Apple Plans to Ditch Intel and Use Custom Mac Chips Starting in 2020

[Freyqq]

I've read through a lot of this thread and gave it some thought. This move is not to enhance OSX. This is a move to enhance iOS. Expect over the next two years iOS apps getting more advanced and OSX apps getting dropped or not updated. I have a feeling that all current OSX apps will not function without major work, but iOS apps will work pretty much seamlessly.

The upside is probably Apple will release a Surface Book competitor.

 

[slooksterPSV]

I've read through a lot of this thread and gave it some thought. This move is not to enhance OSX. This is a move to enhance iOS. Expect over the next two years iOS apps getting more advanced and OSX apps getting dropped or not updated. I have a feeling that all current OSX apps will not function without major work, but iOS apps will work pretty much seamlessly.

The upside is probably Apple will release a Surface Book competitor.

The best part of this entire thing is that they are working towards a convergence platform as well. My iPhone and iPad to the exact same things. I can make calls using my iPad, my Mac, etc. So why not make a single device to do everything? Powerful enough for a developer, small enough for a consumer. =D

 

[ActionableMango]

Again I hope everyone remembers the Emulation tech that we used back in the PowerPC to Intel transition days, it was called "Rosetta Stone" and it was good. Good enough to play OpenGL games.

Yes, I remember "Rosetta" and it was quite good.

I also remember when Apple took it away and my software stopped working, some of which had no Intel replacements.

P.S. "Rosetta Stone" is language learning software.

 

[MrX8503]

Nothing personal, but do you understand that you are just general software user case and your work can be done in a secure web app? Actually iOS is generally optimized frontend with closed hardware and software. There is no Magic in this decision, there is only pure Greed. Every business ecosystem has a bone structure. The Bone of Apple is creative professional, the skin and make up are regular users. This happens when marketing and sells departments take over design and development. Change is not innovation. Ask what drives the change. And if my words don't ring a bell listen to this man saying it clearly:

Steve Jobs also said that companies that are serious about software should build their own hardware. This is the philosophy that got Apple to build their own SoCs for the iPhone.

Those that see this move away from Intel as a bad move can't see the bigger picture. Apple wants to control the entire UX and that means building the CPU themselves.

It'll be painful at first, but we'll look back and see that it was the right move for Apple all along.

 

[techwhiz]

1) You don't think Apple can throw a bunch of PCI-E lanes, more memory channels etc. in? They can build something as good/better than "Core i9".
2) I never mentioned emulated x86, only native Linux ARMV8 virtual machines. I don't think x86 Windows virtualisation would be a realistic expectation.
3) I don't understand this point?
4) They already do way more than "modify the number of pipeline stages". Apple's cores are custom from the ground up they, just implement ARMV8 ISA. They can stick to the ARM spec and still do things like encryption acceleration (like AES-NI in x86), wide SIMD/vector instructions (like AVX) etc. to be at parity with x86. When I say custom accelerators, I mean beyond the CPU cores, custom logic outside the CPU core.
5) They can do it. Apple's core is already in the same league as Intel's, while it's not trivial to build up from that to a desktop/server class part it's not beyond Apple's engineering capability. By far the hardest part is designing the core and they've already nailed that.
6) iOS is macOS. Same kernel, same graphics, same basis for the APIs etc.
7) Broadcom sold their IP to Cavium who've used it to build ThunderX2 (I should have left Broadcom off the list).
Samsung never joined the server race, but they are actively developing high-performance ARM cores.

My point regarding third-party options wasn't that there are off the shelf designs available, but there are companies Apple can work with (Qualcomm, Cavium, AMD, ARM etc.) should they falter. It would be a case of integrating the third-party IP into Apple's design, rather than farming out the whole SoC.

It's worth remembering that the core in your Intel-powered MacBook is a "server" core. AMDs Zen is a "server" core. The logic around the core changes depending on the application (e.g. the number of PCI-E lands, memory channels, cache) but the core remains the same e.g. between Core m and Xeon. (well, the core differs with Skylake EP but that's unusual!)
[doublepost=1522791739][/doublepost]
Why does ARM only have to mean lower power consumption? It's more than that, it can mean better performance (as we've seen with Apple's SoCs already)

1. I think you miss the point and no making an ARM a desktop processor is a non-trivial exercise. The current A11 doesn't support visualization and threads. The current A11 is not a desktop processor. Besides the DDR channels you need memory scheduling. Adding more cores, you need additional snoop filters in the cache.

