Why Apple might consider leaving Intel's x86 for its own ARM chips in future Macs

135678

Comments

  • Reply 41 of 150
    asdasdasdasd Posts: 5,686member
    rick 007 wrote: »
    I think you missed the part about normalization of results. This was a very simple attempt to compute the quantity of work that a single core could do per clock cycle.

    A single core 3.0GHz Xeon (2012 westmere I believe) could do 60% more work per a unit of time than a 3.0GHz single core A8. To be fair, I should have used an A-series chip that was shipped in 2012. Or, found someone with a 2014 Mac Pro. That would have given a wider performance gap.

    In all cases though, the Xeons can do more work per core, per clock cycle than A-series chips released around the same time.

    I don't know much about hardware. However Apples GCD is a very easy way for devs to leverage easy multi threading. In fact a function or method can call what looks like a local "block" of code on a seperate thread (the creation of which is abstracted away) and then call back to the main thread (for UI purposes) within the local scope.

    This is a common pattern and most iOS and Mac apps use it religiously leading to a fairly common multi threaded model. (The GCD api generally decides what threads to use or reuse unless you specify a thread by a unique name). And of course lower level API use this fairly commonly . On a fairly conservative Mac app it's common to see 5-10 threads.
    bdkennedy1 wrote: »
    Apple wrote XCode to handle the switch from PPC to Intel.  Now we have Swift. Apple didn't spend 3 years writing a new programming language for the hell of it.

    Well it was one guy for 2 years. I don't see much advantages to swift on Intel vs ARM or vice versa.
     0Likes 0Dislikes 0Informatives
  • Reply 42 of 150
    Quote:

    Originally Posted by sflocal View Post



    I have zero clue how all the x86 politics work, but couldn't Apple somehow license the x86 architecture (like it did with ARM) and essentially build its own x86-compatible chip solely for its own machines and run with it? That way, Apple could contract with its own, very capable chip suppliers and all Intel has to do is just sit-back, and count the royalty money coming in for doing nothing.



    On the flip side, for me personally I love that I can run Windows on my Macs. It's mandatory for me that it is able to do that. Running ARM with x86-emulation is going to suck, if it is even an option technically. I love macs, but for non-mobile devices, it's gonna be a Windows world.



    If Apple licenses and builds its own x86 chips, they'd be awesome I think.

     

    That's not the way Intel works. Just like Apple, Intel does NOT license it's architecture, but builds a product and sells it for whatever price will maximize it's profits. Apple (or anyone else for that matter) licensing the x86 architecture will never happen.

     

    It makes no sense for Apple to switch, to ARM, for it's MacOS devices. The benefits of Windows compatibility alone would dictate that doing so would significantly hurt sales to Enterprise customers. Why not let Intel invest the R&D and manufacturing necessary to crank out successive generations of their processor? Apple will continue to focus it's chip design resources on the A-Series processor (Apple's unique design of the ARM architecture) for it's iPhone and iPad.

     

    I could certainly see Apple developing a larger (12") iPad and maybe a keyboard to turn it into a more 'PC-like' device, but as I've said many times before it makes no sense to introduce an ARM-based Mac. Of course what do I know? Apple has surprised us many times and they certainly could do so again...

     0Likes 0Dislikes 0Informatives
  • Reply 43 of 150
    dasanman69dasanman69 Posts: 13,002member

    Nope. Windows is on 13% of all computing devices.

    You might want to recheck that.

    http://www.businessinsider.com/chart-of-the-day-the-competition-for-windows-10-2014-10
     0Likes 0Dislikes 0Informatives
  • Reply 44 of 150
    Quote:

    Originally Posted by dasanman69 View Post





    You might want to recheck that.



    http://www.businessinsider.com/chart-of-the-day-the-competition-for-windows-10-2014-10



    Business Insider =/= credible source.

     

    http://www.windowscentral.com/microsoft-admits-windows-only-small-percentage-devices

     0Likes 0Dislikes 0Informatives
  • Reply 45 of 150
    asdasdasdasd Posts: 5,686member
    karmadave wrote: »
    That's not the way Intel works. Just like Apple, Intel does NOT license it's architecture, but builds a product and sells it for whatever price will maximize it's profits. Apple (or anyone else for that matter) licensing the x86 architecture will never happen.

    It makes no sense for Apple to switch, to ARM, for it's MacOS devices. The benefits of Windows compatibility alone would dictate that doing so would significantly hurt sales to Enterprise customers. Why not let Intel invest the R&D and manufacturing necessary to crank out successive generations of their processor? Apple will continue to focus it's chip design resources on the A-Series processor (Apple's unique design of the ARM architecture) for it's iPhone and iPad.

