The big problem is code developed to support SSE and AVX SIMD instruction sets. I doubt that could be translated automatically into the ARM SIMD instructions.
Great analysis here. My hunch tells me that over the next few years as apple expands its processor team and continues to make exponential strides there they may eventually transition, but at least a few years down the road and possibly only after they can safely phase out Samsung as a foundry.
Also lets not forget that apple and intel need to have the no poaching settlement well behind them, as I imagine Apple would want to add hardware x86 recompiling or emulation on chip, and what better resources to do that.
I can't take any article seriously that talks about leaving Intel without also commenting on Intel's Thunderbolt. I don't think Mac market would survive a rapid apandonment of Thunderbolt.
I can't take any article seriously that talks about leaving Intel without also commenting on Intel's Thunderbolt. I don't think Mac market would survive a rapid apandonment of Thunderbolt.
You think entry-level Macs are proped up by TB peripherals?
The article talks about 20 million Mac sold, not simply entry level Macs.
Yes I think TB makes MacMinis and MacBook Airs more useful. Given TB pass through they can handle daisy chaining whereas USB3 can't.
It's also boarding on absurd to have an OS and all developers doing ARM and Intel versions assuming MacBook Pros, iMacs, MacPros remain intel. It brakes the compatibility that users expect when they can use the same TB peripherals on their high end iMac and MacBook Air for example.
It simply makes no economic sense for Apple to move to Arm unless they moved every Mac, developed their own connector with the same or better capabilities as TB and added backward compatibility with current TB peripherals.
Now if you think it's worth it for Apple to make an Apple Laptop running iOS on Arm that might make sense but that would really just be an iPad with a keyboard.
You can't consider Arm for Computers without considering Peripheral compatibility and software development for the entire (not just low end) platform. Maybe Apple has an answer for that but it's a major factor and the article doesn't touch that issue.
Excel for OS X will be feature compatible with the Windows version in Office 2015. You're also missing the fact that MS ported a 99% feature complete version of full-blown Office to ARM for the RT devices. With a few tweaks, that version becomes 100% feature complete.
AMD is a joke anymore; people say they want x86 power, then AMD is not your option. They don't have the talent nor the cash to compete with Intel anymore.
Feature compatibility, or lack thereof, is only one of the problems with Excel for Mac. Lack of performance is also a huge issue for many users. Office for Mac barely crawls along. You can get far superior performance and fewer weird bugs running Office for Windows in VirtualBox (the slowest Windows on Mac solution).
Windows still commands 90% of the PC market despite the missteps with Vista and Windows 8 and the overall movement from desktop to mobile. Apple choosing to be incompatible with that 90% seems crazy.
Apple should, however, add better external external keyboard/pointer support into iOS. And improve multi-tasking.
Getting something roughly akin to Metro-type multi-tasking and application launching, plus support for an external trackpad, should be easy enough and would turn iPads from almost good enough to replace Netbooks to plenty good enough to replace Netbooks, without hurting Macbook Pro sales much.
But, it doesn't seem as though Apple will do this. Perhaps it considers that it is preserving its low-end laptop market. Perhaps it just doesn't want to sully iOS with a non-touch interface and confuse developers with having to support a mouse cursor. More likely, it is just one of a few Apple blind spots that show up and never seem to go away, or at least last far longer than they should, like Mac OS X's questionable multi-monitor support that finally seems to be improving over the last two releases.
If rumors of a larger iPad are true, then it would almost have to fix these limitations in order to hit its market correctly.
Maybe they just might make a laptop out of an iPad. That would make things simple.
I am firmly in this camp though there is significant resistance to this idea for reasons I don't fully get. I think folks think it is an either or proposition. An IOS based laptop would be a new device so nothing really changes. It would be a low end device but no lower end than a current iPad. I'd be all over an IOS based laptop.
Many people misunderstand what it would be, and think an iOS-based laptop would just be like OS X. Which is quite false.
Some others think that an ARM-based laptop (say, MacBook Air) would have to run iOS. Which again is silly. You could have such a laptop and it would run fat binaries (x86/ARM, just like we used to have PPC/x86 fat binaries). And even run x86 binaries if Apple builds Rosetta2. OS X with ARM is no big problem, just a couple of smaller ones.
