ARMageddon

Devil's Advocate · Apr 2, 2015

Benchmarks of the new 14nm Broadwell MacBook are appearing, and it seems the base model's 1.1 GHz Core M is on par in CPU performance with 20nm 1.5GHz A8X in the iPad Air 2

Geekbench
MacBook, 1 thread: ~1900
MacBook, 2 threads: ~4000
A8X, 1 thread: ~1800
A8X, 3 threads: ~4600

Considering Apple probably pays ~$200-300 for the Core M, while it's estimated an A-series SoC (with RAM!) costs Apple $20-40, it seems inevitable Apple will offer an ARM-based MacBook soon with the advent of an even more efficient 14 nm A9 this year. It may even be under $1000.

What will it be called?

DarwinKS · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28782025#p28782025:2yvj7yeu said:
Devil's Advocate[/url]":2yvj7yeu]
What will it be called?

MacBook RT.

GoodOnYaMate · Apr 2, 2015

It'll run iOS X.

Smack · Apr 2, 2015

How do the GPUs compare?

Hap · Apr 2, 2015

I really don't understand this fascination with switching Macs to ARM. None of the OS X software out there now will run on the platform, developers will have to support it (which I see as a very up hill struggle), and emulators are not feasible given the comparability in performance. Most users would loose so much in productivity that I don't see it going over well. There are a lot of iOS ARM apps out there, but I think they would also need a lot of changes to work on a non-touch device.

cateye · Apr 2, 2015

I fail to see the business case for Apple to put the Mac through yet another architecture change when Intel is competently churning the reference platform just fine. Why bother? iOS is all Apple is really interested in, they completely controlled that platform from the outset, and are free to steer it as they see fit for maximum effect. The Mac is a nice side business at this point, stable and mature. Iterating the industrial design here and there occasionally to keep buyers engaged is all Apple needs to do. Intel will do the heavy lifting of the internals for them, as they do for all PC OEMs, thus preventing Apple from getting caught in the same sort of meaningless benchmark bakeoffs that constantly sunk the PowerPC. The real cost-savings for Apple is not having to worry about the Mac architecture at all.

May as well hope for an xMac while you're at it. It's more likely.

MKG · Apr 2, 2015

I think we are a good way away from that type of convergence, but the A8X and A9 chips are shaping up to be formidable processing engines. The specs on the A8X and price point make it a fantastic value and possible desktop compute candidate. Do I think (or want) a mess OS hybrid? Absolutely not.

armwt · Apr 2, 2015

My *gut* reaction was to dismiss this, but the numbers make for some interesting thought. The Intel Core M-5Y31 cpu is priced at $281 when sold by the tray, according to Google... let's say Apple is getting them for $250 just for nice round numbers. If they ARM CPU is $40, that means they could save $210 per laptop in hardware costs by switching... drop price by $100, and they're netting a $110 extra profit per machine (which is better than a 10% increase at these price points).

I still don't think it will happen, due to issues that Hap and cateye listed above, but it does make for some interesting thought.

(also makes for some interesting thought about the "iPad Pro"... again, personally I don't think it will happen, but you could build a A9-powered convertible iOS tablet that would give the MB a serious run for its $$ in pure performance - might be a heavy traveler's dream machine.

Smack · Apr 2, 2015

Given the huge negatives of an architecture switch, I don't think $100 price cut would appease anyone.

Devil's Advocate · Apr 2, 2015

There's no reason the price cut couldn't be more significant because Apple is essentially paying an Intel tax. Remember, the A-series chip goes into an order of magnitude more products (iPhones and iPads) that allows for greater economies of scale.

Simple math:
If Apple has 40% margins and pays $280 for the Core M in the $1300 Macbook, then it costs Apple around $610 to make a MacBook. Dropping it's price to $899 with a $40 A9X, Apple would still have 38% margins.

Is having an Intel Core M over an A9X really worth $400 or 44% more money for the average buyer? There's no objective differentiation to the end-user what architecture their computer uses.

It won't take very long for major apps to have ARM binaries, since a lot of software development already takes place on ARM... in fact, more ARM processors are sold than x86 processors worldwide. x86 is past the point of being doomed.

armwt · Apr 2, 2015

Is having an Intel Core M over an A9X really worth $400 or 44% more money for the average buyer? There's no objective differentiation to the end-user what architecture their computer uses.