    I could certainly see Apple developing a larger (12") iPad and maybe a keyboard to turn it into a more 'PC-like' device, but as I've said many times before it makes no sense to introduce an ARM-based Mac. Of course what do I know? Apple has surprised us many times and they certainly could do so again...

    The iPad certainly needs a keyboard for enterprise use. Text selection and copying and pasting is a mess right now.
     0Likes 0Dislikes 0Informatives
  • Reply 46 of 150
    paxmanpaxman Posts: 4,729member
    Quote:
    Originally Posted by plovell View Post

     

    An iOS-based Mac would be a Very Bad Thing. The primary reason for this is that the user experience is quite different, in ways that we tend to forget. Specifically, the user access to the file system is application-based (all the files for an app are stored in the app's folder) whereas OS X gives the user the ability to mix'n'match what app accesses which files. Totally different UI paradigm.

     

    It wouldn't call it an IOS based 'Mac'. It'd be something else and I am sure it would not be a problem for anyone. People use iPads and Macs without getting confused.

     

    Now, I can well see an ARM-powered MBA. All Apple's apps would be fat-binary (built for both ARM and x86) and many popular apps would quickly follow. For many (most?) it might be a simple recompile since the APIs and system services would be the same. And apps that weren't fat could be run using Rosetta2 - assuming that Apple does it. The reason that Rosetta went away is that the company that built it (Transitive) was acquired and the new owners didn't want to re-license it when the original license expired. But wait - the new owner is IBM. And didn't Apple recently do a major deal with  IBM? Certainly did, so I believe it's quite possible that Apple could now license that technology for an x86-ARM version. Or alternatively Apple could build it itself. The work with Swift shows that it now has the ability. The only question is one of resources and priorities.

     

    Most of the apps most people use on a daily basis also exist for iPad. A low cost IOS based laptop with all the advantages and possibly some added benefits, as well as the limitations of that OS would make excellent sense most of the time for most people. As a superior answer to a Chromebok it would be great. I have a powerful Mac at work and I really don't need my laptop to be top of the line. If so I'd get a MBP which obviously many people would do. My guess is that an IOS based laptop at 699 and up would sell like hotcakes.

     0Likes 0Dislikes 0Informatives
  • Reply 47 of 150
    wizard69wizard69 Posts: 13,377member
    rogifan wrote: »
    What consumer benefit is there to an ARM based Mac? MBAs already get great battery life. And there are new Windows OEM machines that are thinner/lighter than the current Airs. Would an ARM MBA be cheaper than its Intel equivalent?

    In a sense the benefits all lie with Apple, the consumer could benefit from a lower price but for most Apple users that doesn't mean much.

    It all comes back to my old argument that silicon is today what the printed circuit board was to engineers in the 1980's. That is the silicon is where innovation takes place these days. Apple will likely stay with Intel as long as they continue to integrate more and more useful tech onto the dies. Also Intel has to become more aggressive pricing wise. As it is right now AMD's announced chips have a higher degree of integration than Intels Broadwells. The trends in the industry are clear, just about the entire motherboard will be on one chip in the near future. The only thing not on the chip will be analog, RAM and secondary store. Even RAM will eventually be in the same package.
     0Likes 0Dislikes 0Informatives
  • Reply 48 of 150
    tyler82tyler82 Posts: 1,118member
    Good enough is for people who buy PCs and Androids. I want a Mac!
     0Likes 0Dislikes 0Informatives
  • Reply 49 of 150
    tmaytmay Posts: 6,470member
    Quote:
    Originally Posted by karmadave View Post

     

     

    That's not the way Intel works. Just like Apple, Intel does NOT license it's architecture, but builds a product and sells it for whatever price will maximize it's profits. Apple (or anyone else for that matter) licensing the x86 architecture will never happen.

     

    It makes no sense for Apple to switch, to ARM, for it's MacOS devices. The benefits of Windows compatibility alone would dictate that doing so would significantly hurt sales to Enterprise customers. Why not let Intel invest the R&D and manufacturing necessary to crank out successive generations of their processor? Apple will continue to focus it's chip design resources on the A-Series processor (Apple's unique design of the ARM architecture) for it's iPhone and iPad.

     

    I could certainly see Apple developing a larger (12") iPad and maybe a keyboard to turn it into a more 'PC-like' device, but as I've said many times before it makes no sense to introduce an ARM-based Mac. Of course what do I know? Apple has surprised us many times and they certainly could do so again...


