Apple Inc. is planning to use its own chips in Mac computers beginning as early as 2020, replacing processors from Intel Corp., according to people familiar with the plans.
The initiative, code named Kalamata, is still in the early developmental stages, but comes as part of a larger strategy to make all of Apple’s devices — including Macs, iPhones, and iPads — work more similarly and seamlessly together, said the people, who asked not to be identified discussing private information. The project, which executives have approved, will likely result in a multi-step transition.
There is no doubt that Apple knows how to build fantastic chips for its mobile devices, and if this plan received the go-ahead from executives, they must have fast Mac chips on the horizon. When Apple switched to Intel years ago, the transition was pretty smooth—If this is going to happen, the transition will need to be just as smooth.
That’s because Apple had secretly maintained an X86 port of OSX from the beginning…
Switching to new chips will, imo, simply kill OS X as a platform, since EVERYONE will have to rewrite and recompile everything.
One of the HUGE drivers for the acceptance of OS X was it’s underlying Unix platform and that it was intel-compatible which opened tons of OSS to the Mac.
As it is, this flat kills running Windows or a Windows VM on a Mac, and while most people who just use their Mac don’t really care, this is something that’s let me actually have a %@$# career as a Systems Administrator and use a Mac at the same time.
I REALLY don’t want to go back to the dark ages of running emulators instead of VM’s.
I don’t WANT my Mac to just be a big iPhone…
… what?
1) Apple’s had an ARM port of OSX since at least 2007, for iOS, no? iOS is OSX at its heart, and I would equally be shocked to find that Apple didn’t have full-real-honest-to-god OSX running on whatever the fastest ARM chip it has lying around it just as a proof-of-concept.
2) What makes you think people have to rewrite anything just because the target architecture changed? Unless you have a low-level dependence on the CPU for data packing or endianness or you’re doing embedded assembler, you shouldn’t need to touch a single line of code. And very, very, very few projects care about that stuff, as is wise and correct.
3) Telling XCode to emit a fat x86/ARM binary, if they go that route, should take about 30 seconds and require the same basic compilation as … building for x86 by itself. “A recompile” is not remotely a scary thing, I say, as a professional software developer. Builds are cheap. Honest.
And remember, Apple’s done this twice before, and there was not a Great Application Crisis; I was using Macs during both transitions and it just wasn’t a huge deal, either 68k-PPC or PPC-Intel.
4) OSS running on OSX works because OSX is a POSIX-compliant Unix(tm), not because it happens to run on Intel CPUs; you can run nearly any OSS project of any size or importance (apart from virtualization systems, as you note, and apart from the caveats above) on any platform you feel like. OSX doesn’t natively run non-OSX OSS binaries, even if they’re compiled against an Intel CPU; OSX, note, doesn’t even use the same binary format as e.g. Linux [Mach-O vs. ELF].
That’s the whole point of POSIX standards and all that stuff; it’s not platform and OS tied!
5) Yes, it would be horrible for anyone who has to run Windows or any x86 environment.
More importantly, it’d be horrible for the much larger set of people for whom an A11 is not sufficient oomph.
(“Desktop class” does not convince me, Apple.)
ARM is taking over the data center. No reason that it can’t take over the desktop too.
Apple also transitioned from 680×0 to PowerPC in the 90s. Apple is very good at these sort of things even before they bought a chip company.
I somewhat agree with BruceJ’s comment. The killer question is, what is Apple going to do for all their customers who rely heavily on Boot Camp? For those customers who, without the ability to run Windows in addition to Mac OS, might find themselves unable to justify buying a Mac (because they cannot make do without Windows, and if they can only afford one PC it’s going to be a Windows-capable machine)?
Making a translation layer like Rosetta to enable you to run intel-based Mac OS apps on your ARM based Mac is easy enough – Apple already makes the opposite to allow phone developers to run a virtual iphone on their macs. So the OS X/Mac OS ecosystem will survive OK.
But for people who need either Boot Camp or a Virtual Machine running Windows (or both), that won’t be enough. Current VM software (virtualbox, etc) doesn’t support CPU emulation. Every CPU emulator that I’ve seen has run hot, has added heavy CPU overhead, has not been very fast, and has been buggy. Apple will have to have a first party solution that works far better than anything else currently out there.