As I have said elsewhere - the big difference between iOS and OS X is how users interact with the file system. That is, are files stored in the app's folder (iOS) or are they in your home directory (OS X) ?? This of course immediately points up the issue that iOS is single-user, and not multi-user like OS X. Making iOS multi-user would force the file-storage scheme to change, as you couldn't readily have files for several users in the app's directory.
I also would like something like an iOS-based laptop. An iPad with wireless keyboard goes part way but I can't pair a Magic Trackpad or Magic Mouse. Reaching to the screen to select something is a real pain, as Steve said many times.
An ARM based MacBook with a headline "Million compatible apps", why would it be that far away? Apple has a record of changing architecture every decade or so. so now its been a decade with intel. Time is up and I guess it's only a matter of months they announce it. Next WWDC ?
I agree with you, but a lot of folks seem only to focus on why this won't happen.
DED is showing an uncharacteristic lack of imagination here.
this is a dumb article and a dumb hypothesis, high end users be they Apple or Microsoft customers have been loosing out of late due to Intel shifting focus to the low end of the market where the volume and the money is, Apple replacing risc with cisc will only be of benefit to the low end market these chips will never replace Oct core sever processors and we at the high end will be left hanging. Worse, if the OS transitions to the new architecture and all the high end pro apps are still on Intel I predict a huge mass exodus of high end professional users migrating to Windows just as fast as they can.
Except that high end pro apps were on the PowerMac before the transition to Intel, and this Pro market was one of Apple's most profitable and important markets during the "dark days" when Steve Jobs had just returned to the company. Before the iPod, before the iPhone, Apple had to first become profitable again selling just PowerPC-based Macs, and they did.
People don't get it: NOBODY ports anything to a CPU except the OS and compiler engineers. The OS (Darwin) and the compilers for ARM are already finished otherwise there were no iOS software. Apple's MACH/BSD kernel in Darwin always is and always has remained multi-architecture. It doesn't matter if it's PPC, x86, x86-64, ARM-32 or ARM-64.
In essence: OS X for ARM is but a recompilation for Apple, plus a few device drivers, if they run it on hardware using different/additional components than the iPads. Heck, I'd be surprised if that had not been running in Apple's labs for quite some time.
Transition thus, for OS X native apps is recompilation and some QA, nothing more, except for an absolutely tiny fraction of the market that still might use assembly code for some small aspects of their code, but that's negligible.
A move to ARM wouldn't be about cost, but about performance: for many years now (GCD, OpenCL) Apple has been telling developers to increase performance with multithreaded code, and to stop relying on increased clock speed and similar tricks to increase single threaded performance, which simply starts hitting physical limits, particularly in the performance-per-Watt measure.
While an hypothetical A9 or A10 with a clock speed not limited by iPad/iPhone heat dissipation limits and battery limits might have a lower single-core performance than an equivalent Intel Core i3 CPU, Apple could have a 12+ core ARM chip in the power envelope of a dual-core or quad-core Intel chip; so we're talking 12+ cores vs. 2 to 4 cores (4 to 8 threads with hyperthreading) and that could be able to match performance while scaling power usage by turning off anywhere from 0 to 11 cores, while still being able to do low CPU tasks like web browsing or even video streaming if decoding is done in the GPU.
So not only would Apple be able to save costs, it would be able to offer battery life like no other in the industry.
M$ failure with WinRT is irrelevant: it has nothing to do with ARM, but with the fact that they didn't offer the full desktop environment with the RT version in an attempt not to cannibalize their higher profit x86 "real" Windows sales; it was a marketing decision that backfired (combined with the fact that the ARM chips they were targeting at the time were vastly underpowered compared to the 64-bit chips Apple has now)
In less than two years we may have iOS X, the ideal point to merge iOS 10 with OS X, for an OS that can switch between desktop and touch mode and apps that can switch between desktop, full-screen and touch modes. The OS, compilers, binary architecture etc. are all ready. What's needed is simply the integration of Cocoa and CocoaTouch on top of the same OS (remerge the two Darwin branches) and some additional APIs for apps to switch the type of GUI offered based on context. This is within reach: OS X already offers UI modes for windowed and full-screen modes, iOS allows for UI modes depending on screen size (iPhone, iPhone plus, iPad) so the basic mechanism are already present.