So long as the software is there to support it. Got to keep that in mind. I suspect Apple could have OS X running on ARM very quickly, if they don't already.

But as of right now, today, not a single app in the Mac App store would run on that machine, at least not without some sort of "Rosetta", and THAT is why it matters to the end-user... *today*.

cateye · Apr 2, 2015

In what universe would Apple ever pass cost savings on to its customers? The ARMBook would retail for the same as an IntelBook, run no legacy software, for no material gain to consumers or Apple.

sdh · Apr 2, 2015

We have no idea what Apple is paying Intel for the Core M chips. It could be $281, it could be a small discount off that, or it could be half price.

Apple probably loves having their A-series chips to be a threat to Intel to keep their pricing in line (unfortunately AMD is mostly toothless these days). Intel is smart enough to know there are benefits to the x86 architecture but not at five times the price.

cateye · Apr 2, 2015

Apple probably loves having their A-series chips to be a threat to Intel to keep their pricing in line (unfortunately AMD is mostly toothless these days). Intel is smart enough to know there are benefits to the x86 architecture but not at five times the price.

Thank you for this. Apple's AX chip prowess is far more useful as a negotiation cudgel against other partners than it is as an awkward "one size fits all" solution for all their products.

Also, what's Apple's motivation to have to bifurcate the Mac experience into Intel-based ProMacs (iMac/MacPro) where performance and legacy compatibility are critical, and consumer-level Macs running ARM? Now there are two separate versions of OS X, two different lines of Macs, and limited compatibility between them.

Unless you want to argue that Apple will eventually transition out of high-end computing entirely (which is not an entirely crazy idea, given Apple's razor-sharp consumer focus and success), this starts to make even less sense.

Devil's Advocate · Apr 2, 2015

"OS X" aka the Darwin Mach kernel already runs on ARM -- in iPhones, iPads and Apple Watches. All of Apple's development tools have an ARM target. If an app is written in a non-architecture specific language, it should be a simple recompile. Apple could require all Mac App Store apps have to be universal binaries past a certain date (e.g. 6 months after launch).

With x86 gone, Apple can focus all of its software development, driver and compiler peoplepower on ARM whereas right now there's probably two separate teams within Apple.

Devil's Advocate · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28784155#p28784155:2juid51g said:
cateye[/url]":2juid51g]

Apple probably loves having their A-series chips to be a threat to Intel to keep their pricing in line (unfortunately AMD is mostly toothless these days). Intel is smart enough to know there are benefits to the x86 architecture but not at five times the price.

Click to expand...

Thank you for this. Apple's AX chip prowess is far more useful as a negotiation cudgel against other partners than it is as an awkward "one size fits all" solution for all their products.

Also, what's Apple's motivation to have to bifurcate the Mac experience into Intel-based ProMacs (iMac/MacPro) where performance and legacy compatibility are critical, and consumer-level Macs running ARM? Now there are two separate versions of OS X, two different lines of Macs, and limited compatibility between them.

Unless you want to argue that Apple will eventually transition out of high-end computing entirely (which is not an entirely crazy idea, given Apple's razor-sharp consumer focus and success), this starts to make even less sense.

The idea is x86 would be gone. Dead. There's no reason why Apple couldn't simply develop higher-end ARM chips or higher clocked versions. They could also simply throw more cores at the problem, i.e. a 128 or 256-core ARM Mac Pro.

The Intel tax someone pays on a Mac Pro dwarfs the cost of every other component of the system, especially on the top-of-the-line models.

cateye · Apr 2, 2015

I'm not a developer so I'm not qualified to comment on how realistic an assertion that is. However, I'm always highly suspicious of the claim of a "simple recompile" being the solution to architecture changes. I'd love to hear from actual developers on this thought. It'd be especially cool to hear from Schwieb from the APEX Group at Microsoft, as he's stated in the past that there's significant shared code between the iOS and upcoming OS X versions of Office. It would be nice to know how much of the additional work they have to do is related to OS X vs iOS, or Intel vs ARM.

cateye · Apr 2, 2015

The idea is x86 would be gone. Dead. There's no reason why Apple couldn't simply develop higher-end ARM chips or higher clocked versions. They could also simply throw more cores at the problem, i.e. a 128 or 256-core ARM Mac Pro.