It might be easier for them to just pay a few billion dollars to AMD or Intel for a licence to the x86 instruction set, and then build x86 emulation directly into hardware, sort of like how the original PowerPC Macs had hardware level 68000 support built in.
eta: forgot to say, in addition to the relatively small number of Mac customers who absolutely require Windows for something or other, this change will also hurt the rate of people switching to the mac — knowing that they can run Windows if they have to is probably an important psychological safety net for people on the fence about switching OS’s.
Early reviews of Windows 10 running on ARM report that it feels like running on a Core M laptop. There are some early limitations like no 64-bit or Hyper-V support yet (apparently those will be addressed in later releases) but overall it sounds pretty good for a first release. The good news is that 15-20 hour battery life is a real.
Note: these Windows laptops are generally running on Snapdragon 835 chips which are close to 2 years old and were already 1-2 years behind Apple A series when released. Interesting times ahead.
Given how badly their notebooks have been held back by Intel, this isn’t the least bit surprising.
I thought the new Windows OS support ARM…
I had forgotten about that.
??
Apple’s notebooks have been held back by Apple. Going USB C – only 5 years too early, significantly raising the price due to a feature almost no one wants, and shipping a garbage keyboard are not Intel’s fault.
I sure hope Apple has this stuff figured out. I use VMWare and Docker daily. The ability to run “real” Windows and x86 Linux is far more important to me than having a MacBook that is fractionally thinner, or has incrementally better battery life and CPU performance.
Those things aren’t holding them back. Just look at the sales numbers.
It will happen. I’ve said it for years.
What sales numbers? They don’t make them public. They announced in the quarter after the late 2016 models came out that sales were better than before… which doesn’t really say much. There was a lot of pent-up demand because those models were badly delayed, the market is mostly a captive audience, and it wasn’t clear how bad a problem the butterfly keyboard was at that point. Since then, there has been no material increase in overall Mac sales, and little reason to believe the MBP has been a big hit.
I often get asked by people what MacBook to buy, and there is no good answer right now. The 13″ MBP with touchbar is an expensive donglefest with a keyboard that is flawed by design. The 15″ and 13″ without touchbar have similar flaws. The 12″ MacBook is a niche product (again with a flawed keyboard). The once-great Air hasn’t been significantly updated in years.
Switching to ARM doesn’t solve any of these problems.
5,112,000 Macs sold in Q1 2018 (most of which were laptops) for $6,895,000,000 in revenue. Doesn’t sound like “holding them back” to me.
Both prior transitions were pretty smooth, but the transistion to Intel was smoother.
Part of it was that there was an Intel version of the OS running in a skunkworks lab ready to kick start the process, but most of it was that the Power PC chips were SO far behind where Intel was, that apps compiled for PPC ran FASTER on emulation on intel than they did on the old PPC chips.
I doubt that any transition away from Intel would be due to any great speed disparity as bad as the lagging PPC archetecture (effectively abandoned by Motorola and IBM ) was.
But there was also the fact that at the time of the intel transition, an easier means of reusing code from the PPC days than the previous transition to PPC.
BUT…
Does moving away from Intel mean moving to ARM?
Remember, there’s another company producing x86 compatable chips, AMD. Suggesting that transitioning away from Intel may not necessarily mean away from the x86 archetecture. So an Intel emulator may not be necessary.
Yeah, but nobody runs it and nobody cares about ARM binary compatibility for “Windows”.
People want to run Windows for x86 software, in the desktop/Apple context.
(ARM’s mostly for Embedded, since I think MS has given up on “cheap ARM tablets”, pretty much?)
I can understand Windows, but why “x86 Linux”?
I mean, there are Linux ports to ARM, and even for things that don’t have source, you wouldn’t think an ARM binary would be a big deal, what with ARM’s popularity increasing in datacenters, etc.
Apart from weird proprietary stuff, I’d expect ARM Linux to Just Work as well as x86 Linux, on a same-architecture host.
AMD’s chips aren’t better than Intel’s, though.
(I mean, they’re not bad.
