I do not necessarily disagree. But I cannot see the connection with ARM. ARM based computers are as open or closed as the manufacturer designs it to be.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Intel Expecting Apple to Transition to Custom ARM-Based Chips Starting in 2020
- Thread starter MacRumors
- Start date
- Sort by reaction score
I do not necessarily disagree. But I cannot see the connection with ARM. ARM based computers are as open or closed as the manufacturer designs it to be.
Of course you can run Windows and Linux, they both run on ARM processors.
As for your unsupported claim, the issue is not "high-end" chips, the issue is low power performance, and Intel gave up long ago trying to compete with ARM. In the battery powered, portable space ARM is continuing to thrash Intel, and moving into laptops is just more of that progression.
[doublepost=1550970011][/doublepost]
Did you actually read that article? Linus's point is that people need personal machines that use the same architecture as the servers, which this transition would provide. That point is even made in the article.
Apple does have to stick to the ARM instruction set, their license is such that the CPU will always be able to execute 100% of ARM compiled machine level code (assuming it’s only 64 bit code). But, Apple could do things at the motherboard level that would be considered proprietary, because, like you said, they don’t have to be concerned about anyone else using these. It could even be that to have a non-native OS run in an ARMac, you would have to reverse engineer what all the Apple-only parts are doing. Any future Hackintosh is definitely possible, but it could become a lot more difficult for the untrained to do.
I was looking for this story earlier... this is what a company that owns the chip, compiler, operating system is able to accomplish.
https://www.infoq.com/news/2018/11/iphone-xs-javascript-performance
The ARMac is going to scream, the only question is how far ahead of Intel they will be
Apple does have to stick to the ARM instruction set, their license is such that the CPU will always be able to execute 100% of ARM compiled machine level code (assuming it’s only 64 bit code). But, Apple could do things at the motherboard level that would be considered proprietary, because, like you said, they don’t have to be concerned about anyone else using these. It could even be that to have a non-native OS run in an ARMac, you would have to reverse engineer what all the Apple-only parts are doing. Any future Hackintosh is definitely possible, but it could become a lot more difficult for the untrained to do.
I was looking for this story earlier... this is what a company that owns the chip, compiler, operating system is able to accomplish.
https://www.infoq.com/news/2018/11/iphone-xs-javascript-performance
The ARMac is going to scream, the only question is how far ahead of Intel they will be
Have you seen their license? I have not.
(And of course they could make any arbitrary processor not using any ARM IP and my same argument holds.)
Actually, they could POSSIBLY get cheaper, since they’re not paying a chunk of cash to Intel for their CPU’s and x86 license. However, it remains to be seen if they’d pass along those savings.
It’s from here:
https://en.m.wikipedia.org/wiki/ARM_architecture#Licensing
“Companies can also obtain an ARM architectural licence for designing their own CPU cores using the ARM instruction sets. These cores must comply fully with the ARM architecture.”
This is why Apple was able to produce a 64-bit only version of the chip even though ARM had not designed a core schematic to sell. At the time, the only 64-bit designs ARM offered also included 32-bit compatibility.
And you’re right they absolutely could make their own arbitrary processor and I wouldn’t doubt their ability. I mean, they own the entire path and could simultaneously keep the compiler updated in lock step WITH the development of a new chip.
I hadn’t thought about that, but I remember they were using PowerVR reference designs in much the same way... up until they figured out how to do graphics themselves.
I’m not sure they signed the same architectural license as everyone else. I’m also sure whatever they signed is secret. In any event, once they are off x86 and they are supplying all their own processors, there isn’t any particular reason to stay with ARM unless they think ARM has some advantage for them.
All the people who need emulation or Windows will leave, then Apple can kill the Mac and make iBooks.
Back into obscurity? Sure, why not...
Intel is the reason I bought a Mac when I did and truth be told if it hadn't been for Bootcamp I may not even have gotten into Macs. These days a Mac is only a "to have the OS and keep using my Mac stuff" kind of device anymore anyways.
I'll probably buy an ARM Mac too, but obviously eventually the new Rosetta (whatever they'll call it, I'll just nickname it Rosetta 2 for now) will be discontinued, some of my more obscure and older Mac apps will die with it, they ones that haven't already when 32bit support got axed and then I'll be left with the situation: keep an old Mac around for legacy stuff, the rest happens in Linux and Windows going forward.
Good luck flexing on Intel, Apple. Just because Windows 10 has builds for ARM doesn't mean it solves the problem of the transition. And Microsoft is probably wise enough to not strip away x86 support for Windows 10 for ARM as soon as I expect Apple to pull it. (to push devs to updating legacy applications... alright whatever)
Glassed Silver:mac
Intel is the reason I bought a Mac when I did and truth be told if it hadn't been for Bootcamp I may not even have gotten into Macs. These days a Mac is only a "to have the OS and keep using my Mac stuff" kind of device anymore anyways.
