Why Apple will move away from Intel and towards ARM 64 bit | Hacker News Reader

Why Apple will move away from Intel and towards ARM 64 bit

semiaccurate.com

62 pointsmtgx69 comments

I don't necessarily disagree with the conclusions, but

"Next up is the OS. Have you noticed that OSX releases of late, well, to be blunt, suck? It’s not that they suck as a stand alone OS, but they take away a lot of the freedoms and flexibility that the Mac desktop OS user has come to expect. Bit by bit Apple is removing all of the parts that make OSX something other than a phone class OS"

Is nonsense. "Suck" is a matter of opinion, but there's nothing you can do on a 5 year old Mac that you can't do on a new Mac (and several things more). People keep confusing attempts to make the OS easier to use with "closing" it. When I can't run my own software, and drop to a terminal with the standard UNIX tool chain, I'll but the argument. But there's no sign of this so far.

> When I can't run my own software, and drop to a terminal with the standard UNIX tool chain, I'll but the argument. But there's no sign of this so far.

You can run your own software and drop to a terminal with the standard UNIX tool chain on your iPhone, too. You just have to jailbreak it first.

So Apple is really one flip of the switch away ...

That's true of every computational device in the world. Windows, Ubuntu, Android, etc. Everyone could flip that switch, and lockout non-jailbreakers (or Kernel recompilers). It's not an argument for it happening.

>So Apple is really one flip of the switch away ...

And how is that different from OS X 10.1?

> So Apple is really one flip of the switch away ...

... in much the same way that Ubuntu is really one flip of the switch away.

Exactly, it's not like Apple didn't reach out to the unix community last year and work on putting together a slick command line tools package to make projects like homebrew even easier to use and support. If you're the sort of user who compiles a lot of software in the command line it's pretty easy to forget that launchpad even exists. And so what if it does?

If you can tell me how to set up Lion to use my preferred Snow Leopard hot corner settings, I'll be forever grateful:

* Top left corner: Show all four spaces in a 2x2 grid

* Top right corner: Show all current application windows

* Bottom right corner: Show desktop

* Bottom left corner: Show all windows

I've been using this setup (minus the Spaces part) since 10.3, and have yet to find a workflow I prefer for managing windows on a laptop screen's limited space when using a laptop with no external displays hooked up. I'd prefer to also have Leopard-and-before style proportionally-sized Expose to SL's grid, but that's a minor thing.

Lion's Full Screen Mode and Mission Control appear to have made things more "accessible" (and oh, are people often surprised by my setup when borrowing my laptop for a bit and inadvertently hitting a corner!) but I can't find any way to replicate what I had before. As far as I can tell, the options are just gone.

And I find the removal of features I'd come to love really disturbing, especially coupled with the move away from hardware upgradability/expandability.

-----

As far as moving the Mac to ARM goes... that's in all likelihood the day I buy another ThinkPad. And for pretty much exactly the same reasons as the first time, when my PowerPC-based PowerBook G4 just didn't have enough power, and couldn't virtualize x86 OSes for testing/dabbling/whatever.

Fair point, they did rather neuter spaces (I guess to get more people using it, but at the expense of power users). All other options still work. I'm not at my machine to lookup settings, but ML added a "Don't group windows" to Mission Control that pretty much replicates Expose from SL (just ignore the Spaces view at the top).

That's good to know about ML. I've been happy staying on SL for my personal laptop so far, but am glad there's an upgrade path that'll still give me an all-windows Expose functionality.

Spaces was always the least-used of my hotcorners -- especially after getting a separate work laptop that lives with two external monitors hooked up, plus the internal screen open, which greatly reduces the need for that kind of stuff. It's mostly made me not bugged by it being on Lion.

You should check out Total Spaces: http://totalspaces.binaryage.com/

I was dreading the move to Lion since Spaces was a big part of my development workflow. But Total Spaces has filled in the gap for me.

Since its just juggling windows around there is a slight lag in responsiveness compared to the original spaces but in general it does the trick for me.

You can do most things on your list with Total Spaces:

* Top left corner: Show all four spaces in a 2x2 grid * Top right corner: Show all current application windows * Bottom right corner: Show desktop * Bottom left corner: Show all windows

>If you can tell me how to set up Lion to use my preferred Snow Leopard hot corner settings, I'll be forever grateful

This is not the kind of the locking and freedom removal the article is talking about.

