Mac moves to ARM, makes laptops obsolete?

[touchingcloth]

This feels significant. Macs and MacBooks will start using custom ARM chips by the end of this year. I got my first MacBook which had an Intel processor and ran OS X 10.5 (Leopard) and supported PowerPC as well as x86 architectures, before 10.6 (Snow Leopard) came out and only supported x86.

I'm on my second MacBook now (the first lasted six years), which is x86 and running 10.15 (Catalina), and by the looks of things it'll be defunct (or at least unable to take major OS releases) before too long.

The new OS is named Big Sur with the version number 11.0. Even people not familiar with Apples or Macs should find some familiarity this this - the original, garish iMacs with colourful translucent cases were the first hardware to take the OS versioned as 10.0 and branded as OS X, so if you can remember colourful macs like from off of Allie Mcbeal or the phrase "OS X" then that era has now passed.

The only real reason I keep a MacBook rather than a PC or iPad is software like Photoshop. This change means that Adobe will need to - and are - updating their MacOS software to be able to run on a different architecture, which in my case means I'll potentially be able to make the jump from a laptop to a tablet for absolutely everything, so rather than a MacBook I might be able to move to an iPad. C.f. YouTube and other web apps running on Flash before iPhones didn't support the format and most things have moved to things like HTML 5 since, I guess.

[Cloud]

I had a G4 Powerbook back before they switched to Intel.  It'll be fine, I know people love to be as cynical as possible about Apple but they're not really in the business of suddenly making existing hardware useless.  The transition from PowerPC to Intel was remarkably smooth and PowerPC was supported for a reasonable length of time - also worth noting they still have some Intel hardware to release, so they're not going to be ditching it right away.

What I do remember though was first trying an Intel Mac not long after, and realising just how ridiculous the performance difference was.  PowerPC had really fallen behind bigtime.  It'll be interesting to see if similar happens here, as it did age the Powerbook quite quickly from the rapid progress.

Finally seeing a macOS 11 is almost as big a wow moment as seeing them finally do something about the godawful iOS icon dump :). It's played a fair part in driving me to Android 2 or 3 times in its history (I had the original iPhone)

They're definitely not making laptops obsolete.  I think this will be a brilliant leap forward and it's about time.  iPads are great but just not the same as a desktop OS for productivity, even though they're getting nearer.  PC.... yyyyyeah I mean Windows 10 is decent but spend a few weeks with it then come back to macOS and - well to me at least it's a no brainer, macOS is such a joy to use in comparison.  YMMV, I guess.

[earl_sleek]

I can't imagine even a powerful tablet like an iPad Pro being able to cope with loads of tracks and plugins in a DAW, for example. I think that's still years off, and even when it's technically feasible I think the form factor will always be unsuitable for a lot of tasks.

Isn't it only quite recently that ARM has been catching up to x86 in terms of performance? So I wouldn't expect ARM Macs to have a big performance advantage over Intel ones.

[Cloud]

It's not the same ARM chips as in tablets I don't think.  I got the impression they're talking proper desktop class processors, just developed in house.

[touchingcloth]

I can't imagine even a powerful tablet like an iPad Pro being able to cope with loads of tracks and plugins in a DAW, for example. I think that's still years off, and even when it's technically feasible I think the form factor will always be unsuitable for a lot of tasks.

Isn't it only quite recently that ARM has been catching up to x86 in terms of performance? So I wouldn't expect ARM Macs to have a big performance advantage over Intel ones.

It depends on the software and OS I guess, and Apple now have quite a long experience in optimising performance for ARM processors now. In my case I'm on the cusp between desktop and tablet needs - tablets offer better interfaces for photo editing, but desktops have historically had the better of software, so I've found myself in the laptop niche when I'd really rather a tablet, but I think if I were editing videos rather than photos I might feel differently.

"Making laptops obsolete" is probably a bit strong, but it seems like developers who haven't had a userbase with a need for desktop apps will now need to make their desktop apps for ARM chips and in the process make them ready for running on iPads, even if that wasn't there intent. In my case I only ever need the power of a tablet, but some of the software which is only available on desktops, so having that same software able to run on an iPad might make me ditch the MacBook. Though I do my fair share of pirating so maybe not.

[BlodwynPig]

FFS. A big fuck off to bespoke software users.

[buzby]

It's not the same ARM chips as in tablets I don't think.  I got the impression they're talking proper desktop class processors, just developed in house.

