Apple A processors in Macbooks

Discussion in 'MacBook Pro' started by elias latach, May 14, 2015.

  1. elias latach macrumors newbie

    Joined:
    Apr 22, 2012
    #1
    Why does not Apple use its own A processors in its Macbooks? I believe a small number of those synchronised together would produce the same power produced by an intel processor. It will allow the computers to get rid of the fans.

    Why??? Why?????? Why???????
     
  2. opeter macrumors 65816

    opeter

    Joined:
    Aug 5, 2007
    Location:
    Slovenia, EU
    #2
    Because one of the selling points for the Mac computers since 2006 is, that you can run the x86 version on Windows OS on them ...

    And these ARM based CPUs today simply cannot compete with Intel actual CPUs in term of raw power/performance.
     
  3. elias latach thread starter macrumors newbie

    Joined:
    Apr 22, 2012
    #3
    By the way they are throwing away boot camp in their new machines (I read something on Macrumors about that). I'm sure that by combining a number of processors that do all of the video processing on the iphone (imovie), can match up to a low-end intel processor, or at least they can make a relevant one (it's Apple!).
     
  4. mcnallym macrumors 6502a

    Joined:
    Oct 28, 2008
    #4
    Can you link to that as I haven't seen anything about removing bootcamp. Certainly they are restricting the Windows releases that support with Bootcamp, ie you may need Win 8.1 rather then 7 for "official" bootcamp workings but haven't seen anything to say that Bootcamp is going altogether.

    ARM is still behind Intel on the price/performance ratio and the Core M can use the same codebase as the regular Core processors from Intel so less codig.

    Coding to make use of Multple CPU's isn't that straightforward ( ask Software Developers ) so throwing multiple cpu's at it isn't goig to work short term, perhaps long term.

    We know that Apple have had ARM based Laptops ( Based o MBA Chassis) in development/testing so it is something that they have looked/looking at but not made out to production machines yet.

    I suspect that until ARM provides something comparable and can run the Intel Apps through something like how PPC on Intel via Rosetta, otherwise would need separate coding or face a lack of apps on launch which wouldn't be good. (Windows RT is a good example of what can go wrong here) That we won't see ARM based Mac's.
     
  5. wheelhot macrumors 68020

    Joined:
    Nov 23, 2007
    #5
    Nope, just they restrict boot camp to windows latest OS which is a pain for some people and if they remove boot camp without an alternative, they'll certainly expect to lose sales from it as many people buy macs these days cause you can dual boot em.
     
  6. dusk007 macrumors 68040

    dusk007

    Joined:
    Dec 5, 2009
    #6
    Forget bootcamp just the incompatibility with all currently shipping apps would be a huge problem. Switching from x86 to ARM would be big load of trouble. That to be worth it, I don't know what would need to happen.

    Also as mentioned Intel CPU cores are just a lot faster. Just adding cores doesn't help or AMD wouldn't do quite so badly. That is 15W+ cores vs. 2W it is not really fair to compare them. They could probably beat Intel's GPU without much trouble but the CPU still matters more and you cannot just scale that cyclone up to run at 3+Ghz.

    I think the bootcamp issue would be one of the smaller problems in such a move.
    You'd need to get a new OSX that supports the ARMv8 and then get developers to recompile all their apps and just recompiling them will only be enough for some apps. Big important programs like the creative ones would need to be rewritten. The ARM chip lack a lot of the Mediaextensions so it will even be slower than one would expect in media producing film/music/picture applications.
     
  7. darngooddesign macrumors G3

    Joined:
    Jul 4, 2007
    Location:
    Atlanta, GA
    #7
    They are not.
     
  8. dmccloud macrumors 6502a

    dmccloud

    Joined:
    Sep 7, 2009
    Location:
    Anchorage, AK
    #8
    First of all, not everything you read on this site is true. Second, Apple isn't dumb enough to alienate a large portion of its users by either dropping Bootcamp or abandoning Intel for a completely new codebase and platform that lacks the performance of the Intel parts.
     
  9. christarp macrumors 6502

    christarp

    Joined:
    Oct 29, 2013
    #9
    Mac OS X is free now, it makes no sense to force users to use it. The money they get is from the hardware, if the hardware runs more stuff then it will open the market to a bigger audience.
     