Also they need at least a laptop, desktop and a server class CPU.

3. The point was that iOS and MacOS aren't that different anyway.
4. Even though it's custom, they can't do any thing to modify the instruction set. So any customization must be in execution of existing instructions or it's not an ARM. Yes, they can add custom acceleration outside the core. Crypto, etc.
5. once again synthetic benchmarks don't measure system performance and an A11 is not comparable to an i9.
7. Samsung did have a server group that they failed to sell and disbanded. The IP may have been acquired by Cavium, but not the people.

I still disagree that the Apple A11 is a desktop class core, but we will have to "agree to disagree".
I never said ARM meant lower power. In actuality the metrics and characteristics change dramatically for computing on an iPad and a general purpose laptop. iPads are primarily for consumption and lack the performance for multi-stream video encoding for instance.

Let's just summarize this:
Can Apple assemble a team to make processors to compete with Intel? Absolutely.
They might have to pay for some of the patents they will infringe on, but they have cash.
Is it the A11 family? Absolutely not.
Is it a two year project to develop a family of laptop/desktop processors? Not a chance.
The A-series processor is on a one year cycle and its an evolution of previous technology.
How long to replace Intel? More than two years if they ahve to add virtualization and threads.

It all boils down to ROI.
Can Apple make an argument to shareholder for the return on investment?
I don't know, but if they do this, they will need to.

BTW see Sept. 2017
https://www.extremetech.com/mobile/256730-apple-quietly-designing-building-silicon-empire
[doublepost=1522796129][/doublepost]

I've read through a lot of this thread and gave it some thought. This move is not to enhance OSX. This is a move to enhance iOS. Expect over the next two years iOS apps getting more advanced and OSX apps getting dropped or not updated. I have a feeling that all current OSX apps will not function without major work, but iOS apps will work pretty much seamlessly.

The upside is probably Apple will release a Surface Book competitor.

http://fortune.com/2015/10/13/apple-merge-ios-os-x/

Tim Cook - “We don’t believe in having one operating system for PC and mobile,” Cook said, according to those in attendance. “We think it subtracts from both, and you don’t get the best experience from either. We’re very much focused on two.”

So unless Tim changed his mind......

 

[Freyqq]

1. I think you miss the point and no making an ARM a desktop processor is a non-trivial exercise. The current A11 doesn't support visualization and threads. The current A11 is not a desktop processor. Besides the DDR channels you need memory scheduling. Adding more cores, you need additional snoop filters in the cache.

Also they need at least a laptop, desktop and a server class CPU.

3. The point was that iOS and MacOS aren't that different anyway.
4. Even though it's custom, they can't do any thing to modify the instruction set. So any customization must be in execution of existing instructions or it's not an ARM. Yes, they can add custom acceleration outside the core. Crypto, etc.
5. once again synthetic benchmarks don't measure system performance and an A11 is not comparable to an i9.
7. Samsung did have a server group that they failed to sell and disbanded. The IP may have been acquired by Cavium, but not the people.

I still disagree that the Apple A11 is a desktop class core, but we will have to "agree to disagree".
I never said ARM meant lower power. In actuality the metrics and characteristics change dramatically for computing on an iPad and a general purpose laptop. iPads are primarily for consumption and lack the performance for multi-stream video encoding for instance.

Let's just summarize this:
Can Apple assemble a team to make processors to compete with Intel? Absolutely.
They might have to pay for some of the patents they will infringe on, but they have cash.
Is it the A11 family? Absolutely not.
Is it a two year project to develop a family of laptop/desktop processors? Not a chance.
The A-series processor is on a one year cycle and its an evolution of previous technology.
How long to replace Intel? More than two years if they ahve to add virtualization and threads.

It all boils down to ROI.
Can Apple make an argument to shareholder for the return on investment?
I don't know, but if they do this, they will need to.

BTW see Sept. 2017
https://www.extremetech.com/mobile/256730-apple-quietly-designing-building-silicon-empire
[doublepost=1522796129][/doublepost]

http://fortune.com/2015/10/13/apple-merge-ios-os-x/

Tim Cook - “We don’t believe in having one operating system for PC and mobile,” Cook said, according to those in attendance. “We think it subtracts from both, and you don’t get the best experience from either. We’re very much focused on two.”

