly working. Particularly if it becomes a Macbook Air and is replaced every 12 months with a newer, faster CPU and is "fashionable"....

[Fading]

Lots of reasons for Apple to shift.....

and the enforced obsolescence of all the current Intel kit is just icing on the cake.....

[Mage]

They have switched before

68000 and Power PC.

Also I wrote about this when they ditched Apple Server HW and again when they brought out an iPad with keyboard and stylus. Also their desire to have Mac users locked into a walled garden for everything is obvious in newer OS X versions. Also blacklisting older (32bit?) SW on Mac.

Apple will go cautiously. This won't be soon, but inevitable. Apple doesn't have to be CPU native compatible like MS desktop Windows (legacy desktop apps is why people put up with Win 10, though MS has stupidly greatly reduced compatibility).

The waste bin model (round trash can) and other newer Pro models show Apple is more interested in volume sales and cutting cost than servicing the historic professional Video and Publishing markets.

I don't see iOS (at GUI anyway) replacing OSX as a laptop / ultrabook needs a different GUI and more flexible, featured OS than a phone/tablet. See Chrome OS vs Android (sensible) and the Win 8 debacle (Zune / Tile GUI OK phone, stupid on laptop).

[ChrisCabbage]

Wasn't there some talk about emulating Intel/AMD architecture on ARM...?

I mainly use my Mac for audio work - music production etc. I'm not sure it's easy to re-work audio processing code, to work on a new architecture all that easily. Especially given the real-time constraints it is and will be subject to.

This might force me off the Apple platforms and back onto Wintel.

[Charlie Clark]

I think some kind of hardware emulation is a given but it will probably be less noticeable than Rosetta was (hardware is a lot faster and the lessons from Transmeta have been learned).

The other point: code that requires specific hardware acceleration will be a deal-breaker. To be fair Apple already has extensive APIs for this so it should be able to handle most of this at compile time. But I wouldn't make any plans about your future OS yet. Apple generally provides OS support for Macs for at least five years. This applies, of course, to all new devices sold from now until the switch which should give both them and developers time to work out any kinks.

[schofiel]

MC88k, anyone?

Don't forget one of the original pre-PowerPC architectures NeXT were targeting was the Motorola MC88000 RISC, based on the original Mac dev team's experiences with 68k.

[DrBobK]

68K -> PPC

I remember thinking the original 680x0 to PowerPC transition was going to be a disaster but it was actually much smoother than I'd expected. I was programming research stuff at the time as well as using a Mac for day to day office things. I switched to Windows a few years later (my university stopped supporting Macs and hence they became a bit tricky to buy). I'm currently thinking of switching back twenty something years on....

[Waseem Alkurdi]

Re: 68K -> PPC

You're going to be surprised at how much Macs changed.

[DuncanLarge]

Will Apple finally change?

If they switch to arm, will they finally start making hardware that wont die if you sneeze on the keyboard or look at it incorrectly?

https://www.youtube.com/user/rossmanngroup

[DerekCurrie]

Re: Will Apple finally change?

Watch Mr. Rossmann rant enough times and your realize he has tunnel vision. I think he's a great technician and knows how to express himself well. But his daily experience of people bringing him broken gear by no means equals the experience of any but a small minority of Apple gear users. Even he admits he's grinding his ax for his own benefit versus Apple's insistence upon certified repair shops, parts and processes. I'm glad when Rossman points out the all-too-many real Apple Blunders of late. But he's often obsessive in his Apple criticisms to the point of maniacal, as opposed to stepping away from his limited perspective and taking in a more insightful and full perspective. I personally got tired of the rants, despite his skills, and unsubscribed. I prefer clearer heads.

[dnicholas]

My prediction

Apple surprises everyone and releases their own CPU based loosely on OpenRISC, raises two fingers to both Intel and ARM.

Predictions are easy

[Charles 9]

Re: My prediction

Why would they flip the bird at ARM? They already paid for and hold the legal license to build their own ARM-based silicon and have been doing so for years. Why throw out the baby with the bathwater?

[Anonymous Coward]

Re: My prediction

“Why throw out the baby with the bathwater?”

Because if you throw the baby high enough, the baby may see the sarcasm?

OpenRISC however is a dead end. UPU is newer, more efficient, backed by a rising superpower in a country renowned for its technical manufacturing abilities....

[Charles 9]

Re: My prediction

UPU seems to me more like patent-dodging. You have to think there's a reason CPUs and GPUs are kept on separate units, much as how computers and phones are kept as separate UI designs.

[dnicholas]

Re: My prediction

The reason? Cost.

If you sell iWidgits by the million, it all adds up. Multiply by X number of years you plan to be in business... It's a wild prediction of course.

The ideal for a hardware/software biz like Apple would be to control the hardware end to end and supply hardware agnostic APIs to developers for your proprietary OS. But that's just wild speculation

[Balding Greybeard]

Need To Think Differently

Imagine an OSX hypervisor running on an ARM chip on a motherboard that has an x86 low-cost Atom chip.

Problems solved... IOS functionality on the ARM, MS PowerPoint, Visio, etc. on the Atom.

[Charles 9]

Re: Need To Think Differently

Not quite. You forget there are plenty of performance-intensive x86 MacOS applications out there. Plus the whole VM problem (Atom is NOT VM-capable, last I checked).

I think VMs will be the real problem. They weren't a thing in the 68K->PPC or the PPC->x86 transitions, but now this can be a major obstacle to transitioning away from x86.

[Herby]

In further news....

Apple just announced that after the transition to ARM chips in 2020, they are thinking about switching to 68080 based chips in 2025. The new chips will be 64 bit compatible and run faster with many cores.

Yes, I am dreaming.

[Claptrap314]

Techincal & Marketting difficulties

You can probably spin a custom ARM chip suitable in a year. You cannot build a component validation team in a year. I don't know Apple's internal structure, but unless they already have competent verification and validation teams in place, this will fail badly.

As for a translation layer, remember that x86 chips already are translating into a risc-type internal instruction set. There might be less of a performance loss than we expect.

[razorfishsl]

Er no.......

This will basically BREAK all VMware products, you know all the WINE & x86 VM programs currently running on OSX.

Talk about shutting down , PCB,silicon & engineering tools & for what?

[Waseem Alkurdi]

Think Different. Isn't that Apple's catchline?

[dnicholas]

It's now iSlave

[DerekCurrie]

This Old Techno-Ignorant Rumor Now Bores Me

*yawn* I'm bored out of my mind attempting to explain to those with willful tech-ignorance why this rumor is not gonna happen. I considered this issue dead two years ago. And yet, I end up having to explain it over and over anyway.

So, just watch, all you willfully tech-ignorant gossip mongers. No ARM Macs. ARM chips IN Macs we already have and why not! But any transition from Intel CISC based CPUs back over to RISC CPUs at this point is impossible without throwing the baby out with the bath water. If you don't understand why, it's because you have some Computer 101 facts to learn. /rant.

Hey Ming-Chi Kuo! Is this your rendition of August Syndrome? You went on vacation and phoned in this BS as filler? Don't do that.

[Waseem Alkurdi]

Re: This Old Techno-Ignorant Rumor Now Bores Me

you don't understand why, it's because you have some Computer 101 facts to learn.

So you don't have anything to back up your dismissal of the rumor, and you're accusing the other side of not knowing enough because they can't back up your own argument.

Wow.

We're all 1D10Ts here. Care to explain to us, professor?