  10. yjchua95 macrumors 604

    Joined:
    Apr 23, 2011
    Location:
    GVA, KUL, MEL (current), ZQN
    #10
    Because x86-64 architecture is completely different from ARM. Do you have any idea how much of a job it is for developers to rewrite all their code to support a completely different architecture?

    It's just like back from the transition of PowerPC to Intel.

    You don't look like you know much about computers and how hardware+software go hand in hand.
     
  11. brdeveloper macrumors 68020

    brdeveloper

    Joined:
    Apr 21, 2010
    Location:
    Brasil
    #11
    I think there is room for a low cost Retina Macbook equipped with an ARM processor. It would be a pilot project allowing to "test" the reception from the market.

    The upcoming iPad Pro could achieve this goal. It could support all iOS apps inside non-resizable windows, while OSX apps which don't require more than a recompilation would run natively in a OSX fashion (i.e., into resizable windows).

    With time, more complex and legacy apps would be ported, but the important thing is bring the concept, so it will bring interest from 3rd-party developers.
     
  12. christarp macrumors 6502

    christarp

    Joined:
    Oct 29, 2013
    #12
    A few years ago I would've agreed with you, back when the atom processor was intel's foray into the world of low power processors. However, the core M holds its own very well performance wise and is also pretty low power. Maybe not up to the efficiency that an Apple designed ARM chip is, but those don't have the performance of the Core M's yet either.

    Intel has made huge strides into the mobile market with Core M and I think it's not going to be worth transitioning from x86 to ARM anymore with the introduction of that chipset.
     
  13. brdeveloper macrumors 68020

    brdeveloper

    Joined:
    Apr 21, 2010
    Location:
    Brasil
    #13
    Unless they want a better margin for bargaining along with Intel. Double the cores of the iPad Air 2 processor and I think it would probably surpass Core M numbers.
     
  14. MyopicPaideia macrumors 68000

    MyopicPaideia

    Joined:
    Mar 19, 2011
    Location:
    Trollhättan, Sweden
    #14
    Agreed completely with the Core M processor. Even one year ago, the argument for ARM was pretty compelling, right after the A8X benchmarks starting hitting the Web. With the new MB now and the Core M processors, the base 5Y31 is almost double the single core scores of the A8X and at only marginally higher TDP.

    Single core is the most important for day to day use, as most things are single core ops. Doubling the A8X to 6 cores wouldn't improve its single core performance.

    Intel did a good job with the Core M, hats off to them. I was actually in the "bring on ARM" camp last year when this was being discussed more. I do feel that Apple's ARM processors did their bit to put pressure on Intel though. It will be interesting to see what Apple has done with the next gen A9 series chips.

    At some point, all of these chips wil be "good enough" to be able to virtualise the other's environment without noticable performance drop from the user experience perspective.

    I still think with SWIFT, METAL, rumoured iPad Pro, USB-C, etc. that Apple have a longer term plan or goal to do just this, they are just lining up all the pieces right now, and when the time is right, they could very easily start the transition, which will feel natural at first. Starting with an entry level new product category, like the iPad Pro, that runs maybe a freed up version of iOS that includes apps like Finder, Text Edit, Preview and Terminal, supports BT trackpads/mice as well as keyboards, and allows access to the MAS as well as the iOS App store (fat binaries on the MAS being required for apps to show up there, so you get the proper binary for an ARM or x64 system installed) and an inhouse developed virtualisation tool equivalent to Fusion/Parallels optimised for ARM as a viable Bootcamp alternative.
     
  15. leman macrumors 604

    Joined:
    Oct 14, 2008
    #15
    ... which would only matter to benchmark geeks. Single core performance is what matters the most to the average user. Until ARM can catch up in this area, there is not much sense in contemplating their use in personal computers — unless you want to go back to performance standards of 2008.
     
  16. brdeveloper macrumors 68020

    brdeveloper

    Joined:
    Apr 21, 2010
    Location:
    Brasil
    #16
    Are you sure that single core speed really matters to the potential buyer of a Retina Macbook? That is, perhaps only real-time audio processing needs the best single core performance available. With video processing, you can distribute batches of frames to multiple cores. When browsing the web, you would probably improve the overall performance with multiple cores (each ajax request can be processed by a different core, each tab can run on a different core and so on).