Kind like how they threw GPU cores at the MacPro as a way to differentiate it from more aggressively-priced and value-based Wintel equivalents? How'd that work out for them? You should read the MacPro thread here to find out how much of a disappointment it's been to pretty much everyone except a narrow band of of folks using software that has been specifically re-written to take advantage of OpenCL. You're asking the development community to target desktop ARM similarly for Apple alone. Won't happen.

The Intel tax someone pays on a Mac Pro dwarfs the cost of every other component of the system, especially on the top-of-the-line models.

Apple would never pass cost savings on to consumers. They're not Dell (does that saw still work anymore, now that Dell is moving upscale with products like the XPS13? They're not... Acer? Lenovo? Hmm. Doesn't have the same ring).

Mr Beardsley · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28784341#p28784341:3d9cio70 said:
cateye[/url]":3d9cio70]I'm not a developer so I'm not qualified to comment on how realistic an assertion that is. However, I'm always highly suspicious of the claim of a "simple recompile" being the solution to architecture changes. I'd love to hear from actual developers on this thought. It'd be especially cool to hear from Schwieb from the APEX Group at Microsoft, as he's stated in the past that there's significant shared code between the iOS and upcoming versions of Office. It would be nice to know how much of the additional work they have to do is related to OS X vs iOS, or Intel vs ARM.

The iOS simulator is proof of the ability to easily move code between architectures. When an app is run in the simulator, the code is compiled as x86. When the app is run on a device, the code is compiled as ARM. It seems that the simulator works very well, and I have not heard of much in the way of trouble between actual devices and the simulator. Also of note, the endian issues that existed between PPC and x86 are not present with x86 and ARM. ARM can be either big or little endian, although it defaults to little. x86 is a little endian architecture. That removes one of the major issues that was present in the PPC to x86 transition. I would guess the only major issue is for code written to target SSE. This would need to be changed for the ARM NEON. I would guess this impacts very few applications. Hopefully at this point most major applications are not using inline assembly anymore.

cateye · Apr 2, 2015

But this is for software that begins life written for iOS. This may be a stupid question, so bear with me, but is it accurate to assume the reverse would necessarily be true? Is SSE optimization the only sticking point that differentiates code written assuming OS X on Intel? I'd be curious to know more about what would constitute a "non-architecture specific language" as DA indicated a few posts earlier.

Devil's Advocate · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28784429#p28784429:2awte3w2 said:
cateye[/url]":2awte3w2]

The idea is x86 would be gone. Dead. There's no reason why Apple couldn't simply develop higher-end ARM chips or higher clocked versions. They could also simply throw more cores at the problem, i.e. a 128 or 256-core ARM Mac Pro.

Click to expand...

Kind like how they threw GPU cores at the MacPro as a way to differentiate it from more aggressively-priced and value-based Wintel equivalents? How'd that work out for them? You should read the MacPro thread here to find out how much of a disappointment it's been to pretty much everyone except a narrow band of of folks using software that has been specifically re-written to take advantage of OpenCL. You're asking the development community to target desktop ARM similarly for Apple alone. Won't happen.

The Intel tax someone pays on a Mac Pro dwarfs the cost of every other component of the system, especially on the top-of-the-line models.

Click to expand...

Apple would never pass cost savings on to consumers. They're not Dell (does that saw still work anymore, now that Dell is moving upscale with products like the XPS13? They're not... Acer? Lenovo? Hmm. Doesn't have the same ring).

Disappointing issues can be waved away if the price is right. The reason the new Mac Pro is a disappointment to some is price, since, as you said, Wintel boxes (really Linux boxes, no one I know really uses Windows anymore) are much cheaper. However, with an ARM Mac Pro with tons of general-purpose cores as well as GPU compute, Apple does have a strong differentiation from generic boxes. A 128-core ARM Mac Pro with 1/3 the single-core performance of an Intel chip would still be equivalent to 40 Intel cores for parallel compute, which is what a lot of the current Mac Pro is used for anyway. Now imagine it priced at $3000, the same as the current 6-core Mac Pro.