But they’re not Obviously Superior; at best, generally, AMD is a wash, at least last time I checked into the competition, which was pretty recently, when I rebuilt my home gaming PC.
“Intel” here means more like “the entire licensed x86/IA32 family” more than “literally just Intel’s products”.
Apple could switch to AMD CPUs tomorrow if it felt like it, figuratively.
But there’s no compelling reason to.
Moving to in-house ARM chips has some more plausible benefits, especially for laptops (energy efficiency, mainly) – and complete control of the CPU design, which they won’t get from AMD any more than they will from Intel.)
The A11 chip has higher single core benchmark results than the i7 chips from three years ago. That’s on a chip that is designed for mobile usage. If Apple can get those kinds of results on mobile chips they should be able to do significantly better on a desktop chip. Apple also has the ability to add custom features or accelerators to their chips which could dramatically speed up certain tasks.
Most people just want to run the software they want to run and don’t care about the underlying architecture.
My point wasn’t that AMD made better chips, but to illustrate that other companies CAN make Intel-compatible chips. It’s not a proprietary lock-in. Therefore, Apple could theoretically be designing its own chips for the Mac that are x86 based, and not necessarily ARM based.
They were/are, according to Tim Cook in their financial reports, the best selling MacBook Pros in company history. You may not choose to take him at his word, but in that setting he can’t really lie. SEC trouble and all that stuff…
By held back I’m not talking about personal preferences or design quibbles. I’m referring to the issue Pros (I’m in this category) complain about; performance and memory. We don’t really care about USB-C ports (Apple is always early), the Touch Bar (neat, but needs a desktop keyboard equivalent to matter), or even the keyboard (although that needs to be improved to better deal with dust and debris, no argument there). Whatever, I’m attached to large 4k displays much of the time and using external keyboards.
We do care, a great deal, about CPU and GPU performance, and memory, and how they’re balanced in terms of power usage on a device that relies on a battery. Intel is the primary reason my mid-2104 MacBook Pro with 16 GB of RAM barely behind a MacBook Pro sold in 2018. They’re the reason Apple couldn’t move to 32 GB with the new MacBook Pro, and they’re the reason performance improvements have lagged terribly compared to mobile (phone/tablet) and desktops.
The last time this happened was with the G5 and PowerBooks. Apple took matters into their own hands then, and it’s the reason we’re on Intel now. I look for Apple to do the same thing this time, except now they’re in a position to do it entirely themselves.
If “desktop” means “casual users who might honestly be fine with an iPad if it wasn’t a tablet”, sure.
Currently ARM just can’t compete in serious single-thread power, which is what a lot of “desktop” is, once you get out of casual-land.
(Data centers? Care more about efficiency in power/heat than raw power. They can leverage parallelism much more than most desktop uses, natch.
Completely different problem space, and one where ARM has serious strengths, making it a compelling player.
But data centers aren’t like desktops.)
In theory. But in my experience there’s usually a gap between theory and practice, filled with a big time sink. I actually do have some “weird proprietary stuff” requirements. I don’t want to have to recompile everything from source. Historically, the JVM has been far faster on x86 than any other architecture. Etc..
Maybe it will all work, but I’m not confident that my priorities rank very high on Apple’s list.
Apple could ship an notebook with 32GB today. Windows vendors do. Intel makes the chips to support it, and Apple chose not to use them. Same with faster GPUs, and (to some extent) CPUs. They picked thin instead, and shipped the 15″ MacBook “Pro”.
By “casual users” I assume that you mean people wearing blue jeans.
Also, don’t think that the performance of Ax chips in phones is the best Apple can do. If this is only 2 years out, they have been secretly testing desktop-style chips with much higher thermal limits for years now, and you can bet those chips leave their mobile cousins in the dust.
They’d need to have a license to the x86 instruction set. Intel will not sell that, and IIRC, the deal between AMD and Intel does not allow AMD to sell x86 IP to anyone else.
Apple could buy AMD outright without making a dent in their cash reserves, but would it be a wise purchase?
Ah, I mistakenly assumed that the x86 archetecture wasn’t proprietary, given that AMD made compatable chips. A trip to the wikipedia page for AMD shows that it’s a rather… complicated history between them and Intel.