Apple could further integrate their offerings with anything between an iPhone to a massively parallel MacPro running on the same CPU and OS architecture.
Yes, even a MacPro could go ARM: one could have 64 cores where there are now 12, and the GPUs could be just as powerful as what there is now, as these are not directly depending on the CPU choice.
So really, the only real obstacle is x86 emulation for running legacy Windows apps. With more and more corporate computing being cloud based, and with a more OS agnostic M$ (Ballmer is gone, as is evidenced by M$ Office for iOS), this may be less of an issue than ever; and possibly, the streamlined ARM-64 instruction set might be particularly well suited for CPU emulation, or for Macs, Apple could ditch the ARM-32 legacy instruction set and use the freed silicon for some special instructions needed for efficient x86 emulation.
I can't take any article seriously that talks about leaving Intel without also commenting on Intel's Thunderbolt. I don't think Mac market would survive a rapid apandonment of Thunderbolt.
I know, right? People who own Thunderbolt devices might leave the platform. Both of them!
<a data-huddler-embed="href" href="/u/27437/digitalclips" style="display:inline-block;">@digitalclips</a>
I know you're being funny, but plenty of people still have to run Windows, unfortunately. Excel doesn't even compare between the platforms, and seems the Mac version can not keep up in a business environment. 3D Studio Max is still a large standard, and has yet to make its way to the Mac platform (Autodesk, get off your ass!).
I tend to agree with the article. I think ARM is a ways off. I think the AMD option argument is the most interesting of all. If price truly is the defining factor, Apple could help fund AMD to get back on track and use that as a bargaining chip against Intel.
My fear is that Intel has dropped the ball the last year, they missed their Broadwell ship time and just seem to be off lately. A sign that was quite similar to the way IBM was handling PPC towards the end of that era. I think the mobile market caught Intel off guard.
At a minimum, we have common developer tools to compile for both ARM and Intel. I don't think the transition would be much, if at all, assuming the app is on a modern codebase. Swift being released could also point to a transition path as well, though some years off which may align with an ARM transition.
Personally, I'm not a fan of it at all. I'm quite happy with Intel as a power user, and wish Intel would work something out so Apple and them could co-design mobile cpus that are both powerful and energy efficient. Intel has a lot of great minds, they need to figure something out, soon.
Perhaps i should have said 'who needs it ' on a consumer Mac. I already stated I think higher end Macs will be Intel for a while so they have VMWare.
I agree with you, but a lot of folks seem only to focus on why this won't happen.
DED is showing an uncharacteristic lack of imagination here.
Funny. I was thinking DED was showing an uncharacteristic lack of fanaticism here.
This article reminded me of the excellent series of articles AI published during Apple's big hardware and software transitions in the long ago past. Those articles were objective, educational, mercifully lacking in extreme hyperbole, and prompted constructive commentary. I much prefer that kind of reporting to u-ra-ra we're-number-one-everyone-else-sucks cheerleading articles.
I know, right? People who own Thunderbolt devices might leave the platform. Both of them!
Live everyone using Adobe Creative Cloud on Mac, professional NLE users (FCPX and Avid). Blackmagic users (both software and hardware) just to name a few. People spending $2000+ dollars on a computer are all likely using Thunderbolt Storage and Thunderbolt i/o devices.
Comments
You think entry-level Macs are proped up by TB peripherals?
The article talks about 20 million Mac sold, not simply entry level Macs.
Yes I think TB makes MacMinis and MacBook Airs more useful. Given TB pass through they can handle daisy chaining whereas USB3 can't.
It's also boarding on absurd to have an OS and all developers doing ARM and Intel versions assuming MacBook Pros, iMacs, MacPros remain intel. It brakes the compatibility that users expect when they can use the same TB peripherals on their high end iMac and MacBook Air for example.
It simply makes no economic sense for Apple to move to Arm unless they moved every Mac, developed their own connector with the same or better capabilities as TB and added backward compatibility with current TB peripherals.