Unlike other ARM server or HPC competitors, Apple has an advantage in leverage on fabrication and economies of scale -- it's got to make 100 million ARM SoCs anyway in iPhones and iPads.

Mr Beardsley · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28784595#p28784595:fwwvg2s2 said:
cateye[/url]":fwwvg2s2]But this is for software that begins life written for iOS. This may be a stupid question, so bear with me, but is it accurate to assume the reverse would necessarily be true? Is SSE optimization the only sticking point that differentiates code written assuming OS X on Intel? I'd be curious to know more about what would constitute a "non-architecture specific language" as DA indicated a few posts earlier.

Off the top of my head assembly, SSE, potentially the Quick Sync video encoding (although the iOS ARM devices have H264 hardware, so the issue would be differences in the software calls if they exist), the Intel AES encryption extensions (again I would guess ARM has an equivalent, and it would be down to differences in the calling implementation), and lastly iOS uses OpenGL ES whereas OS X is a full OpenGL implementation. I'm sure there are more, but those are what jumps out to me.

My opinion is this would be a much smoother transition than PPC -> Intel, at least from a code compatibility and porting perspective. Also I would hope that there is less processor specific code floating around. However, I also doubt there would be anything like Rosetta to help keep existing apps running.

As to your question about MS Office, Microsoft was able to make an ARM version to run on their RT tablets. Between that and the iOS versions it gives hope that it is doable to bring to ARM, although we don't know how much work went into the RT port.

Mr Beardsley · Apr 2, 2015

One additional note. During the PPC - X86 transition, Apple had a great desktop processor in the G5. They did not have a good position in mobile. With ARM, the circumstances are almost exactly reversed. Apple has proved they have top of the line mobile processors. The big question is how to handle the desktop Mac Pro. If the market was heading increasingly toward desktop workstations, I would agree with those saying it is stupid to leave Intel. However, desktop computing, especially workstations, are becoming increasingly niche. I think the market direction is clear. Price and being subject to Intel's timeline are very real issues. I suspect it is more a matter of when Apple moves to ARM for the OS X machines, not if they will. The current MacBook shows that it could technically happen very soon for at least some of the products.

aengblom · Apr 2, 2015

Question: Are more likely to see ARM chip added to Mac OSx or a keyboard added to an iPad? I suspect Apple would take the latter version first. That product could also look a lot like a Macbook. Yes, Apple has to solve the problem where navigating a touchscreen in a laptop form sort of sucks, but iOS is a product with a leading marketshare. OSx is a product with a niche marketshare. That's the platform they want to grow.

cateye · Apr 2, 2015

The current MacBook shows that it could technically happen very soon for at least some of the products.

Technically, yes. But I remain convinced there's little practical reason to do so. And you yourself share the reason: Apple sees the general purpose computer as a niche. The iPad is their answer to the question of what is the future of computing. You or I might argue about how realistic that is, but I have zero doubt in my mind that's Apple's vision, even if it takes five or ten years to play out (see the IBM relationship as Apple's answer to pushing the boundaries of what iOS and the iPad/iPhone form factors can do). As a result, I really don't see what the long term benefit to Apple is to enforce another architecture transition on the Mac. It makes far more practical and business sense for it to continue to coast along as a reasonable money-maker, attracting people via sharp industrial design, while putting in place the steps toward what they believe is the real transition.

While we're all just guessing, I think putting that ultra-powerful ARM chip in an "iPad Pro" and extending iOS' capabilities to eventually replace OS X is the more likely scenario.

Step 1: XCode for iOS. When that happens, the internet will explode.

cogwheel · Apr 2, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28782025#p28782025:1qx702mf said:
Devil's Advocate[/url]":1qx702mf]Benchmarks of the new 14nm Broadwell MacBook are appearing, and it seems the base model's 1.1 GHz Core M is on par in CPU performance with 20nm 1.5GHz A8X in the iPad Air 2

Geekbench
MacBook, 1 thread: ~1900
MacBook, 2 threads: ~4000
A8X, 1 thread: ~1800
A8X, 3 threads: ~4600

Considering Apple probably pays ~$200-300 for the Core M, while it's estimated an A-series SoC (with RAM!) costs Apple $20-40, it seems inevitable Apple will offer an ARM-based MacBook soon with the advent of an even more efficient 14 nm A9 this year. It may even be under $1000.