This is a feature that's changed, removed, as they experiment to find the best (according to their team) way to implement it. I, for one, never use Hot Corners (find them disturbing), and I also hate "autohide dock". So it's not like it was a feature with unanimous love that got removed (something like Maps on iOS or Rosetta that people either needed or didn't care about).

Also, I don't see how those changes in the software level has anything to do with the ISA its targeting. Same thing with the upgradable components. Just because you are using an ARM core doesn't mean you can't support SATA and DDR3 buses.

Nor would we reasonably expect those avenues to be closed off to developers even if, from a consumer standpoint, the device does become 'closed off' by guiding consumers through the app store to get their software. Developers need under-the-hood access, and Apple has always understood that devs, at least, need freedom.

The whole idea that a laptop-like device has a consumer and developer mode is what's disturbing to many. Unless the developer version is free and open.

I don't think they'd bother with different modes. It's just that, like it is now, if you're a dev you can install iterm2 and xcode and lift the hood and bash around, but if you're a consumer, you not only don't have to ever open the terminal, you don't even need to be aware it exists.

This is why the hyperventilating about closing the OS is so overwrought: It doesn't need to be closed to serve Apple's purpose of having a walled garden, it just needs to be massively more convenient to stay in the walls of the garden if you're a consumer. Devs will find the gate and wander out on their own.

I wish they would bring back keychain syncing through iCloud.

I get the feeling there was some weird support or security issue, because its removal is odd given the rest of their iCloud moves. I wish it was back too.

I often think Apple should have bought a source code control company or a document versioning company to get the expertise for making iCloud sing.

Oh, yes; that would be awesome. It would be absolutely amazing if it would sync everything (including website credentials) between iOS and Mac also (I wasn't a MobileMe user, so I don't know if it synced data only between Macs or between all devices).

I've been struggling with a corrupt file system for the past couple days (some inodes are currpot, and anytime you want to access them, either through Finder or Terminal or Time Machine or even SuperDuper, the system crashes. Nasty, nasty bug) and there are basically two things that are keeping me from just erasing the whole fucked-up volume and install ML anew: I'm not sure if I mive ApplicationSupport/MobileSync folder to the new machine, it would still sync my iPhone and iPad, and that I can't export the entire keychain (I have to do it bits by bits, and most items are not exportable).

I like OS X 10.8 a lot better than say 10.5. Full-screen spaces is a killer feature for me.

At a developer level, OS X is now more accessible than it ever has been. Do you remember OS X 10.0, back when you could program the system in any language as long as it was Objective-C? Today you have multiple languages that are ABI-compatible with the Objective-C runtime. Isn't that a huge win? You have POSIX support that's better than it has ever been. You finally have thread-local variables. You've got a great toolchain in LLVM/LLDB/Clang that Apple has been bankrolling.

Mac is the development platform for iOS and Apple knows it needs to keep developers happy and the iOS apps flowing. It has nothing to gain from neutering Mac.

I'm sure that Apple has a plan in place to do this, but whether they pull the trigger on this plan is another question. It's quite likely that Clover Trail is faster than A6, and Clover Trail is Intel's low end chip. Ivy Bridge is almost an order of magnitude faster than Clover Trail. Most of the PA Semi guys at Apple will be working on chips for the next iPads and iPhones, but I'd be willing to bet lots of money there's a skunk works project to design a chip that could go into the Air. Apple loves squeezing the margin out of their suppliers, and Intel's margin is very juicy.

Unlikely, especially on the GPU side. But let's say Clover Trail is slightly faster than A6. That's not really relevant, because Apple would care much more about using their own chip. It's not like Apple hasn't used slower CPU's than the competition. Heck, for the entirety of the iPhone 4S' and iPad 3's life, both had slower CPU's than the competition. And they didn't care.

Plus, Clover Trail is not used in smartphones for a reason - it uses too much power. That's why they are only using it for $800 Windows 8 tablets with big batteries. So for this chip, the A6X, or whatever they call the chip inside the next iPad, will be its competition anyway.