Now if you think it's worth it for Apple to make an Apple Laptop running iOS on Arm that might make sense but that would really just be an iPad with a keyboard.
You can't consider Arm for Computers without considering Peripheral compatibility and software development for the entire (not just low end) platform. Maybe Apple has an answer for that but it's a major factor and the article doesn't touch that issue.
Lol no.
Excel for OS X will be feature compatible with the Windows version in Office 2015. You're also missing the fact that MS ported a 99% feature complete version of full-blown Office to ARM for the RT devices. With a few tweaks, that version becomes 100% feature complete.
AMD is a joke anymore; people say they want x86 power, then AMD is not your option. They don't have the talent nor the cash to compete with Intel anymore.
Feature compatibility, or lack thereof, is only one of the problems with Excel for Mac. Lack of performance is also a huge issue for many users. Office for Mac barely crawls along. You can get far superior performance and fewer weird bugs running Office for Windows in VirtualBox (the slowest Windows on Mac solution).
Windows still commands 90% of the PC market despite the missteps with Vista and Windows 8 and the overall movement from desktop to mobile. Apple choosing to be incompatible with that 90% seems crazy.
Getting something roughly akin to Metro-type multi-tasking and application launching, plus support for an external trackpad, should be easy enough and would turn iPads from almost good enough to replace Netbooks to plenty good enough to replace Netbooks, without hurting Macbook Pro sales much.
But, it doesn't seem as though Apple will do this. Perhaps it considers that it is preserving its low-end laptop market. Perhaps it just doesn't want to sully iOS with a non-touch interface and confuse developers with having to support a mouse cursor. More likely, it is just one of a few Apple blind spots that show up and never seem to go away, or at least last far longer than they should, like Mac OS X's questionable multi-monitor support that finally seems to be improving over the last two releases.
If rumors of a larger iPad are true, then it would almost have to fix these limitations in order to hit its market correctly.
Maybe they just might make a laptop out of an iPad. That would make things simple.
I am firmly in this camp though there is significant resistance to this idea for reasons I don't fully get. I think folks think it is an either or proposition. An IOS based laptop would be a new device so nothing really changes. It would be a low end device but no lower end than a current iPad. I'd be all over an IOS based laptop.
Many people misunderstand what it would be, and think an iOS-based laptop would just be like OS X. Which is quite false.
Some others think that an ARM-based laptop (say, MacBook Air) would have to run iOS. Which again is silly. You could have such a laptop and it would run fat binaries (x86/ARM, just like we used to have PPC/x86 fat binaries). And even run x86 binaries if Apple builds Rosetta2. OS X with ARM is no big problem, just a couple of smaller ones.
As I have said elsewhere - the big difference between iOS and OS X is how users interact with the file system. That is, are files stored in the app's folder (iOS) or are they in your home directory (OS X) ?? This of course immediately points up the issue that iOS is single-user, and not multi-user like OS X. Making iOS multi-user would force the file-storage scheme to change, as you couldn't readily have files for several users in the app's directory.
I also would like something like an iOS-based laptop. An iPad with wireless keyboard goes part way but I can't pair a Magic Trackpad or Magic Mouse. Reaching to the screen to select something is a real pain, as Steve said many times.
I agree with you, but a lot of folks seem only to focus on why this won't happen.
DED is showing an uncharacteristic lack of imagination here.
I dont' think Apple buys the xeon chips. I think they use HP Proliant DL servers in their datacenter.
Except that high end pro apps were on the PowerMac before the transition to Intel, and this Pro market was one of Apple's most profitable and important markets during the "dark days" when Steve Jobs had just returned to the company. Before the iPod, before the iPhone, Apple had to first become profitable again selling just PowerPC-based Macs, and they did.
The OS (Darwin) and the compilers for ARM are already finished otherwise there were no iOS software.
Apple's MACH/BSD kernel in Darwin always is and always has remained multi-architecture. It doesn't matter if it's PPC, x86,
x86-64, ARM-32 or ARM-64.