    The problem with the x86 hardware market is that it is essentially profitless for most of the OEM's while Intel maintains its high margins by metering out its releases. This isn't acceptable to Apple, which is eager to push for more customization than Intel is willing to allow, and at a development pace that would leave the OEM's in the dust.

     

    What is Apple to do other than abandon Intel where it can in its Mac product line and/or add ARM in new products, and deprecate x86 over some period of time, divorcing itself from x86 and Intel?  

     0Likes 0Dislikes 0Informatives
  • Reply 50 of 150
    wizard69wizard69 Posts: 13,377member
    rick 007 wrote: »
    Not too long ago, I updated a very computation-intenstive piece of software and compared a 2012 3.0 GHz 12-Core Mac Pro to an iPhone 6.  To be fair, the software only leveraged the CPU (i.e. didn't tap into OpenCL which would have only been available on the Mac).
    Very interesting! It would be most interesting to see your results running this app on an iPad Air 2. The idea is to see what the faster RAM would do for you. Right now the Apple implementations of the ARM processors suffer from a small cache, faster RAM might indicate if that small cache is impacting the performance numbers.
    Of course the Mac beat the phone.  But I then "normalized" a bit of the results.  For the sake of ease, I assumed everything would be linear.  I basically computed a theoretical amount of work for a single ARM core running at 3.0 GHz.  Then compared to what a single core (Xeon) did on the Mac.  The Xeon core was about 60% faster.
    That is pretty interesting, is the XEON locked to a clock rate or is it free to scale up and down?
    Now, this is just a single app that was very CPU-bound so obviously not the only measure.  Still, the 60% figure was quite telling.  Furthermore, such apps would realistically be tuned to tap into the GPU.
    I would expect the Xeons to be faster, that is no big deal. I'm more interested in determining how the cache and internal data paths impact performance here. I have this feeling that right now the core in Apple latest A series chips end up starved for information from time to time. The reason I mentioned the latest IPad Air above is that it has an enhanced memory subsystem which has been shown to be very helpful for some apps.
    Can the gap close between ARM and Intel?  Quite possible.  
    All Apple really needs is to leverage the incremental performance increases they have been getting each year.
    But, how will such a gap be closed?  If the answer is to just throw tons of ARM cores to compete with fewer Intel cores, I'm not sure that will work with the vast majority.
    Currently 4-6 cores seems to the the sweet spot for desktop type usage in a non trivial manner. Apple could easily pump out a six core chip that would make an excellent laptop chip. Build it on a process that allows for 2.5 GHz and you have a decent solution for an AiR type laptop.
     Computation-intensive software would argubly be an equal match since it would saturate all the cores.  But often times, sotware is single-threaded and most of your computer's resources are idle.  So you end up comparing just a single core.
    It isn't as bad as it use to be as far as single threaded software goes. Of course there will always be software that bottlenecks on a single thread, however there are far more apps that can leverage multiple cores these days.

    Beyond that most users are not sitting there running just one app. People seem to forget this which really skews the value of a fast single core.
    As a side note, not sure what percentage of Mac users use Boot Camp.  I would imagine that if now on ARM, that would present a problem for those users.

    Nope, haven't run windows on my Mac in ages. See here is another failure with logic, everybody claims that all most users need is a machine to web browse and read E-Mails. For some that is true, the thing is if that is true then Arm doesn't make a difference because those machines would be shipped with native mail, web browser and other apps as part of the system. The reality is this argument is confused or maybe better illogical, as apps would not be a problem for the people claiming to be the target audience.
     0Likes 0Dislikes 0Informatives
  • Reply 51 of 150
    wizard69wizard69 Posts: 13,377member
    Why do you believe this? If the rumors are true, Apple is about to release a laptop that is most certianly a regression performance wise. It isn't about the current user experience it is rather selling the type of user experience customers want. (I'm not too sure customers want the rumored machine).

    Besides the user experience is so dependent upon how the customer uses his machine that you can't realistically say it will be a negative for everybody. In some cases more cores will win for a user.
    kkerst wrote: »
    "Good enough" won't cut it. It has to be equal or better than what the current user experience brings with Intel. Probably already been done - managers are probably just waiting for the AX series chips to come up to speed before actually considering RTM to the masses.
     0Likes 0Dislikes 0Informatives
  • Reply 52 of 150

    Apple won't transfer off of x86 until there's a compelling reason. "Having to wait for Intel" is not compelling - every major PC manufacutrer is waiting too. I'm running on PC stock that's 7 months old now because until we know what's going on with Windows 10 and the 5th gen chips come on out, they're not builiding anything they weren't in the run up to the Holiday season. 