Laptop-class processors, not desktop. They are going to be based on the A14 chip family that is being developed for the next iPhone, but with the number of cores increased. The A12Z system-on-a-chip that is used in the current iPad Pro has 4 high performance cores and 4 low-energy cores. The chips being developed for Macs are supposedly going to have 8 high-performance cores and 4 low energy cores.

As a guide on what the performance mighrt be like, the 2020 iPad Pro is roughly equivalent in benchmarks to the current MacBook Air (Core i5, 8GB of RAM) but about half as powerful as a MacBook Pro (Core i9, 32GB of RAM, AMD Radeon Pro 5500M GPU with 8GB GDDR6):


They will probably start moving the lower-spec MacBooks over to being ARM-based first.

[Sebastian Cobb]

I can't imagine even a powerful tablet like an iPad Pro being able to cope with loads of tracks and plugins in a DAW, for example. I think that's still years off, and even when it's technically feasible I think the form factor will always be unsuitable for a lot of tasks.

Isn't it only quite recently that ARM has been catching up to x86 in terms of performance? So I wouldn't expect ARM Macs to have a big performance advantage over Intel ones.

Actually daws are an example of jobs that require lots of parallelism which is something an arm chip with lots of cores (even mobile phones have loads these days) can provide. Cloudfare use arm servers for their caching systems as they can handle lots of simultaneous requests.

What they'll be shooting themselves in the foot with is the tech community, who will miss hyper-v virtualisation used for both running vm's and containers. This might be bad news given windows now can have a Unix-like shell and Linux provides just as good, if not better (thanks to osx's weird old bundled versions of php and python) than osx. I used osx in some jobs, it was fine. I now use a debian machine and the same tools.

People who mostly get apps from the app store or big vendors likely won't notice.

As someone who grew up on acorn machines and likes an underdog to Intel I'd like to see how this pans out. I'd like an arm-based laptop, and there have been ones for Linux but the support has never been quite good enough.

[MojoJojo]

What they'll be shooting themselves in the foot with is the tech community, who will miss hyper-v virtualisation used for both running vm's and containers. This might be bad news given windows now can have a Unix-like shell and Linux provides just as good, if not better (thanks to osx's weird old bundled versions of php and python) than osx. I used osx in some jobs, it was fine. I now use a debian machine and the same tools.

I don't know much about this stuff and am a bit confused. Isn't hyper-v a microsoft product that runs exclusively with a Windows host? ARM have been trying to push for servers for a long time and do have their own virtualisation tech - it's probably not as mature as intel's because they haven't actually had much luck in the server market, but it should be more than sufficient for desktop/laptop use. And containers don't actually need any virtualisation tech, beyond the fact you run linux in a VM (which from what I've heard doesn't work great and you're much better off just running native Linux).

I imagine the goal with this will be fanless laptops with amazing battery life, rather than the performance boost seen with the switch to Intel.

[Sebastian Cobb]

I was using hyper-v as shorthand for hardware virtuslization (Intel call this VT-x AMD call it AMD-v) I imagine arm virtuslization works fine when executing arm instructions, it won't be able to execute x86 instructions natively.

Osx and windows use a light vm when running containers too, but yes on a Linux server running containers on a vm makes little sense unless you've already got a host with guest machines already set up.

[Huxleys Babkins]

I imagine the goal with this will be fanless laptops with amazing battery life, rather than the performance boost seen with the switch to Intel.

Intel's laptop chips have stagnated badly though, both thermally and and speed-wise, so it's inevitable that we'll get a performance boost too.

I'm quite excited to see how these chips will perform in desktop form factors. ARM chips are generally stuck in the smallest, most cramped spaces and almost always passively cooled. But stick one in a tower, up the voltage and give it decent cooling?

[Sebastian Cobb]

Heat has definitely been an issue on laptops the last macbook Pro got some stick as the processor couldn't reach its potential due to self-throttling under load to keep the heat down. Apple weren't the only offender but they're prickly about fan noise.

[Blumf]

An ARM based supercomputer is topping the charts at the moment:

https://www.anandtech.com/show/15869/new-1-supercomputer-fujitsus-fugaku-and-a64fx-take-arm-to-the-top-with-415-petaflops

At the heart of Fugaku is the A64FX, a custom Arm v8-A CPU-based chip optimised for compute. The total configuration uses 158,976 of these 48+4-core cards, running at 2.2 GHz peak performance (48 cores for compute, 4 for assistance). This allows for some substantial Rpeak numbers, such as 537 PetaFLOPs of FP64, the usual TOP500 metric. But A64FX also supports quantized models with lower precision, which is where we get into some fun numbers for Fugaku:

[Zetetic]

FFS. A big fuck off to bespoke software users.