You're comparing the fastest ARM processor to the slowest x86 processor. There's no current step up in the ARM line for anything above the Macbook. Such steps up could be developed, but...

[url=http://arstechnica.com/civis/viewtopic.php?p=28784679#p28784679:1qx702mf said:
Devil's Advocate[/url]":1qx702mf]Disappointing issues can be waved away if the price is right. The reason the new Mac Pro is a disappointment to some is price, since, as you said, Wintel boxes (really Linux boxes, no one I know really uses Windows anymore) are much cheaper. However, with an ARM Mac Pro with tons of general-purpose cores as well as GPU compute, Apple does have a strong differentiation from generic boxes. A 128-core ARM Mac Pro with 1/3 the single-core performance of an Intel chip would still be equivalent to 40 Intel cores for parallel compute, which is what a lot of the current Mac Pro is used for anyway. Now imagine it priced at $3000, the same as the current 6-core Mac Pro.

The Mac Pro is the lowest volume Mac by far. Non Mac Pros won't be doing easily parallelizable number crunching, and general computer usage is really, really hard to massively parallelize. You're either suggesting that Apple migrates the Mac Pro only to ARM, or they develop ARM into a low core count, high frequency design that will only be used in Macs, and not iOS stuff (way too power hungry, and the PoP packaging that's right for iOS devices is wrong for higher power high frequency stuff). Both options completely evaporate any economies of scale.

[url=http://arstechnica.com/civis/viewtopic.php?p=28784679#p28784679:1qx702mf said:
Devil's Advocate[/url]":1qx702mf]Unlike other ARM server or HPC competitors, Apple has an advantage in leverage on fabrication and economies of scale -- it's got to make 100 million ARM SoCs anyway in iPhones and iPads.

You may wish to actually understand what you're saying. The ARM variant you're talking about Apple developing is not the same chip as goes into their iOS devices in the hundreds of millions, it's a separate chip design (remember, you've posited a lot more cores). It will have less, not more, economy of scale than the Intel chips they currently use, since Intel sells their chips to companies other than Apple, but Apple won't sell their ARM chips to other companies. Remember, Intel gains no economy of scale (in the way you're using the term) on the Core M line because it also makes Xeons. The only way to actually gain economy of scale from the iOS chips would be if Macs above the Macbook incorporated more than one of the same chip as the machines below it. That would require redesigning the chip to support multi-chip SMP purely for a small corner of the Apple ARM device range.

noisy narrow band device · Apr 2, 2015

The never-ending tale of Apple-switches-to-ARM. Another benchmark, another chapter.

Yawn.

Apple certainly likes the negotiation power these rumors provide but a switch to ARM is not going to happen in this decade.

So far Apples' consumption-centric devices - iPhone, iPad, Watch are ARM based, production-centric devices - laptops, desktops, uhm servers - are X86 based.

Being part of the X86 ecosystems in itself has benefits. Windows remains the dominant desktop ecosystem by far and this is where the action is for productivity but also for consumer software development. Keeping the transition barrier low for developers is in Apples interest and benefits consumers.

On the side of the consumption-centric devices Apple has successfully erected a walled garden. Close to 700 million iOS devices make the price of admission worthwhile for developers.

It's a different story for laptops and desktops. Companies like Adobe, Microsoft but also like Blackmagic need to be able to address both the OSX and the Windows (Linux not to forget) market with as little friction as possible.

As a platform company Apple needs production-centric devices and hence cannot afford to drive developers away by yet-another-transition and the need to maintain two builds for different target architectures.

If ARM begins to make inroads into the server-space and a large portion of developers shift over and the tool-chains mature it might make more sense. But go look up what smart people like Linus Torvalds have to say about the competitiveness of ARM chips for "real work" and see yourself.

Last not least: the same day this benchmark came out Microsoft showed an Atom-based tablet. Don't underestimate Intel.

gregatron5 · Apr 2, 2015

The switch to an Arm platform will have nothing to do with consumers. It will have everything to do with developers. Consumers won't really care about the hardware if the software isn't there.

noisy narrow band device and others have nailed the current physical reasons why any possibility of this is on the distant horizon.