In essence: OS X for ARM is but a recompilation for Apple, plus a few device drivers, if they run it on hardware using different/additional components than the iPads. Heck, I'd be surprised if that had not been running in Apple's labs for quite some time.
Transition thus, for OS X native apps is recompilation and some QA, nothing more, except for an absolutely tiny fraction of the market that still might use assembly code for some small aspects of their code, but that's negligible.
A move to ARM wouldn't be about cost, but about performance: for many years now (GCD, OpenCL) Apple has been telling developers to increase performance with multithreaded code, and to stop relying on increased clock speed and similar tricks to increase single threaded performance, which simply starts hitting physical limits, particularly in the performance-per-Watt measure.
While an hypothetical A9 or A10 with a clock speed not limited by iPad/iPhone heat dissipation limits and battery limits might have a lower single-core performance than an equivalent Intel Core i3 CPU, Apple could have a 12+ core ARM chip in the power envelope of a dual-core or quad-core Intel chip; so we're talking 12+ cores vs. 2 to 4 cores (4 to 8 threads with hyperthreading) and that could be able to match performance while scaling power usage by turning off anywhere from 0 to 11 cores, while still being able to do low CPU tasks like web browsing or even video streaming if decoding is done in the GPU.
So not only would Apple be able to save costs, it would be able to offer battery life like no other in the industry.
M$ failure with WinRT is irrelevant: it has nothing to do with ARM, but with the fact that they didn't offer the full desktop environment with the RT version in an attempt not to cannibalize their higher profit x86 "real" Windows sales; it was a marketing decision that backfired (combined with the fact that the ARM chips they were targeting at the time were vastly underpowered compared to the 64-bit chips Apple has now)
In less than two years we may have iOS X, the ideal point to merge iOS 10 with OS X, for an OS that can switch between desktop and touch mode and apps that can switch between desktop, full-screen and touch modes. The OS, compilers, binary architecture etc. are all ready. What's needed is simply the integration of Cocoa and CocoaTouch on top of the same OS (remerge the two Darwin branches) and some additional APIs for apps to switch the type of GUI offered based on context.
This is within reach: OS X already offers UI modes for windowed and full-screen modes, iOS allows for UI modes depending on screen size (iPhone, iPhone plus, iPad) so the basic mechanism are already present.
Apple could further integrate their offerings with anything between an iPhone to a massively parallel MacPro running on the same CPU and OS architecture.
Yes, even a MacPro could go ARM: one could have 64 cores where there are now 12, and the GPUs could be just as powerful as what there is now, as these are not directly depending on the CPU choice.
So really, the only real obstacle is x86 emulation for running legacy Windows apps.
With more and more corporate computing being cloud based, and with a more OS agnostic M$ (Ballmer is gone, as is evidenced by M$ Office for iOS), this may be less of an issue than ever; and possibly, the streamlined ARM-64 instruction set might be particularly well suited for CPU emulation, or for Macs, Apple could ditch the ARM-32 legacy instruction set and use the freed silicon for
some special instructions needed for efficient x86 emulation.
Before we get too far ahead of ourselves, how about Apple updating the Mac Pro to Haswell v3 Xeons? And a pair of Xeons would be very nice.
Perhaps i should have said 'who needs it ' on a consumer Mac. I already stated I think higher end Macs will be Intel for a while so they have VMWare.
I agree with you, but a lot of folks seem only to focus on why this won't happen.
DED is showing an uncharacteristic lack of imagination here.
Funny. I was thinking DED was showing an uncharacteristic lack of fanaticism here.
This article reminded me of the excellent series of articles AI published during Apple's big hardware and software transitions in the long ago past. Those articles were objective, educational, mercifully lacking in extreme hyperbole, and prompted constructive commentary. I much prefer that kind of reporting to u-ra-ra we're-number-one-everyone-else-sucks cheerleading articles.
Excellent article DED!
DED showed both sides.
I know, right? People who own Thunderbolt devices might leave the platform. Both of them!
Live everyone using Adobe Creative Cloud on Mac, professional NLE users (FCPX and Avid). Blackmagic users (both software and hardware) just to name a few. People spending $2000+ dollars on a computer are all likely using Thunderbolt Storage and Thunderbolt i/o devices.