Why?

I guess some of your stuff might have hand-written x86-64 assembly or something, maybe.

[Sebastian Cobb]

Why?

I guess some of your stuff might have hand-written x86-64 assembly or something, maybe.

Developers/security people often choose the macbook pro because it's powerful enough to spin up (sometimes multiple) vm's of windows/linux machines as well as run Unix-like stuff easily.

Wine will be out also, if any Osx users use that. It looks like wine on arm is too run apps made for windows-rt/arm stuff.

[canadagoose]

I feel like I replied to this earlier, but I clearly didn't. Must have dreamt it.

Anyway, I'd heard rumours about this a while ago but it's interesting to see it's starting to pan out. It'll be a bit irritating for people who have just bought an x86-64 Mac, especially if it was one of those several-thousand-dollar ones, but I'm guessing it'll be a bit like the transition to x86 - what was that, about 5 years? - and I think that's already been mentioned in the thread so I won't bother waffling. With regard to emulation and virtualisation, I wonder if it'll become more tenable as Windows for ARM becomes more of a thing? Isn't Windows 10 X (?) an ARM-based thing too - maybe that will drive demand for better performance of x86(-64) programs on ARM systems.

[Claude the Racecar Driving Rockstar Super Sleuth]

It could start an ARMs race.

RISCy business.

[Sebastian Cobb]

I feel like I replied to this earlier, but I clearly didn't. Must have dreamt it.

Anyway, I'd heard rumours about this a while ago but it's interesting to see it's starting to pan out. It'll be a bit irritating for people who have just bought an x86-64 Mac, especially if it was one of those several-thousand-dollar ones, but I'm guessing it'll be a bit like the transition to x86 - what was that, about 5 years? - and I think that's already been mentioned in the thread so I won't bother waffling. With regard to emulation and virtualisation, I wonder if it'll become more tenable as Windows for ARM becomes more of a thing? Isn't Windows 10 X (?) an ARM-based thing too - maybe that will drive demand for better performance of x86(-64) programs on ARM systems.

I don't think Windows ARM has been successful or will be, which is a shame (I actually want arm to displace Intel) because it's in things like the surface pro X. Which costs a grand and is not compatible with most windows software.

We have a basic surface and a surface pro at work. I hate them. My boss loves his pro, and makes sure we all go to the breakout area and watch him set it up and login to everything at every meeting and stand up. These two things might be related.

[canadagoose]

I don't think Windows ARM has been successful or will be, which is a shame (I actually want arm to displace Intel) because it's in things like the surface pro X. Which costs a grand and is not compatible with most windows software.

We have a basic surface and a surface pro at work. I hate them. My boss loves his pro, and makes sure we all go to the breakout area and watch him set it up and login to everything at every meeting and stand up. These two things might be related.

That's a bum. I'd thought Windows on ARM emulated at least x86 32-bit apps, but the performance is probably a bit lacking, like running Windows 2000 on my OnePlus 6T (in a VM, of course). The two-screen ARM devices are still to come, apparently, so there might be a chance yet, but I do remember Windows RT and it didn't go well, so...

[Sebastian Cobb]

There were some android devices that went the other way - Intel atom as the main cpu then an arm co-processor to execute arm instructions for certain apps. Although given android is (or at least - was this pre-dated Cotlin) Java I'm not sure why it needed to given things should've been compiled to bytecode.

It looks like lenovo have some nice looking arm devices, but again, they're competing in price with £1000 thinkpads and the advancement in battery life is going to sacrifice performance. And let's face it, outside of macs thinkpads are one of the few computers that can be seen a bit like a used car and still have a good bit of life in them after a couple of years, even if I was buying new, I wouldn't trust Microsoft not to shelve things if it wasn't popular.

[MojoJojo]

I thought Windows on ARM was long dead, because most Windows software isn't compatible and without that Windows is a bit pointless. I don't hate Windows, but I only have it around to work with software and devices that wouldn't work on linux.

I suppose one point to note is unlike the PowerPC->Intel change Apple are actually designing and getting the ARM chips fabbed, which means they're going to be keeping significantly more of the shelf price which is probably a bigger motivator than any performance/battery life gains.

Then again maybe the volumes aren't bit enough to really make much of a saving.

[Zetetic]

Developers/security people often choose the macbook pro because it's powerful enough to spin up (sometimes multiple) vm's of windows/linux machines as well as run Unix-like stuff easily.