So unless Tim changed his mind......

Not really changing his mind. MacOS will just be iOS with a mouse/keyboard UI layered on top. Different "OS" in form but not function.

 

[techwhiz]

Not really changing his mind. MacOS will just be iOS with a mouse/keyboard UI layered on top. Different "OS" in form but not function.

So you don't understand.
iOS *IS* MacOS stripped of features.
Different GUI but they both started as the same BSD OS.
Tim Cook's indication is that they will not merge back.

 

[AppleScruff1]

If Apple wants a large market share they need to:

1) Drop the price significantly.

2) Create a better file browser than Finder. Check out Path Finder or Windows Explorer.

3) Include a supported VM by default, so you can easily run Linux and Windows (embrace and extend). Even if few uses Linux, power users are the ones who recommend computers to others.

4) Support NVIDIA and work for better compatibility with (Windows) games and GPU acceleration (AI etc)

5) Switch to a robuster / faster Web browser by default, or make Safari on par with Firefox,Chrome or Edge.

Switching to a new CPU architecture will not help, and it's not what customers are asking for.

They don’t want a bigger market share, they want bigger profits.

 

[Falhófnir]

I wonder if they will bifurcate the line into iMac and iBook (In house chips) and a continuing line of iMac Pro, Mac Pro and MacBook Pro (intel chips) until such a point as one (presumably intel in their plans) are so outclassed they are finally dropped?

 

[DNichter]

Seeing a lot more openness in the thread is refreshing. Too many will hold onto their old ideas forever. Embrace the change, it could potentially be better in the long run. There’s nothing to be afraid of.

 

[gnasher729]

Wow! I remember the transition from power PC chips to Intel. Yes, I am that old and have been around Apple that long. This will be quite the transition similar to that time.

One difference is that the ARM64 and x86_64 programming models are very, very similar. So similar that 90% of iOS developers build Intel-versions of their apps to run on their Mac under a simulator. PowerPC and Intel used different byte ordering, which was a major problem at the time. ARM and Intel are the same that way.

 

[pankajdoharey]

Bottomline is Apple or Tim Cook are not naive, they are in tech business since the inception, no body has illusion that an Apple chip can compete with Intel until they actually do. If Apple has decided on this, then there must be a reason.

 

[gnasher729]

I don't think it would be a massive challenge to convince Adobe and Microsoft to re-write their desktop apps for the Mac. Not that I think that moving away from Intel would be all that great of an idea.

Hint: There is no ''rewrite". There is recompiling with different compiler switches.
[doublepost=1522797405][/doublepost]

Please no please no please NO! This is a TERRIBLE idea! Intel leads the industry in high power CPUs. Going any other route besides Intel will just result in slower computers!

The ARM chip in and iPad 12.9" is bloody fast, while still optimised for low energy usage. For a desktop machine, you jus put four of them into a case, and that will outrun anything but the high end MacPro.

 

[pankajdoharey]

Any Mac, Linux or Windows OS .
You should try it one day; computers can do wonderful things .

They both are different types of Devices for different applications. How is a Mac more modern than an iOS, i still didn't get the point you are tryign to make.

 

[gnasher729]

iOS *IS* MacOS stripped of features.

iOS has lots of features that are missing in MacOS.

 

[Sydde]

I wish I could remember the name of the chip, but Intel Macs produced in late 2007/early 2008 included a hardware chip which prevented installation of non-blessed OS software (translate "Linux").

But that is the opposite of what I meant. If Apple were concerned about Hackintosh, they would have had logic in their machines that OS X/macOS would demand, so that if someone wanted to install on alternate hardware, the OS would see that at boot and refuse to run (or brick, even). Since they did not do that, evidently it did not bother them.

 

[gnasher729]

If Apple wants a large market share they need to:

1) Drop the price significantly.

2) Create a better file browser than Finder. Check out Path Finder or Windows Explorer.

3) Include a supported VM by default, so you can easily run Linux and Windows (embrace and extend). Even if few uses Linux, power users are the ones who recommend computers to others.

4) Support NVIDIA and work for better compatibility with (Windows) games and GPU acceleration (AI etc)

5) Switch to a robuster / faster Web browser by default, or make Safari on par with Firefox,Chrome or Edge.