From a software perspective, too many companies are writing code too close to the metal. Give Apple five to ten more years to optimize Swift/LLVM, deprecate Obj-C, and get all development on Swift, and then, assuming hardware development has advanced enough, another shift becomes possible.

halse · Apr 2, 2015

I got my grubby paws on a 1.2 for a bit and it was plenty snappy for the types of things it will be used for.
Don't think the target group does 30 layers in Photoshop or 4K video editing, that is 15" rMBP (and desktop) territory.

The Professor · Apr 2, 2015

What do you mean a 1.2? Did you get advanced access to the new MacBook, or are you talking about some Windows machine with the 1.2GHz Core M?

omniron · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28783501#p28783501:2pdmxuq8 said:
Hap[/url]":2pdmxuq8]I really don't understand this fascination with switching Macs to ARM. None of the OS X software out there now will run on the platform, developers will have to support it (which I see as a very up hill struggle), and emulators are not feasible given the comparability in performance. Most users would loose so much in productivity that I don't see it going over well. There are a lot of iOS ARM apps out there, but I think they would also need a lot of changes to work on a non-touch device.

I'm not at all saying apple is planning this, or will do this, but you realize that Apple, within a few years switched from OS 9 to OS X, then switched from PPC to x86?

I guarantee Apple could pull of a switch from Intel to ARM just as smoothly if they wanted to. I don't think they would do this, just because the ability to run Windows in a VM at fast speeds is a compelling feature of Macs, but there's no real technical reason it couldn't be done.

omniron · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28785893#p28785893:3andz0gz said:
noisy narrow band device[/url]":3andz0gz]The never-ending tale of Apple-switches-to-ARM. Another benchmark, another chapter.

Yawn.

Apple certainly likes the negotiation power these rumors provide but a switch to ARM is not going to happen in this decade.

So far Apples' consumption-centric devices - iPhone, iPad, Watch are ARM based, production-centric devices - laptops, desktops, uhm servers - are X86 based.

Being part of the X86 ecosystems in itself has benefits. Windows remains the dominant desktop ecosystem by far and this is where the action is for productivity but also for consumer software development. Keeping the transition barrier low for developers is in Apples interest and benefits consumers.

On the side of the consumption-centric devices Apple has successfully erected a walled garden. Close to 700 million iOS devices make the price of admission worthwhile for developers.

It's a different story for laptops and desktops. Companies like Adobe, Microsoft but also like Blackmagic need to be able to address both the OSX and the Windows (Linux not to forget) market with as little friction as possible.

As a platform company Apple needs production-centric devices and hence cannot afford to drive developers away by yet-another-transition and the need to maintain two builds for different target architectures.

If ARM begins to make inroads into the server-space and a large portion of developers shift over and the tool-chains mature it might make more sense. But go look up what smart people like Linus Torvalds have to say about the competitiveness of ARM chips for "real work" and see yourself.

Last not least: the same day this benchmark came out Microsoft showed an Atom-based tablet. Don't underestimate Intel.

I do see Apple making iOS more robust to the point where you see major productivity apps being run on iOS. I think apple is slowly inching this direction (iCloud drive, rumors of iPad pro, killing of Aperture, Quicktime X,etc.). it's not farfetched at all to see "universal binaries" where end users can start editing a video on their iPad Pro, the continue working on it on their Mac, if they wish, within the same application.

Then if these universal apps start to become popular, then maybe switching from Intel might be reasonable.

HomeKit is going to be very significant too I think, it'll be the beginning of everyday-computing shifting to the background.

Hap · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28791101#p28791101:11j99lqi said:
omniron[/url]":11j99lqi]

[url=http://arstechnica.com/civis/viewtopic.php?p=28783501#p28783501:11j99lqi said:

Hap[/url]":11j99lqi]I really don't understand this fascination with switching Macs to ARM. None of the OS X software out there now will run on the platform, developers will have to support it (which I see as a very up hill struggle), and emulators are not feasible given the comparability in performance. Most users would loose so much in productivity that I don't see it going over well. There are a lot of iOS ARM apps out there, but I think they would also need a lot of changes to work on a non-touch device.

Click to expand...

I'm not at all saying apple is planning this, or will do this, but you realize that Apple, within a few years switched from OS 9 to OS X, then switched from PPC to x86?