But I'm not sure how any of this stuff affects BlodwynPig - if it's bespoke software (and you've fucked up and not obtained a source license), surely they can provide a macOS/ARM version in most cases?

Wine will be out also, if any Osx users use that. It looks like wine on arm is too run apps made for windows-rt/arm stuff.

Interesting if Rosetta 2 can be leveraged for things like this, albeit with a substantial performance penalty.

[Sebastian Cobb]

I suspect, as apple say it'll work fine with things that go through X-code, bringing things even more into their walled garden and distribution channels. Anything outside of those is merely tolerated.

I always thought something like the chromebook would've been an ideal arm vehicle. Given the ecosystem there was no real reason to need x86 compatibility like windows does and macs might.

[buzby]

But I'm not sure how any of this stuff affects BlodwynPig - if it's bespoke software (and you've fucked up and not obtained a source license), surely they can provide a macOS/ARM version in most cases?
Interesting if Rosetta 2 can be leveraged for things like this, albeit with a substantial performance penalty.

The announcement said that they are going to use Rosetta 2 to emulate an x86 environment to allow existing  applications to be run on the new ARM-based Macs. They said all developers will need to do is recompile their existing applications using the new version of Xcode (which I'm sure is easy to say, but there's bound to be problems), and that during the transition they should expect to deliver packages containing both Intel and ARM binaries.

[Zetetic]

I specifically wonder how easy it'll be to bend Rosetta 2's binary translation to support something like Wine (and an alien binary format).

[Sebastian Cobb]

The announcement said that they are going to use Rosetta 2 to emulate an x86 environment to allow existing  applications to be run on the new ARM-based Macs. They said all developers will need to do is recompile their existing applications using the new version of Xcode (which I'm sure is easy to say, but there's bound to be problems), and that during the transition they should expect to deliver packages containing both Intel and ARM binaries.

In theory all you need to do to make a pro*c application move from solaris sparc to x86 solaris is recompile it. And most of the time that's true. The times it aren't, are not fun, sometimes it's not even your fault.

I don't see how they're going to be able to magic converting x86 instructions into arm instructions in a way the thre rest of the *nix haven't thought of.

[Blumf]

But I'm not sure how any of this stuff affects BlodwynPig - if it's bespoke software (and you've fucked up and not obtained a source license), surely they can provide a macOS/ARM version in most cases?

A lot of business software is based on old proprietary ways, because it dates back decades. So no, you're not going to get the source (and you wouldn't know what to do with it even if you did, as your business isn't software development, it's a 2 person company dealing with other complex stuff)

Windows RT (their original ARM attempt) failed, in part, because MS tried to lock it all down with an app store ($$$ in their eyes staring at Apple and Google's). That model only works for software devs if you're selling lots of copies  that don't change much. For complex software that is specialist, and thus only has a few hundred customers, if that, it's a non-starter. Lots of costs, and MS skimming the profits. Plus, with WinCE/etc., a proven track record of screwing over partners and devs. Not sure what their current ARM attempt means, but history says to avoid.

[Sebastian Cobb]

Why what you've said is true blumpf but far more true on windows tbf. When I left a job in 2017 some of our old software products used excel reports written in 1997 with visual basic code that used an oracle client to get data from databases directly, shove it around in hidden worksheets behind the scenes. Not long before that, I had to reingineer part of one of them to use a different SOAP library when making Web requests because the office 2003 to one's finally stopped working after 10 years of ignoring deprecation warnings.

The chances are none of that shit would've worked on Osx office ever. And it would have never have mattered because apple never made any inroads to enterprise in departments that had to adhere to software policy.

[ProvanFan]

Has it got Snake II though?

[Blumf]

And it would have never have mattered because apple never made any inroads to enterprise in departments that had to adhere to software policy.

This is to Apple's benefit. Most Mac users just have straight forward tools; Photoshop, Illustrator, etc. Software that works within itself, so that, come a big architectural change, it's pretty easy to cope. As such, the switch from PowerPC was awkward, but not impossibly painful, and I expect for the traditional Mac user, the switch to ARM will be the same.

Enterprise software, on the other hand, is deeply and complexly tied together across many systems. Ugly legacy messes that are set in stone[nb]As a side note, open source isn't the answer here. See the Python and Perl communities dragging their feet. See how long it's taking to replace X Windows. This is a complexity problem, not a proprietary vs FOSS issue.[/nb]. The problem here is that, since Mac's switch to Intel, some people have made the mistake of bringing that enterprise mess onto their shiny Apple laptops, not a good idea.