Switching to a new CPU architecture will not help, and it's not what customers are asking for.

Any of your five points, do you have any actual evidence? I think none of these are things that 99% of customers are asking for.

But that is the opposite of what I meant. If Apple were concerned about Hackintosh, they would have had logic in their machines that OS X/macOS would demand, so that if someone wanted to install on alternate hardware, the OS would see that at boot and refuse to run (or brick, even). Since they did not do that, evidently it did not bother them.

Actually, Apple has done exactly that for about forever. There is one chip on every Apple motherboard containing a 64 bit code that is required for decrypting essential parts of MacOS. This chip is easily simulated, but doing that to run MacOS an a non-Apple computer means you are commitiing a DMCA infringement. Apple doesn't go after individuals for this, they used it to stop Psystar back in the day (which was actually ordered to pay $2500 for every computer running MacOS that they sold, except they hand no cash to pay), and of course if Dell, HP, Lenovo decided to make MacOS compatible computers without Apple's permission, that is what they would expect to pay.

 

[Queen6]

Well, so it's beginning. I wonder how many people will choose to move on from Apple due to lack of X86 support. I for one will not be buying a Mac that I cannot run windows on

Given recent efforts don't expect very much, just sleeker shinier toys. Another factor is Intel, I don't see them taking such a move likely, no doubt Intel will respond in anger which is great for the consumer.

Going to happen sooner or later Tim & Co have had the Mac's cards marked for years. If they cant cull the desktop hardware & OS, they look very likely to dilute it ever more, to point when it's just a glorified IOS device that can be marketed at a higher price point...

Q-6

 

[ActionableMango]

Translating CISC code to RISC on the fly will result in colossal performance loss.

Intel processors are famous for being CISC, but all Intel processors since the Pentium Pro were redesigned with a RISC core. CISC instructions are translated/divided into smaller parts and then executed by the RISC core.

I see no reason why Apple couldn't do the same by adding a CISC translation engine to their ARM CPUs.

 

[Mescagnus]

You are not doing whole lot on your Mac. I have monthly crashes on all of my Macs.

Whatever you're doing, you must be doing it wrong.

 

[aaronage]

1. I think you miss the point and no making an ARM a desktop processor is a non-trivial exercise. The current A11 doesn't support visualization and threads. The current A11 is not a desktop processor. Besides the DDR channels you need memory scheduling. Adding more cores, you need additional snoop filters in the cache.

Also they need at least a laptop, desktop and a server class CPU.

3. The point was that iOS and MacOS aren't that different anyway.
4. Even though it's custom, they can't do any thing to modify the instruction set. So any customization must be in execution of existing instructions or it's not an ARM. Yes, they can add custom acceleration outside the core. Crypto, etc.
5. once again synthetic benchmarks don't measure system performance and an A11 is not comparable to an i9.
7. Samsung did have a server group that they failed to sell and disbanded. The IP may have been acquired by Cavium, but not the people.

I still disagree that the Apple A11 is a desktop class core, but we will have to "agree to disagree".
I never said ARM meant lower power. In actuality the metrics and characteristics change dramatically for computing on an iPad and a general purpose laptop. iPads are primarily for consumption and lack the performance for multi-stream video encoding for instance.

Let's just summarize this:
Can Apple assemble a team to make processors to compete with Intel? Absolutely.
They might have to pay for some of the patents they will infringe on, but they have cash.
Is it the A11 family? Absolutely not.
Is it a two year project to develop a family of laptop/desktop processors? Not a chance.
The A-series processor is on a one year cycle and its an evolution of previous technology.
How long to replace Intel? More than two years if they ahve to add virtualization and threads.

It all boils down to ROI.
Can Apple make an argument to shareholder for the return on investment?
I don't know, but if they do this, they will need to.

BTW see Sept. 2017
https://www.extremetech.com/mobile/256730-apple-quietly-designing-building-silicon-empire
[doublepost=1522796129][/doublepost]

http://fortune.com/2015/10/13/apple-merge-ios-os-x/

Tim Cook - “We don’t believe in having one operating system for PC and mobile,” Cook said, according to those in attendance. “We think it subtracts from both, and you don’t get the best experience from either. We’re very much focused on two.”