I'll probably buy an ARM Mac too, but obviously eventually the new Rosetta (whatever they'll call it, I'll just nickname it Rosetta 2 for now) will be discontinued, some of my more obscure and older Mac apps will die with it, they ones that haven't already when 32bit support got axed and then I'll be left with the situation: keep an old Mac around for legacy stuff, the rest happens in Linux and Windows going forward.
Good luck flexing on Intel, Apple. Just because Windows 10 has builds for ARM doesn't mean it solves the problem of the transition. And Microsoft is probably wise enough to not strip away x86 support for Windows 10 for ARM as soon as I expect Apple to pull it. (to push devs to updating legacy applications... alright whatever)
Glassed Silver:mac
I'm late to the discussion anyway, but I'd like to add my thought on this.
As long as Arm processors provide better performance to justify inherent disadvantage of abandoning x86 (no native x86 apps through Windows, and option to dual boot), I'm ok with it. However, going Arm meaning giving up the performance, this will mark the end for me.
Personally, I'm curious about the multi-tasking performance of Arm processors as running synthetic benchs on ipad pro is not grasping the true performance level. Nowadays, I do so many things simultaneously at once that the raw power of multi-core cpu Intel and AMD is providing is essential for my need. Arm is yet to be proven on this field.
I just hope that Apple's reasoning is not based on forcing its users to adapt to less powerful, ipad like Mac OS with gimped multi-tasking.
As long as Arm processors provide better performance to justify inherent disadvantage of abandoning x86 (no native x86 apps through Windows, and option to dual boot), I'm ok with it. However, going Arm meaning giving up the performance, this will mark the end for me.
Personally, I'm curious about the multi-tasking performance of Arm processors as running synthetic benchs on ipad pro is not grasping the true performance level. Nowadays, I do so many things simultaneously at once that the raw power of multi-core cpu Intel and AMD is providing is essential for my need. Arm is yet to be proven on this field.
I just hope that Apple's reasoning is not based on forcing its users to adapt to less powerful, ipad like Mac OS with gimped multi-tasking.
Apple is planning to ditch Intel and transition to Mac chips starting in 2020, based on multiple rumors we've heard in the past from Bloomberg. Axios today confirmed Bloomberg's reporting and said that multiple sources have suggested Apple will transition to custom ARM-based chips next year.
According to Axios, developers and Intel officials are expecting Apple to begin using ARM-based chips in 2020.
The move to ARM-based chips is said to be part of Apple's effort to make Macs, iPhones, and iPads work together and run the same apps. Bloomberg earlier this week said that by 2021, Apple wants developers to be able to create one app that will work on iPhones, iPads, and Macs.
Apple's transition to a single app for all devices has already begun. Last year, Apple ported several of its iOS apps, such as Voice Memos, Stocks, and Home, to macOS. This year, Apple plans to let developers transition iPad apps to macOS, and in 2020, that will include iPhone apps. In 2021, then, developers will be able to make just one app that users can download on any of Apple's platforms.
This transition will greatly increase the number of Mac apps available, and it will cut down on the amount of work developers have to put in to create a Mac app. It will also better unify Apple's operating systems across all of its devices.
There have been rumors about Apple transitioning to ARM-based Macs for years now, and they have ramped up given the many Intel chip delays that have resulted in subsequent delays for Mac products. With its own ARM-based chips, Apple will not be tied to Intel's chip release cycles.
Apple already makes its own A-series chips for the iPhone and the iPad, and there are also custom Apple chips in recent Macs -- the T2. The T2 chip, in the iMac Pro and 2018 MacBook Pro, MacBook Air, and Mac mini models, integrates several components including the system management controller, image signal processor, SSD controller, and a Secure Enclave with a hardware-based encryption engine. It powers the Touch Bar in the MacBook Pro and the Touch ID feature in the MacBook Pro and MacBook Air.
Apple is a major Intel customer, responsible for approximately five percent of Intel's annual revenue, so the transition to ARM-based chips will be a major blow for Intel, but a win for customers in the long run. Apple's modern A-series chips for iPhone and iPad are already more powerful than many Intel chips on the market.
Article Link: Intel Expecting Apple to Transition to Custom ARM-Based Chips Starting in 2020
Maybe if the new Mac Pro is truly expandable, we'll wind up having the equivalent of the old DOS compatibility cards that used to be available for Nubus and later PCI.
Attachments
The "point" the article makes is the "user" architecture and "server" should ideally be the same. We don't use ARM servers and Apple transitioning to ARM desktop chips won't change this.
I'm just pointing out what someone said; I never gave an opinion that falls on either side of the argument.