Yes, and in both cases Apple was switching to a faster architecture. This allowed older applications to be run in emulation without any appreciable speed loss. That's not possible with ARM at the moment and I don't ever see ARM getting far enough of Intel to make it possible. Don't get me wrong - Apple has done amazing things with their ARM processors, but I don't see them surpassing Intel in the desktop space (I wouldn't mind being proven wrong though).

Aelix · Apr 3, 2015

No VMware or Parallels would be a deal-breaker for many, myself included.

omniron · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28791797#p28791797:1wmevusm said:
Hap[/url]":1wmevusm]

[url=http://arstechnica.com/civis/viewtopic.php?p=28791101#p28791101:1wmevusm said:

omniron[/url]":1wmevusm]

[url=http://arstechnica.com/civis/viewtopic.php?p=28783501#p28783501:1wmevusm said:

Hap[/url]":1wmevusm]I really don't understand this fascination with switching Macs to ARM. None of the OS X software out there now will run on the platform, developers will have to support it (which I see as a very up hill struggle), and emulators are not feasible given the comparability in performance. Most users would loose so much in productivity that I don't see it going over well. There are a lot of iOS ARM apps out there, but I think they would also need a lot of changes to work on a non-touch device.

Click to expand...

I'm not at all saying apple is planning this, or will do this, but you realize that Apple, within a few years switched from OS 9 to OS X, then switched from PPC to x86?

Click to expand...

Yes, and in both cases Apple was switching to a faster architecture. This allowed older applications to be run in emulation without any appreciable speed loss. That's not possible with ARM at the moment and I don't ever see ARM getting far enough of Intel to make it possible. Don't get me wrong - Apple has done amazing things with their ARM processors, but I don't see them surpassing Intel in the desktop space (I wouldn't mind being proven wrong though).

The Intels and PPCs at the time were in the same performance class. There was definitely a performance loss running PPC apps in Rosetta. Apple still owns the Rosetta technology, and this could easily be used for an x86 -> ARM transition.

Apple has more money than Intel, and they bought a very capable semiconductor team. While Intel is the defacto king of the hill now ,it's not a forgone conclusion that Apple couldn't eclipse them with the right leader and resources put in place.

I don't see Apple doing this, because Intel's not a slouch, but it would be interesting position to be in if Apple had the fastest chips per power usage, and if the iPad is faster than some Ultrabooks.

Put more RAM and flash storage, and iOS could grow in features to be a productivity powerhouse (but i don't see this happening any time soon).

ClarkGoble · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28792235#p28792235:1z234wcz said:
omniron[/url]":1z234wcz]Apple has more money than Intel, and they bought a very capable semiconductor team. While Intel is the defacto king of the hill now ,it's not a forgone conclusion that Apple couldn't eclipse them with the right leader and resources put in place.

I don't see Apple doing this, because Intel's not a slouch, but it would be interesting position to be in if Apple had the fastest chips per power usage, and if the iPad is faster than some Ultrabooks.

Intel still wins for fab. They've been adjusting to the new marketplace, albeit after missing out on the whole Arm move. But as we've seen in the past they are nimble and do adjust. I think they know they need to keep Apple happy. I'm sure Apple keeps ARM ready, but I bet they'd rather keep it as a threat to maintain Intel's attention on them.

People do forget that there are big costs to going dual CPU. And it's more complicated than "they're already running on two CPUs." Most code isn't. That's a lot more things that can go wrong. Plus, to be frank, Macs make up a smaller and smaller part of Apple's profits. Why make such a big move on the laptop front? I think we'd expect to see them first blur the Mac and iOS lines more. But if that happens it'll likely primarily be about Apple expanding iOS more. i.e. I think Ive and Cook see iOS and not OS X as the future.

ClarkGoble · Apr 3, 2015

[url=http://arstechnica.com/civis/viewtopic.php?p=28791797#p28791797:ypwdivbi said:
Hap[/url]":ypwdivbi]Yes, and in both cases Apple was switching to a faster architecture. This allowed older applications to be run in emulation without any appreciable speed loss. That's not possible with ARM at the moment and I don't ever see ARM getting far enough of Intel to make it possible. Don't get me wrong - Apple has done amazing things with their ARM processors, but I don't see them surpassing Intel in the desktop space (I wouldn't mind being proven wrong though).