So unless Tim changed his mind......

"The current A11 doesn't support visualization and threads" I don't think a CPU needs SMT to be considered desktop/server class. Virtualisation definitely. I'd be surprised if Apple doesn't support this already (though I couldn't find hard info.) ARM themselves had that with the ol' A15 core.

Of course it's not trivial to design a desktop part but considering Apple's team has literally designed a core at least as good as Intel's best along with everything needed to support that, they clearly have the engineering capability for it. It doesn't feel like a big stretch, they have the talent and funds.

"Also they need at least a laptop, desktop and a server class CPU." I personally would expect distinct designs for iPhone, iPad/MacBook, MacBook Pro/iMac and Mac Pro - so 4 distinct SoCs based on the same IP.

"3. The point was that iOS and MacOS aren't that different anyway." I agree
"4. Even though it's custom, they can't do any thing to modify the instruction set. So any customization must be in execution of existing instructions or it's not an ARM. Yes, they can add custom acceleration outside the core. Crypto, etc." that's basically what I said too. Crypto, SIMD etc. are all there in the ARMV8 spec.
"5. once again synthetic benchmarks don't measure system performance and an A11 is not comparable to an i9." Apple's core is comparable to Intel's "lake" architecture, though. I'm not basing that only on synthetics (I know anyone bigging up Apple cores is assumed to have looked at the Geekbench numbers alone, I have looked a little deeper )

"Is it the A11 family? Absolutely not." A10X/A11 would be great for the MacBook (Core m) but I agree higher models would need a different design, and that's what I'd expect.
"Is it a two year project to develop a family of laptop/desktop processors? Not a chance." This rumour has been around for years. I expect the work has been underway for many years now.
"The A-series processor is on a one year cycle and its an evolution of previous technology.
How long to replace Intel? More than two years if they ahve to add virtualization and threads." It's not going to happen overnight, that's for sure!

Regarding the Tim Cook quote. A transition to ARM wouldn't nullify what was said. They can remain distinct platforms. I think we agree in that I don't see Apple merging iOS and macOS regardless of this potential ARM transition.

Regarding ROI. Couldn't you make the same argument for iPhone? They could have continued using off the shelf Qualcomm/Samsung/whoever SoCs, they could have used off the shelf Cortex cores, they could have continued using PowerVR etc. They're rumoured to be working on their own radio too. Why spend the cash when there's a generic solution available?

 

[Freyqq]

So you don't understand.
iOS *IS* MacOS stripped of features.
Different GUI but they both started as the same BSD OS.
Tim Cook's indication is that they will not merge back.

Not really. It may share some of the same codebase, but an iOS app can't run on OSX without a lot of work.

 

[imacken]

Well, so it's beginning. I wonder how many people will choose to move on from Apple due to lack of X86 support. I for one will not be buying a Mac that I cannot run windows on

+1.
My new 2017 iMac could be my last if this true.

 

[gnasher729]

Thanks for the info. So if Apple developed a new chip/architecture for their laptops and desktops and created the tools needed for development and support of the platform, you don't feel that developers would consider adoption of that platform? I am trying to understand the details, but it seems like wouldn't consider any other platform viable in the future. I'd imagine it is iOS based, but built for the point and click environment. Your insight is valuable here so I appreciate it, I know my views can be short sighted.

Apple wouldn't develop a new chip, the would take an ARM chip. Maybe one with more cores than usually because power consumption is not as important for a desktop / laptop as for a phone / tablet.

Apple wouldn't create any new tools, because the tools are all available. You just use Xcode and tell it to produce code for macOS or for iOS, and for which processor. You can build ARM versions of any macOS app today, except that you won't find a Mac today able to run it.
[doublepost=1522800473][/doublepost]

Not really. It may share some of the same codebase, but an iOS app can't run on OSX without a lot of work.

It's about two seconds work. In Xcode, don't choose a real device as the target but any of the provided emulators, and the app will run on macOS just fine.

 

[Sydde]

How long to replace Intel? More than two years if they ahve to add virtualization and threads.

Explain what you mean by “threads”. Because iOS is definitely a multi-threaded system with PMT. I remember Mac OS 7-8-9, which were not threaded, and you can tell from experience that iOS is. It even goes beyond threading, with GCD.

Do you mean hyperthreading?