     0Likes 0Dislikes 0Informatives
  • Reply 53 of 150
    wizard69wizard69 Posts: 13,377member

    Everyone seems to forget that the reason why Rosetta worked in the first place is because Intel Core-series chips had so much horsepower in comparison to what Apple had been getting from IBM in the form of PowerPC 970 that they could afford to waste cycles on instruction translation.  That is completely not the case with a theoretical x64-to-ARM shift.  An ARM CPU that could take a 10-15% overhead to translate x64 to ARM64 and have the same performance simply doesn't exist, and probably won't for quite some time, if ever.
    It simply doesn't matter any more. The overwhelming vast majority of Apple users simply don't care about windows compatibility any more. IPad weaned them from that dependency. Beyond that I would expect Apple to ship the machine with all native apps to begin with.

    Even if Apple did want to extend i86 compatibility there are new ways to do that such as recompiling the executable.
    Second, it would put Apple in the position of not being able to differentiate their products on the things they do best - industrial design and software design.
    You don't think that through as Apple is leading the pack with ARM design.

     Intel spends billions every year on CPU designs and lithography research and fabrication process research.  Apple would definitely have to do the CPU design work, and probably have to spend loads of cash helping some other foundry play catch-up.  If they stay with Intel, they don't have to do that because Intel already is, and is spreading the cost amongst all PC OEMs.
    You do realize that Apple has been doing CPU designs for some time now? As for the foundries that is certainly a big expense but it is still far chepaer than paying Intel. The fact is Apple is shipping a better tablet processor right now than Intel is. That is pretty awesome but it isn't a laptop chip. That chip can be easily extended to be a laptop chip though.
    I can't for the life of me figure out why they would want to give up the hardware parity they have enjoyed for the last 9 years in order to have less performance, and far more expense.

    It is pretty clear that the expenses are in fact over all lower. It would be far more expensive for Apple to try to run one of Intels tablet chips in an iPad. Even if they did the thermals would suck. Beyond that you make an assumption about performance that isn't reflecting the potential of Apples A series design. Anything they would put into a laptop would be beefed up just like Intels laptop chips are faster than its tablet chips. I'm not sure why people are so sure that performance would suck so much. Apple has improved ever release of its ARM based chips, they have yet to hit a brick wall.
     0Likes 0Dislikes 0Informatives
  • Reply 54 of 150

    Apple won't transfer off of x86 until there's a compelling reason. "Having to wait for Intel" is not compelling - every major PC builder is waiting too. I'm running on PC stock that's 7 months old now because until we know what's going on with Windows 10 and the 5th gen chips come on out, they're not building anything they weren't in the run up to the Holiday season. 

     0Likes 0Dislikes 0Informatives
  • Reply 55 of 150
    kkerstkkerst Posts: 330member
    Quote:
    Originally Posted by wizard69 View Post



    Why do you believe this? If the rumors are true, Apple is about to release a laptop that is most certianly a regression performance wise. It isn't about the current user experience it is rather selling the type of user experience customers want. (I'm not too sure customers want the rumored machine).



    Besides the user experience is so dependent upon how the customer uses his machine that you can't realistically say it will be a negative for everybody. In some cases more cores will win for a user.

    I agree with you, and "good enough" is relative. Of course no one can anticipate every single use case. It's been my impression of Apple's approach is to take the average use case, ratchet it up many notches so that what you're left with is an above average experience. If they can achieve this with an AX ARM processor, more power to them and I see no reason why it wouldn't sell. I was just stating that they better not screw up what works as is. 

     

    However, I have PPC 2005 Mac mini where I think the performance has never been what I think it should have been from the start, but they fixed subsequent problems with later minis. 

     

    On the notion of adding more cores to get the job done, there's a limit to where this approach won't work anymore. Sooner or later the design (electrical design, not industrial) you are left with is more complicated and not streamlined enough to just meet an average speedy use case. Think battery power here. There's a reason why all iPhones/iPads don't have four cores running at 3GHz. This means the processors are more efficient and don't need to rely on more cores to get the job done.  If they can get OS X to run just as speedy, they'll have pulled another cat out of the bag. 

     

    SELF EDIT: The iPad Air 2 does have three cores in the A8X. So let's go with that. They were able to increase the cores without sacrificing battery life. Great. 

     0Likes 0Dislikes 0Informatives
  • Reply 56 of 150
    wizard69wizard69 Posts: 13,377member
    jameskatt2 wrote: »
    Apple doesn't do "just good enough".
    That is not in its genes.
    Obviously you haven't looked closely at some of Apples newest hardware. We have iMac, Minis and iPad Minis that where done "just good enough".
    Apple does the best products in the world it can make.
    And Intel i3 is not good enough.
    Again Apples current line up proves you wrong. Further if the rumored new MBA is an actual product then they will have gone off the deep end of just good enough.
    It has to be better than Intel's best processors.