    Single core performance is a nice thing when you have intensive (legacy) processing needs, like running an old Photoshop filter. But it seems that the average user wouldn't care with niche legacy software when buying an ARM Mac. A Rosetta-like app for running x86-only apps would be needed, though, but I think most users would get some benefit of more cores in comparison to a higher clock speed.

    ----------

    Well... the Core M is a more recent release than the A8X. Hopefully the next A-processor release will be more efficient and faster, probably surpassing the Core M in multicore performance and diminishing the gap in single core speed.
     
  17. maflynn Moderator

    maflynn

    Staff Member

    Joined:
    May 3, 2009
    Location:
    Boston
    #17
    Why because many people would stop buying them, developers may not follow suit and rewrite their apps (no small undertaking), and you lose out in a lot of the performances and multi-core processing that Intel has built into their chips.
     
  18. leman macrumors 604

    Joined:
    Oct 14, 2008
    #18
    Yes, I am quit sure. There are very few opportunities to multithread a web application on the client (extensions like web workers help somehow, but are still situational). And sure, you can run every tab on its own core — but this is quite pointless, given that usually only one tab is visible and thus only one really requires computational resources. In addition, the average UI-based application is essentially single threaded. The execution model is based around very long periods of waiting, with brief bursts of activity to accommodate the user request. Because event processing, controller logic and view updates are usually done on one thread, you need a CPU that can very quickly power up to its maximal P-state, perform the useful work (often coalesced) and then power down such as quick as possible to conserve power.

    Multiprocessing ability is of course important to ensure smooth operation, but having two hardware threads is usually enough for a contemporary desktop OS. The important realisation (continued from what I said above) is that the usual multiprocessing pattern is very asymmetrical. You have one high-priority thread (usually the one that has to react to user's action) and then a number of lower-priority service threads that either serve data or spend most of their time waiting for something (such as your AJAX response). Often, the most efficient way is to schedule the high-priority thread on one core and the background ones on the second core. This is true for any kind of everyday application that comes to my mind: browsing, office, email (sure, you'd want to fetch you mails on a background thread — but that is an extremely low-priority operation which mostly consists of idle waiting on server response).

    Highly parallel designs, such as groups of weaker ARM cores, only become an advantage if you are dealing with symmetrical multiprocessing patterns. Ideally, you want to subdivide the task in N equal portions, where N is the number of your CPUs. But very few everyday computing tasks exhibit this character. This is commonly the case with image processing, and — yes — real-time audio processing ;) Fourier transforms are easily parallelizable, so you can load up those additional cores with some useful work. In comparison, there is not much chance you can parallelize word processing...

    There is a good reason why a dual-core Intel CPU outperforms a hexa-core AMD one in games.
     
  19. old-wiz macrumors G3

    Joined:
    Mar 26, 2008
    Location:
    West Suburban Boston Ma
    #19
    tell me about it! (retired kernel OS SW developer.)

    ----------

    I worked on a project once that did something similar (different architectures) and trust me it was NOT fun or trivial.

    The compilers can do a lot of the work, but a lot also depends on big-endian/little-endian.
     
  20. leman macrumors 604

    Joined:
    Oct 14, 2008
    #20
    The truth is, virtually any Cocoa application will compile to ARM straight away, unless they do some hacky stuff internally. The alignment and size of basic types are the same between 64 iOS and Intel, as is endianness.\
     
  21. brdeveloper macrumors 68020

    brdeveloper

    Joined:
    Apr 21, 2010
    Location:
    Brasil
    #21
    Not really. Think on a Facebook, a Gmail and a Netflix tab. It's not about visibility, but doing updates even when they're hidden. You can see notifications of new messages on the tab's title while you watch Netflix in another tab. A Java applet could run your home broker's app in another tab while Disqus is refreshing comments in another tab even when you're not looking at it. Scheduling these tasks between cores seem easier than doing all of them in a single core. Don't know why a multicore architecture can't help improving the user experience at this scenario.

    Still don't know why iPad Air2-class speed (that is, it's already faster than a 2010 Core 2 Duo according to Geekbench) can't suit the average user needs - with the advantage of reaching way higher speeds when paralellism is explored by the application.

    Well, a Pentium 4 (perhaps a III) with the help of 8GB of RAM and a PCIe SSD would be enough to execute these tasks at decent speeds with the bonus of having other cores for loading previews or rendering internal panels if needed.