...provided someone builds a version configured to support whatever the actual "ARMintosh" hardware turns out to be, with appropriate drivers etc. For an operating system, the fact that the OS already exists for ARM makes an ARMintosh version far more likely, but it won't just spring into existence by itself. One example: will Apple release open-source drivers, specs & any needed firmware for whatever GPU and other hardware accelerators in their chip?
...and although it would probably be more economical to go with ARM, having a proprietary processor wouldn't necessarily be a problem.
People are fixating on the CPU instruction set as if that were the only thing and forget that the "x86/amd64" label drags with it a whole set of de-facto standards descended (if much expanded and mutated) from the IBM PC.
Its 2019 - much code that would have to be written in CPU-specific assembler 15 years ago can now be written in portable C/C++ - stuff that would have been written in C/C++ is now capable of being written in Javascript or Python. Apple has already stated that the next version of MacOS will drop 32 bit support, so everything is written for a 'clean' 64 bit model. Even operating systems and drivers can be written predominantly in high-level language and modern compiler toolchains use some sort of intermediate bytecode so only the bytecode-to-binary backend needs to be CPU specific (...assuming they even bother with binary: Java, Android and MS .net all use a runtime virtual machine).
...what an OS can't do is have drivers for graphics, networking, hardware acceleration etc. - in any language - if the hardware specs are closed or depend on proprietary firmware.
The instruction set is far less of an issue today that it was in the past, and if we're talking about 2022/23 before Apple totally pulls the plug on Intel it will be even less important by then.
What is important is the OS-level API and frameworks used by C/C#/ObjC/Swift whatever - because using frameworks like Metal and Accelerate will be what takes advantage of the hardware bells and whistles that are likely to give the A-series chipset its edge over Intel. Alongside that, the 'core' CPU instruction set is almost irrelevant for most development. Apple have already shown an intention to push developers into using standard frameworks - as well as a regrettable tendency to depreciate 'cross platform' solutions like OpenGL.
...its not just pure performance. Currently Apple is stuck with whatever dozen-or-so permutations of i3/5/7, clock speed, mobile vs. desktop, nunber of cores, iGPU power etc. Intel deigns to release. With mix-and-match A-series chips, Apple can build the system-on-a-chip to fit the form factor rather than vice-versa. That's far more suited to Apple's design approach (I don't think its any secret that they're primarily concerned with ultra-portables, SFF and all-in-one systems).
Actually, that should be ARM's superpower: a single ARM core might be less powerful than a comparable Intel core, but the cores are smaller, simpler, and ARM is beating Intel in the die-shrink race: ARMs way of beating Intel is packing more cores (plus extra SIMD/vector units/shaders and other hardware acceleration gizmos) into the same space and thermal profile.
We've had 8-core ARM smartphones for years...
So if (for example) the wilder speculation is true and the new Mac Pro turned out to be the first ARMintosh - expect it to start at 16 or 32 cores + a shedload of on-chip hardware accelerators chosen to make FCPx fly.
That's kinda why there's all the interest in ARM right now: so far, Intel has always ruled on single-core performance (well, unless you go back to 1987 when the 286 was choking on the ARM2's dust) but now they're hitting the wall, struggling with smaller die sizes and they're having to look at more cores as the way forward.
...from the POV of the lead developer of a monolithic OS kernel: AKA one of a shrinking pool of "developers" who still need to give a pair of foetid dingoes kidneys about what CPU instruction set they're running on.
Possible but incredibly unlikely. Adds lots of complexity to the design, complicates the OS tremendously, etc.
I don't see it as a move to a universal OS as much as it is a move to a universal development platform based on ARM + UIKit
Exactly. Two main goals:
1) open up macOS to the massive iOS development community by allowing UIKit apps to work well on macOS (hopefully by improving UIKit to add features like menus, key up/down events, etc.)
2) allow much more form factor flexibility and roadmap predictability in Macs by moving them to a CPU that Apple controls. This will allow them to eliminate hacks like the T2 and build the secure store right into the processor, add neural engines and the like, bring parity with iOS to thinks like iMessagea apps, and do lots of things we haven’t even thought of yet.
There is no reason an ARM chip can't be made as powerful as an Intel one. ARM instruction density is lower than x86-64, but not by all that much.
Apple have proved they have a chance at it. There is no real question that Apple's ARM chips have been more powerful than those from competitors like Qualcomm or even ARM themselves.
There are only a very few cases where new programs will have any code written in assembler. That's likely to be for things like SSE. ARM has NEON to perform similar tasks. But a lot of that kind of code has been offloaded to the GPU, and GPU shader code is not architecture-specific in the same way.
