Jason Snell, Macworld:
This week’s report from Bloomberg that Apple is planning on moving the Mac to its own chips starting in 2020 is the culmination of years of growing speculation about the future of the Mac. I’ve been impressed by Apple’s use of ARM chips in new Macs while being skeptical about the prospects of a full transition.
But if we accept the Bloomberg report—and it’s from reporter Mark Gurman’s sources, which are generally excellent—it’s time to shift from speculating about whether or not Apple would do this and start to analyze why the company would make this move, and what form the transition might take.
Thoughtful piece by Jason Snell. Definitely worth reading.
Obviously, this post is based on speculation. But, as Jason says, Mark Gurman has an excellent track record. The question of why Apple would do this is an obvious one. If this is Apple’s plan, no one outside the company can answer it. But two things spring to mind for me.
First, moving the Mac to a chipset that they design and build would give them that much more control over the full stack. Less reliance on outside vendors, the ability to create a more efficient and more powerful set of devices.
Second, moving the Mac to the same chipset as the iPad would (and this is way out of my league conjecture here) make it that much easier to merge macOS and iOS, somewhere down the line.
I don’t get the ‘having the same processor would ease share apps’ argument. All iOS applications are developed on x86 processors and can run on x86 (if you use the iOS simulator).
If Apple want to have a single app for both iPad and macOS, they don’t have to change the Mac processors (which would kill BootCamp by the way), they just have to support fat (aka universal) binaries store submission (they don’t even have to support it at runtime as binary may be stripped on distribution).
The machine CPU and code sharing are completely orthogonal for every day developers (some low level tasks may require some work, but they are corner cases).
I thought there was going to be new low power DDR4 memory out by now that would ease the issue of maxing memory at 32 gigs instead of 16. But 16 seems to be what most people would need anyway. But 32 would be idea for us power users. Disappointing that there’s no new GPU for the 15 inch MacBook Pro. That means it’s much more likely now that I will get the 13 inch MacBook Pro and add an eGPU now that they’re officially supported. The Sonnet Puck seems ideal for traveling power users.
One thing that I do know, this will be one hell of an interesting WWDC.
Apple would only do this if they were confident they could produce powerful chips. But I was around for the PPC transition when everyone was confident RISC would destroy CISC… and the lesson there was “don’t discount intel’s ability to innovate.” Yes, their chips seem to have languished. But look what happened when AMD popped out with genuinely competitive chips. (I’m applauding AMD more than intel on that score). Other challenges exist too. I also remember how much trouble Apple had getting certain companies (I’m looking at you, Adobe) to sign on with the Classic to OS X transition. And as so many people have pointed out, one of the great advantages moving TO x86 was the ability to run Windows apps. Emulation was a thing during PPC, and almost nobody did it. I myself bought Windows machines when I wanted to play games, because emulation frankly sucks. Nevertheless, I’m excited to see what Apple plans to do. People declare this transition a failure or a success before they’ve even seen it. LOL. This year we won’t see any of that except for maybe Marzipan. That seems tremendously exciting. Is Marzipan also going to open up a third UI with AR too? Looking forward to WWDC.
The API issue is the only real challenge in a merger of “OSes” … by all accounts, iOS is really OSX/macOS at its heart, anyway, running the same Darwin kernel.
Just some differences in the UI layer and some different APIs.
(The compiler can, easily, handle the differences in how the CPU families work internally; that’s its job, after all.
LLVM is a fine compiler, and can handle it just fine – their documentation suggests that most of the competing architectures are simpler than x86/x64 and thus easier to output for.)
Is anyone discussing (speculating, wild ass guessing) how Marzipan (rumours + rumours) would support this transition?
For example, would Marzipan provide the direct support to the new architecture so iOS and Mac apps can run efficiently on ARM chipsets? And unify much of the API stack. And provide a better Swift native library?
Back in the day, all forms of life needed to be Carbon based run on OSX. I wonder if Marzipan be the path for Macs to transition from x86 to ARM.
x86 is only a small part of it. Intel Processors and Apple’s processors are very different animals. But even if they were identical and you only had to check a box to compile for one or the other, there’s also the challenges of libraries and APIs designed for touch or mouse and keyboard. Not to mention lots of other things that are in one OS but not the other. It’s easy to say it’s easy when you don’t know what doing it actually involves.
I agree, this will be a multi-year journey and the difficulty.
I don’t understand the whole run Windows (Bootcamp, virtualized, emulation, etc) argument. Windows 10 runs on a Snapdragon 850 ARM chip today. And x86 apps in emulation.