    Nope! All we need is overall performance equal to or better than what would have gone in the product if it was built with Intel hardware. Apple never uses Intel best processors anyways so I'm not sure what you are talking about.
     0Likes 0Dislikes 0Informatives
  • Reply 57 of 150
    wizard69 wrote: »
    That is pretty interesting, is the XEON locked to a clock rate or is it free to scale up and down?

    My Mac Pro is fixed at 3.0 GHz.

    And you raised a good point about the RAM. I only have access to an original iPad Air, so cannot do the gen 2 test you outlined.
     0Likes 0Dislikes 0Informatives
  • Reply 58 of 150
    pfisherpfisher Posts: 758member
    Quote:

    Originally Posted by Rogifan View Post



    What consumer benefit is there to an ARM based Mac? MBAs already get great battery life. And there are new Windows OEM machines that are thinner/lighter than the current Airs. Would an ARM MBA be cheaper than its Intel equivalent?

    Many people now don't need or want a full-Intel computer. More specifically, a lot of people are only using iPhones as their primary computer. 

     

    iOS apps on an ARM laptop, but emulated or can run on OS X.

     0Likes 0Dislikes 0Informatives
  • Reply 59 of 150
    wigginwiggin Posts: 2,265member
    Quote:

    Originally Posted by plovell View Post

     



    An iOS-based Mac would be a Very Bad Thing. The primary reason for this is that the user experience is quite different, in ways that we tend to forget. Specifically, the user access to the file system is application-based (all the files for an app are stored in the app's folder) whereas OS X gives the user the ability to mix'n'match what app accesses which files. Totally different UI paradigm.

     

    Now, I can well see an ARM-powered MBA. All Apple's apps would be fat-binary (built for both ARM and x86) and many popular apps would quickly follow. For many (most?) it might be a simple recompile since the APIs and system services would be the same. And apps that weren't fat could be run using Rosetta2 - assuming that Apple does it. The reason that Rosetta went away is that the company that built it (Transitive) was acquired and the new owners didn't want to re-license it when the original license expired. But wait - the new owner is IBM. And didn't Apple recently do a major deal with  IBM? Certainly did, so I believe it's quite possible that Apple could now license that technology for an x86-ARM version. Or alternatively Apple could build it itself. The work with Swift shows that it now has the ability. The only question is one of resources and priorities.


     

    Except that Rosetta relied on the fact the Intel processors had a big performance advantage over PowerPC which made the translation penalty mostly unnoticeable. Similarly, PowerPC chips had a performance advantage over 68k chips which made the Classic environment feasible. And as others have already point out, the current projections for A-series performance is "less than i3" level performance. In other words, a fair deal less performance than even the slowest chip Apple puts in Macs today. Add that to the "Rosetta 2" translation to run Intel code on an A-series chips, and performance would be pretty dismal by comparison.

     

    The question remains...would it be "good enough"? For a lot of users, I think yes. Email, web surfing, etc. But it would be far below what an Intel Mac would be capable of. As others have stated, Apple wants to be a high-rung retailer. Would the sell such a low performing Mac? I think not. But they could market it as a new line of computers all together. Think Apple's incarnation of a netbook done right. But by not calling it a Mac they could avoid a lot of marking confusion and developer angst. It would be the big brother of the iPad rather than the wimpy little brother of the Mac.

     0Likes 0Dislikes 0Informatives
  • Reply 60 of 150
    wizard69wizard69 Posts: 13,377member
    rick 007 wrote: »
    My Mac Pro is fixed at 3.0 GHz.

    And you raised a good point about the RAM. I only have access to an original iPad Air, so cannot do the gen 2 test you outlined.

    I'd offer but all I have is an iPad 3. The faster RAM in IPad Air 2 does help GPU performance a lot, how it impacts CPU performance isn't documented as well. Even if you had Air 2 it would be only one data point, however let's say it improved things modestly and left us with the XEON only being say 50% faster. It wouldn't be unrealistic to see Apple cut that in half with A9.

    The other interesting test here would be to see how fast your code is on say a MBA. For some workloads the cores might actually be faster on the Air when normalized. In any event it would be interesting to see the normalized results when compared to an Air. If ARM goes into anything it will likely be an Air like laptop.
     0Likes 0Dislikes 0Informatives
Sign In or Register to comment.