There will be some pain for app developers, but that is par for the course for Apple. The deprecation of OpenGL, the removal of Carbon, the switch to x86 in the first place (which took us from big-endian to little-endian, a much trickier change than a switch to another little-endian 64-bit architecture like ARM). If developers have come along this far then at least some of them will come along again. Unlike some of the other controversial changes (headphone jack, pricing, touch bar, keyboard) this change could bring big benefits when it comes to cost and battery life.
Old low-level code for things like sound processing will probably have to be emulated, or people will have to switch to newer versions compiled for ARM.
Windows AAA game developers aren't super interested in the Mac anyway - but I expect a lot less of their code is x86 specific than ten years ago. The switch to Metal/lack of Vulkan support is likely to trip up Windows game developers a lot more than switching the CPU. Remember there's potentially full, unlocked Windows on ARM to support now, too.
I imagine Apple probably already have internal prototypes, running macOS 10.14 recompiled for ARM and doing graphics with Metal on the A12X GPU.
As great as x86 has turned out to be - a lot of the criticisms from the 90s did turn out to be overblown - I think this could be a great move, as long as:
If ARM starts gaining more traction in the cloud then an ARM developer machine might be even more desirable.
Apple have proved they have a chance at it. There is no real question that Apple's ARM chips have been more powerful than those from competitors like Qualcomm or even ARM themselves.
There are only a very few cases where new programs will have any code written in assembler. That's likely to be for things like SSE. ARM has NEON to perform similar tasks. But a lot of that kind of code has been offloaded to the GPU, and GPU shader code is not architecture-specific in the same way.
There will be some pain for app developers, but that is par for the course for Apple. The deprecation of OpenGL, the removal of Carbon, the switch to x86 in the first place (which took us from big-endian to little-endian, a much trickier change than a switch to another little-endian 64-bit architecture like ARM). If developers have come along this far then at least some of them will come along again. Unlike some of the other controversial changes (headphone jack, pricing, touch bar, keyboard) this change could bring big benefits when it comes to cost and battery life.
Old low-level code for things like sound processing will probably have to be emulated, or people will have to switch to newer versions compiled for ARM.
Windows AAA game developers aren't super interested in the Mac anyway - but I expect a lot less of their code is x86 specific than ten years ago. The switch to Metal/lack of Vulkan support is likely to trip up Windows game developers a lot more than switching the CPU. Remember there's potentially full, unlocked Windows on ARM to support now, too.
I imagine Apple probably already have internal prototypes, running macOS 10.14 recompiled for ARM and doing graphics with Metal on the A12X GPU.
As great as x86 has turned out to be - a lot of the criticisms from the 90s did turn out to be overblown - I think this could be a great move, as long as:
- The boot loader is standard unlocked UEFI and people have the option to Boot Camp into Windows on ARM, or install Linux. The fact Linux can't read a T2-secured SSD is already putting people off.
- macOS itself remains macOS and there is no lockdown or restrictions for non-App Store apps.
- The machines still have Thunderbolt. So far not even AMD laptops have Thunderbolt - if it remains Intel-only Apple might have issues. I don't think asking people to switch connectors again so soon is feasible.
- It still runs Photoshop and Microsoft Word.
- The keyboards are good.
- There is still a headphone jack. Please.
If ARM starts gaining more traction in the cloud then an ARM developer machine might be even more desirable.
The context was the game Civilization. Aspyr is a company specialised in porting Windows games to the Mac. Porting an application from Windows to Mac is such a huge undertaking that the developer and publisher (Fireaxis/Take-Two) outsourced it to Aspyr.
Aspyr's ability to port does not require Boot Camp nor I am sure they would benefit from it.
So what Aspyr does is: Windows application on Intel -> Mac application on Intel
What an ARM Mac will require of the developer is Mac application on Intel -> Mac application on ARM. It is a completely different kind of problem and it is extremely simpler since the operating system and therefore the APIs are the same.
We do not know the details, but it would probably only require a recompile of the source code for most application. A recompile takes a few minutes to maybe an hour. Even if you include all the work getting a new executable in Mac App Store or on your website, we are talking a few days.
For a small business which is not a developer, it depends on the exact details how macOS on ARM will work with regards to Windows programs. There might be a version of Boot Camp running Windows on ARM. Or maybe there will be a virtual machine support or some kind of emulation.
But let us assume that it will be impossible to run Windows programs on macOS for ARM. All that a small business will need to do is buy a PC at the time the current Macs they have needs a replacement.
I consider you a technical Mac user and would not be surprised that you run Boot Camp. But with the exception of Windows games, I would suggest using a virtual machine which is superior to Boot Camp.
What I took issue with is the notion that Boot Camp is used by a large majority of Mac users. I believe a very small minority of Mac users use Boot Camp and then mostly for gaming. Even if you include everyone who uses Boot Camp and virtual machines I do think it is less than 10% av Mac users. Maybe as few as 2-3%.