Granted there are limitations since this is a first release (64-bit to come, Hyper-V to come). However everyone across the industry is leaning into ARM (server, desktop, containers, data, storage). Problems being solved in one domain help other domains, and vice versa.
Second, moving the Mac to the same chipset as the iPad would (and this is way out of my league conjecture here) make it that much easier to merge macOS and iOS, somewhere down the line.
Naw.
Lots of Unixes (including macOS*/OSX itself) handle “running on multiple architectures” without any particular issue.
The only thing the same architecture gets you, really, is not having to have multiple binaries around for different architectures.
“Merging macOS and iOS” is a matter of changing APIs and UI guidelines to converge, not a matter of hardware.
(* I have a devil of a time remembering it’s “macOS” again, since I’ve recently been restoring my old Mac collection, and as far as I’m concerned “MacOS” isn’t capitalization-specific and means “MacOS 8 and 9”.
I get why Apple doesn’t care about that, and they really shouldn’t, and macOS is a fine branding choice, but …)
I think the merging iOS and macOS argument carries little weight. The use cases for each are very different and frankly, I don’t see why the CPU is a bottleneck there.
This makes sense for Apple for a few reasons:
First, it frees Apple from Intel’s ability to deliver new generation CPUs. Right now, they can’t do a new Mac with a new CPU until Intel ships that CPU. This would put the CPU factor under Apple’s control.
Second, and related to that, is that it frees Apple from being tied to the performance of the Intel CPUs which, frankly, aren’t impressive over the last 5 years.
Third, it may allow Apple to deliver radically better battery life than they can now for a given level of power. My MacBook Pro 2016 doesn’t have better life than the laptops from them that I had 5 years ago and that’s about 5-8 hours max. If I could get the same power but 10-15 hours? Love that.
As an enduser I don’t think we’ll see too much upside to this aside from the battery life unless the Apple chip team can push chip performance ahead much faster than Intel was able to. But if I can get, say, double the battery life and similar or better performance than the current Intel chips? Sign me up.
I think an ARM processor would be much more interesting as a coprocessor than a replacement, plus we’ve already seen some steps towards this.
There are a lot of things the Mac does that just don’t require much power, and keeping the x86 in an extremely low power state while the ARM handles them makes a lot of sense.
We’ll see.
The Intel schedule complaint seems odd. Aren’t MacBooks on Skylake architecture? Since then, Kaby Lake and Coffee Lake have been released. Cannon Lake was delayed until this year, but there have still been two new microarchitectures that Apple hasn’t even touched.
side note, Gurman’s accuracy record is mixed, hampered by his pro troll position and aggressive eagerness to expose apple. he often gets things wrong.
As a software developer, apple’s most popular pro profession per Craig, my 16GB is more than fine. And it’s not just me:
https://daringfireball.net/linked/2016/11/07/zdziarski-mbp-ram
Sure he could run all that stuff. Because he wasn’t running Chrome! ?
And he wasn’t running a Video editing program with Red RAW video, for example. Sure that’s impressive how Apple compresses memory without any apparent slowdown. Right? But even he admitted there are legitimate reasons to want 32 gigs of RAM.
Kaby Lake I believe.
Glad someone agrees with me on that point. I find Bloomberg to be particularly bad at predicting Apple’s future products and decisions.
The Carbon based life thing is a quote out of context; Cocoa was supported back then, too. It was just that classic apps needed to be updated to be Carbon based. 🙂
Yes, but they had to wait for Skylake mobile chips. The point is that in the past we’ve excused delays in update by noting the dependence on Intel chips schedules (specifically the series that can be used in laptops). This would free them from that entirely.
Kaby Lake brings no performance gains over Skylake aside from video encoding which is niche. I don’t know about Cannon Lake or even if the chips needed for MPBs are available in quantity yet. IF they are, I would think we’d see them at WWDC perhaps.
But that’s the other thing about Intel CPUs… they aren’t bringing significant performance gains with new generations. So aside from the dependency issue, you also have the perception issue of “Apple is using older chips!” but moving to new chips doesn’t actually bring significant, tangible benefits.
No. You just don’t get it. They could do that today if desired. It isn’t desired because the use cases of the two platforms are entirely different.
Do you actively follow Apple, reading what their top software guys and managers explain regularly? I expect a deeper understanding of Apple from the Loop.
Yup.
Edit: And I think the Coffee Lake chip Apple needs came out on April 2nd.