As for "ARMbooks", I'm pretty sure Apple will wait until they have an 64 bit ARM chip first, but I also wonder if they will make a "higher-end" chip, that uses say 5-10W of power and is specifically made for those ARMbooks - not for iPads or iPhones, while in parallel continuing the development of A7, A8, etc.

Being able to run Windows applications with Fusion or Bootcamp was one of the big draws for me for the Mac. Not sure if I'm ready to give up running x86 applications yet (or ever?).

A valid point. However, Microsoft is betting heavily on ARM too, with Win8 RT (if I'm not mistaken and RT is the ARM one). So, Win8 and a lot of Win8 apps are going to support ARM, and those who don't can be virtualized!

(Note that I'm not saying it's a good plan)

Wait, you're suggesting that you're going to be running an x86 emulator on winrt for apps access? That is pretty unlikely to be a realistic solution unless you're only looking for very old apps - try running an arm emulator on x86 to get a taste, and then slow it down by 5x or more.

You're right :) It probably would be unusably slow.

Interesting read with some good points albiet biased towards the aspect they will move. But does show that it is not impossible, though nothing soon and I'd agree that laptops would be there first target area. I also liked the aspect that OSX is becomming more and more phone OS like and that could indeed also be said of the upcomming win8 due to metro. Overall I see both Apple and Microsoft going for consumer level devices, remember calculators are computers as well, only biased for the application and in that both apple and microsoft are transitioning towards a unified interface and standard making computers more and more consumerised. Thats ok as there will always be Linux and many other operating systems to play with if you need that level of interaction and in that whilst its a change I don't like, I can understand it.

One aspect not touched upon in this article is the whole area of a server, one which apple walked away from and is a whole empty cavity that could become one which a 64bit version of ARM chips down the line could end up making that area viable for Apple again. Though it would probably be a solid white cube called iPortal j/k.

Either way, even if they do move they can still brow beat Intel for even better prices milking every avenue of options along the way. Though any move to a ARM base for OSX would at the earliest be 2 years away and relisticaly 4-5. Though for servers and processing farms it does become a whole new avenue.

"Overall I see both Apple and Microsoft going for consumer level devices, remember calculators are computers as well, only biased for the application and in that both apple and microsoft are transitioning towards a unified interface and standard making computers more and more consumerised. Thats ok as there will always be Linux and many other operating systems to play with if you need that level of interaction and in that whilst its a change I don't like, I can understand it."

A huge chunk of Microsoft revenue comes from business customers, pros and the ilk, who run applications that can't all be distilled down to a touch interface. There's still a CLI in Windows 8 and much of the MS new server products are driven with CLI in the form of Powershell. You can pick your level of interaction that you need.

It seems like Apple would be a pretty good candidate for the "Heterogeneous System Architecture" work AMD has been doing, x86_64 and ARM on the same SOC. Leave the x86 cores off when you don't need them.

If anyone has both the technical know how to come up with a "good enough" development scheme to make that work, and the leverage with their developers to ensure support, it's Apple. But yikes that sounds like a hard problem. How do you migrate a process from one ISA to the other? Some kind of compiler-enforced "safe points" where there's a sensible mapping of the IP and other instruction stream-sensitive state?

Lots of systems have been buils with multiple processor architectures, including the early Macs, the Nintendo DS, the PS3, and every PC the last decade (cpu+gpu).

You would usually write your code to use one specific cpu, rather than trying to balance the load automatically like on a heterogenous system.

I thought trotsky was suggesting an ARM/x86 mix along the lines of ARM's big.LITTLE, which allows process migration between two very differently implemented (though ISA-compatible) cores.

While they haven't announced any of it in public, it does seem hard to believe they'd be migrating running applications. It's much more likely that they're targeting nx page sharing, you'd still have build the os for both isas, but it'd allow you to run many tasks on the low poer cores and only spin up the hot ones when the application required the performance or binary compatibility. Think more like switchable graphics than big.little, though the low power architecture would stay usable all the time.

Bootcamp sells a lot of Macintoshes.

OS X sells more.

Maybe, but the ability to flip to Windows is a huge advantage that makes buyers feel safe about the leap. I cannot see them giving up without good reason or an alternative (WinRT).

I was at a talk recently about some C#/.Net framework aimed at .Net developers and half the laptops in the room where Macs running Windows. Even the lead developer giving the presentation had a Mac.