    Ok, real time, frequency domain stuff can be parallelized, but in time domain not really, since you work with a fixed, arbitrary window which can't be splitted into smaller subsamples. Otherwise you would lose information and affect the quality of the filter. Also, latency is important, so you want the best single core performance. Legacy plugins are also single core, so you would need a faster core to get acceptable performance.

    Word processing you mean Office stuff like MS Word? I can imagine multitasking here producing grammar correction, summarization, and other inimaginable natural language processing tasks. Maybe this isn't just feasible on a dual core architecture.


    This can only be explained by legacy A.I. and graphics frameworks. Let's say you have a poor iGPU and the graphics stuff needs help from the CPU. If it could make use of a lot of cores, of course it would run faster if the sum of speeds surpassed a single core one.

    In short, we have to go ahead even if legacy software pushes technology to obsolete standards.
     
  22. leman macrumors 604

    Joined:
    Oct 14, 2008
    #22
    @brdeveloper: what you say makes sense of course, but you are leaving out a lot of details. Most importantly — usually, the most time-critical and work-intensive task in a user app is to update the view (identify, layout and redraw dirty areas). A background tab might still subscribe to notifications and process them in some way, but it does not need to redraw its user area. Most of the time (we are talking about 99% here), it is just idling away, waiting for something to come over the WebSocket/whatever connection it is using. And in a word processor application — sure, it is a good idea to have a background process do the spellcheck etc. — but the really 'slow' task is laying out the text and setting it up to be rendered.

    Ultimately, what matters to the user most is to have a responsive computer. Its all about the illusion that things happen instantaneously (even if they not). Its perfectly ok to wait 0.2 sec more for your notification to come in or your spellchecker to get a word; but layout and redrawing need to happen as quickly as possible. This is why a flexible CPU, which can allocate as much resources as possible to a single, critical thread, is IMO the best choice for the everyday user. Once you get to other application domains, this can change. But overall, a single core system is more flexible than a multi-core system given the same overall performance. The reason is simple — the single core system can emulate the multi-core one essentially without performance loss; but not the other way around.

    Of course, there is some reasonable limit here. As you have said, a P4 can do all this quite easily, so an ARM CPU could do it too of course. The problem is that using one of those CPUs would probably result in noticeable degradation of user experience. However, once we reach some sort of performance milestone, where further increase of single-thread performance will not result in any psychologically measurable improvement in the human-machine-interaction, ARM would certainly become a viable choice.
     
  23. dmccloud macrumors 6502a

    dmccloud

    Joined:
    Sep 7, 2009
    Location:
    Anchorage, AK
    #23
    Actually, the ARM processor wouldn't be able to run OS X apps without some sort of virtualization service, and that would degrade the performance of the apps. The A series processors aren't nearly as powerful as Intel x86 parts, and when you add the virtualization overhead that would be necessary for the scenario you describe, you really wouldn't have much to brag about when it comes to running OS X.
     
  24. leman macrumors 604

    Joined:
    Oct 14, 2008
    #24
    It would, if they were recompiled to the ARMv8 instruction set. Apple already has the necessary technology (cross compilers + fat binaries) as well as the experience (PPC/Intel and Intel32/Intel64 transitions) and could pull this off easily. I am sure they have full OS X versions running on ARM machines in their test labs...
     
  25. TechZeke macrumors 68020

    TechZeke

    Joined:
    Jul 29, 2012
    Location:
    Rialto, CA
    #25
    The whole reason for the original ARM-based Mac rumors was because of the rapid increase in performance of the A Series chips and the fact it was believed that you would need the low-power efficiency of ARM in order to make a super thin, fanless Macbook Air.

    Apple was able to do this with the new Macbook with core M. Core M pretty much kills any need for ARM. Intel pulled through.'

    Considering the huge mess and the amount of compromises that would have to be made(such as losing bootcamp, creating fragmentation, etc) Apple isn't going to switch to ARM just for the sake of switching to ARM. Also, unlike the switch from PowerPC to Intel, there no ARM chips powerful enough to replace the whole line up of Macs, especially Macs like the Mac Pro. Fragmentation would be for years, make it hard for Mac developers(Now they would have to make sure their Mac apps run on ARM if they want to be in the Mac App Store), and would probably cause more harm than good.
     

Share This Page