I think users also forget just how disruptive this was to developers. While the Cocoa switch would make this easier now, this would be a stupid thing to throw at developers who already are grumbling quite a bit at the moment. Look at how fast API changes are coming. Throw in Swift, App Store complaints, and the real concern of the viability of a lot of software on iOS and this would make things worse not better. I'm not saying Apple needs worry about a switch. Windows 10 might be a big improvement but isn't apt to slow the leaking and Linux on the desktop for regular users is still a dream. But it may mean fewer people make new important apps and Apple should be worried about that. The dynamics of the App Store right now simply aren't good for the viability of the platform IMO long term. I think Apple can avoid worrying just because there's not a lot of competition on the desktop. On the phones though I think they still have to worry in China about alternatives. Even if Android isn't quite the behemoth they once were, that's a real competition that problems could elevate. (I recognize Android has at least as many problems as iOS for developers - but that could change)

Anyway, given all those issues why on earth would Apple bother switching their lower line laptops to ARM? I really don't see the benefits until some device that runs both iOS & OS X arrives. And they are different enough I can't see that happening any time soon.

secretmanofagent · Apr 4, 2015

It's not necessarily about overtaking x86 as it is being good enough for the given situation. Have the basic user's needs increased drastically over the past ten years? The only major thing that comes to mind is Youtube and Netflix. Swift is capable of optimizing for whichever it's being thrown at, and it would make sense to switch to an in-house processor if it meets the needs. Apple did it, Samsung did it, and it makes sense long-term if you can hold an advantage that is important to you. It reduces the cost of your own devices, and it gives you negotiation pressure points to Intel to say, "Bring your costs down or we'll do it ourselves" for the devices that don't.

noisy narrow band device · Apr 4, 2015

Swift does not offer a performance benefit (if at all) that negates the ARM speed-disadvantages. Above 5 watts the delta between ARM and Intel X86 remains huge and Intel is pushing the watts down faster than ARM is pushing perf up.

It might well be that in 12 months time X86 chips beat ARM chips in the sub 5 watt realm. Then what?!

There is another factor to be considered: the age of Moore's Law is coming to an end. The smaller stuff gets the more expensive design and fabbing become. Intel's game seems to become the last fabber standing. They have 14nm fabs now, there are two or three more process shrinks each about 24-30 months long to be expected before physical limits make further progress very hard indeed.

That is about 8 years away.

Now if you are in the business of selling gadgets build on the premise of ever more cheap transistors - would you piss off the one company that has the best long-turn prospects of dealing with the end of Moore's Law?! Where would that leave you: at the mercy of Samsung, a real enemy; of Global Foundries, a company with shaky success; or TMSC another company with plenty of ups and downs to boot.

Intel and Apple are frenemies and Apple needs something to make the enemy part look plausible. That is what ARM and ARM-rumors are good for and we here are doing our job being part of it.

secretmanofagent · Apr 4, 2015

For Swift, I'm not thinking of it as performance benefits as much as it being even more platform-agnostic than Objective C is.

You're asking the eternal question of "is it easier to scale ARM up or scale x86 down?" So far, it seems easier for ARM.

ARMageddon

Ars Praetorian

Ars Centurion

Ars Scholae Palatinae

Ars Praefectus

Ars Legatus Legionis

Ars Legatus Legionis

Ars Praefectus

Ars Legatus Legionis

Ars Praefectus

Ars Praetorian

Ars Legatus Legionis

Ars Legatus Legionis

Ars Scholae Palatinae

Ars Legatus Legionis

Ars Praetorian

Ars Praetorian

Ars Legatus Legionis

Ars Legatus Legionis

Wise, Aged Ars Veteran

Ars Legatus Legionis

Ars Praetorian

Wise, Aged Ars Veteran

Wise, Aged Ars Veteran

Ars Praefectus

Ars Legatus Legionis

Ars Tribunus Angusticlavius

Ars Praetorian

Ars Legatus Legionis

Ars Praefectus

Well-known member

Ars Praefectus

Ars Praefectus

Ars Legatus Legionis

Ars Scholae Palatinae

Ars Praefectus

Ars Legatus Legionis

Ars Legatus Legionis

Ars Praefectus

Ars Praetorian

Ars Praefectus