Apple are arrogant enough to believe that cutting bootcamp won't significantly affect sales.

I'm pretty sure they have a rough estimate of the number of copies of VMWare and Parallels sold, plus know the number of bootcamp downloads. They are a very profitable business and do things for business reasons.

Such as declaring thermonuclear war on android, that's a business decision which was made for business reasons, and has been great for business.

Seems pretty silly to claim it's "obvious" as this guy does repeatedly. I'm not sure you can claim that the move from being able to upgrade and service computers to not being able to is a move toward being an iOS device. Seems more likely that Apple just makes those sacrifices because they're required to make thin and light products, and at today's speeds few people need to upgrade things anymore. (Not that it wouldn't be _nice_ to be able to upgrade things, but we've seemingly moved to a time when a lot of people don't mind the lack of this as a feature)

Yay, great. Apple go. The day Apple moves to ARM on the Macbook will be the start of the downfall of Apple.

Also, when is Apple moving to Intel on the iPhone?

Actually, i can imagine running a server farm on ARMs, that'd be great (and there is ARM server hardware available). BUt for the desktop? nope.

Why not? They've done ppc to x86 before.

And what is now called OSX started on 68K processors

ARM started as a desktop platform and, at the time, it was much faster clock for clock than what Intel offered. I believe you can find benchmarks of that time comparing the Archimedes to high end PCs.

Current ARMs are designed to save power, but there is nothing preventing Apple (or anyone else) from designing one for speed.

What makes you think Apple would have more luck competing with Intel on speed this time around than they+IBM/Moto did in the past, or than AMD has had?

Well, it depends on what you mean by 'competing with Intel on speed'. If you mean having higher-performance CPUs than the Intel equivalents then Apple probably won't achieve that. If you mean having CPUs close enough to Intel's in overall performance that few potential Apple buyers will be put off by speed issues, then that seems very possible. If Apple had to use AMD processors from now on they'd manage fine (potential supply/vendor relations problems aside).

It's not that IBM and Motorola couldn't make PPC faster than Intel Core. It's that Apple wasn't a client with volume enough to justify the effort. Instead, IBM focused on making Cell (which is high-volume, so-so margin) and POWER (low volume, very high margin) while Motorola focused (IIRC) on the embedded market.

BTW, Apple is no longer focusing on pure performance machines. The MacPro family is getting much less attention than it once did. Once your crowd thinks your machine is good enough, there is little point in making it even faster.

haha. Yeah, because a benchmark from the 80s means something today.

So you honestly think Apple will forsake the x86 ecosystem, all the apps, maybe put another emulator approach to it AND lose the benefit of running windows on a macbook? For what? What's the benefit? It's just fucking over customers. Well, as always. So maybe, they'll just do it. Opportunity to take even more control of the users, yay.

Out of curiosity, how much code nowadays is tied to an architecture? Operating systems, sure (but even then not much outside the kernel). Most code I write is architecture-agnostic... Yes, in C++.

You are kidding right?

Actually, as a linux and open source user i am probably the next likely person to switch to ARM on the laptop. But i can cleary see that this would be next to impossible for 90% of "normal" users.

Homework for you: Pick the next best non-geek person (secretary job is good, or some family member) and go through all the desktop software they have. How much of it is not available on ARM? How much of it is available on Windows, MacOSX AND Linux. Because... multiplatorm is easy, right? Everyone has it. No, no, no. Unfortunately that's faaaaaaar away. :(

I like to consider my mom very normal and I'm very sure everything she uses is just a compile away from running on ARM. Ubuntu, Firefox and LibreOffice are usually sufficient for 99.9% of what PC users do.

Even those who are doomed to use Exchange can rely on OWA, which runs just fine on any decent browser (and at least one very crappy browser)

Yourline of thought is the same that i had for the last decade with regards to Linux. It's still not the year of Linux on the Desktop ;)

And a big part of why is: There is no MS Office, no Photoshop and no Games. Same will apply for ARM/MacOSX.

Apple has done it on the Mac now three times with ISAs (68K to PPC to Intel, with multi-ISA binaries) and once with operating system architecture (MacOS classic to OSX). NeXT has ported their OS from the original 68K implementation to x86, HP-PA, POWER and SPARC before making it run on Macs. Their IDE, the predecessor of Xcode, also ran on Windows NT.

Apple has also shifted its strategy from "we build really fast computers" (PPC, MacPros) to "we build tools you love" (current Mac line). I see not big problem to gradually moving again to a new ISA, in special if it promises very long battery lives and even lighter machines.

I'm not saying they can't do it. And not saying they never did it. Does it mean that they should constantly switch the architectures? Why?! It's just some stupid "ARM is cool, let's do ARM, it's different" sentiment more then for technical reasons.

Again: If ARM would build a speed equivalent (in major important benchmarks) processor, why do you think that the battery life would increase significantly? Your battery life would _not_ double or triple or whatever. You would maybe get.. i don't know.. 30% more? And why would the machine be lighter?!

So again, where is the real benefit here. Keep in mind that you would likely lose Windows support (that'd even make sense in closing up the Apple ecosystem even more!) and most legacy app (except Apple does another crappy emulator support multibinary thing).

Anyway. As i said in the original post: I hope Apple does this. I hope Apple abandons Windows/Bootcamp and goes further along the road of the Apple ecosystem. If they do, Apple will fall. History repeated :)

What'd be wrong with an ARM-based Air? I don't really do anything on my Macbook that I need all that CPU horsepower for. I'd trade it in a second for 10+ hour battery life.

Sooo... you say you could actually just use an iPad? ;)

Also, i guess you are really overestimating the role of CPU power usage in a modern PC. You wouldn't like double the time, just by swapping Intel with ARM, not nearly.

They're moving away from x86 - maybe. I remember Intel's XScale ARM processors were the best around at the time. But Apple without x86 = suicide...

Anyone who says something like this deserves to be laughed at and shown the door out of technology.

Intel is light years ahead of ARM for desktop CPUs, where power usage is irrelevant.

Do you really think Apple is going to run ALL THE APPS EVER WRITTEN FOR OS X on an x86 to ARM emulation layer?

Really?

Well, then I have a bridge to sell you.

They've done it before, more than once. http://en.wikipedia.org/wiki/Mac_68K_emulator and http://en.wikipedia.org/wiki/Rosetta_%28software%29

Apple already dropped Rosetta, abandoning PPC apps. Gatekeeper is doing away with apps compiled before 2012. 10.9 may drop 32-bit. Perhaps 10.10 will be App Store only with fat x86-64/ARM apps. Then 10.11 can drop x86-64.

Maybe.

Based on the current revenue breakdown I think the only reason Apple still makes computers is the margins they get. If they got the margins a lot of other pc makers get they would have killed the computer side already.

Nobody here's a wee bit worried about the direction Apple is going with Gatekeeper? Only signed and sandboxed apps in a few years? Only to be dl and purchased in the App Store with Apple receiving 30%?

As long as the user can disable it, it's not a problem. Android does the exact same thing.

I disagree. They will either hang themselves with x86-64 or make it a success as it is now. They won't go near ARM as a mainstream CPU.

Will Windows 8 be able to run on that chipset as well?

One thing about ARM is generally ARM chips need SoC specific drivers. If Windows 8 has drivers for this chip, Windows 8 will run. Chances are, if Apple does this and wishes for Boot Camp to continue, they will package the drivers for Windows 8. If they wish to discontinue Boot Camp, they won't.

Not going to happen, Windows RT will ship only with devices just like Windows phone, there will be no standalone software that you can buy and install.

You can't license WinRT unless you're selling the hardware.

It reminds me of this: http://apc.io/

At some point, we have enough power in a small space that the hardware powering your computer or laptop is basically the size of a credit card and it doesn't matter if it's ARM or Intel.

Apple can save money and get better battery life by making their own chips (as they do in phones now) so why not move that up the hardware food chain once the performance is "good enough".

Apple is about selling mobile devices to consumers and probably the #1 thing the average buyer wants is longer battery life, even at the expense of performance.

Also, Apple knows the power of controlling their own destiny and inevitably when they partner with a company, the same company tends to turn around and resell the innovation as their own products - Microsoft Windows, Google Android, and Intel Ultrabooks. Apple is big enough now they might not need to partner for such innovations much longer.

That could lead to an even bigger and more sustainable competitive advantage.

Reply on news.ycombinator.com