• Explore the magic and the mystery!



  • Newsletter Issue #993

    May 17th, 2020

    REVISITING ARM ON MACS — AGAIN!

    It’s fair to say that if you repeat a somewhat reasonable rumor about Apple year after year, it might eventually come to pass. So as its ARM-based CPUs have become more and more powerful, it seems a given that Apple is poised to ditch Intel one of these days and make the third processor switchover in its history.

    Now when Apple first moved to the PowerPC in 1994, it didn’t seem so much of an issue at the time, although it survived for 12 years. But the Intel rumors were around for quite a while. One of the most interesting ones focused on Apple having a secret project, “Star Trek,” where new Macs were being tested with Intel CPUs at the same time that Steve Jobs was reassuring everyone that they were perfectly happy with the progress of the PowerPC.

    Lest you forget, the most powerful Macs with the G5 required liquid cooling to keep the chassis from frying. It was never tamed for notebook use, so year after year the fastest PowerBooks lost traction against even cheaper Intel-based Windows portables.

    But when Steve Jobs announced that the Mac would move to Intel before the end of 2006, at the 2005 WWDC, my reaction was essentially “it’s about time.” But it was also due to Intel’s introduction of its Core series processor, which delivered better performance and more power efficiency.

    Even as Intel appeared to hit the wall with building faster and less power-hungry silicon, Apple has stuck with Intel for some 14 years now. But as it touts its own A-series CPUs as faster than most Intel-based notebooks, the main question is not if but when things change?

    This year, Apple took a giant step towards turning the iPad into a useful notebook alternative, with improved support for input devices and the introduction of a Magic Keyboard variant for these tablets. Together, they are priced in the same range as the equivalent MacBook, with equal or better performance.

    And remember that Apple’s iOS gear runs its hardware in an extremely resource constrained environment. Just imagine, for the moment, how a future A14 something-or-other, running full bore with more cores, will benchmark against the best from Intel. That possibility appears to be closer than ever.

    Indeed the first Macs-on-ARM may appear as early as late this year or the next. The betting is that a future MacBook Air or MacBook Pro might be among the first models to switch. The higher-end gear that includes the iMac, iMac Pro and Mac Pro won’t migrate till much later. The Mac mini appears to be out of the running for the initial transition, but it would seem a good candidate for an early conversion.

    Why not high-end Macs? Well, the assumption is that professional users would be far more cautious about making the switch and would prefer to wait until the key productivity apps they require are ported over. Performance may not be a serious matter since ARM-based CPUs are now considered suitable for use in datacenters.

    With its Project Catalyst, Apple has worked to make it really easy to build an iOS and macOS app from the same source code. But it will still take time for the usual stragglers, such as Adobe, Microsoft and the companies who build specialized software for movie special effects and other tasks, to make the move.

    But if all it requires as to click a few checkboxes and do some simple optimization in Xcode, things may change quicker than one might expect.

    Now I have little doubt that Apple has already built Mac prototypes that are noticeably faster than anything with Intel Inside. But even when reasonable numbers of apps are compatible, there is another X factor, and that’s virtualization.

    From the early days when Intel-based Mac virtualization software appeared, it was possible to run most any Windows app, plus others that required Intel, at pretty decent speeds. Apple’s Boot Camp made it possible to set aside a special disk partition on your Mac to run Windows in its native environment.

    When, and I’m no longer saying if, Apple switches to ARM, it will have an emulation feature that’ll allow apps that require Intel hardware to run with decent performance. But what about virtualization? What about the expected performance hit emulating Intel? What sort of toll will it take?

    Now perhaps Apple can boost speeds of its native CPUs quite enough to compensate, more or less. Also remember that Microsoft has already ported Windows to ARM, but without direct compatibility for apps that require Intel. Still, it might give Apple some alternatives, even if it required licensing technology from Microsoft.

    By focusing on the lower-end gear at the start, Apple might not make a big deal about Windows virtualization at the start, other than allowing it to occur with passable speeds. When the transition moves up the line, that issue might very well be solved for the most part. I’ve suggested, in fact, that Apple might funnel Intel compatibility through its Metal 2 graphics hardware scheme, thus resolving the performance barrier.

    Nothing I’ve written here should present a problem for Apple. It realizes the obstacles, and has no doubt fired up some solutions in its test labs.

    Regardless, as with previous processor transitions, the people who care the most will be power users. Most people who use Macs — even in the enterprise — will be able to use their existing apps and never or rarely notice that a major hardware change has occurred.

    But if perfectly good CPUs are already out there, why would Apple bother to make the switch? Other than the obvious “because it can,” it’s becoming increasingly clear that Intel has hit the wall scaling down its chips and boosting performance in noticeable ways. Promised silicon upgrades tend to run late, so the newest Mac might not be so much faster than the motel it replaces.

    And if you look at the usual bills of materials, you’ll see that it cost Apple a whole lot less to build its own CPUs than to buy up chips from Intel. This may not make much of a difference for entry-level gear, but those top-of-the-line Intel Xeon processors are hugely expensive. The top-of-the-line adds $7,000 to the cost, and much of that is what Intel charges Apple.

    Now imagine, just imagine, if Apple could build an ARM-based equivalent workstation chip and only exact $3,000 or $4,000 extra. A souped up Mac Pro may begin to become almost affordable.

    All right, ECC RAM and solid state drives are still costly, and it doesn’t appear that Apple is interested in entering either market.

    Other than cost, why would Apple want to undergo another processor transition? Even if it’s done as smoothly as the Intel transition was, it will still be a costly proposition where glitches are a given.

    One obvious answer is that Apple craves full control over its hardware, and the more it brings in house, the less it has to depend on third parties to deliver the goods. Even as Apple is buying G5 cellular modems from Qualcomm, which required settling a nasty set of lawsuits, it also bought Intel’s modem division and will, a few years hence, do it all by itself.

    In addition to erratic ship schedules, there are many features in Intel silicon that support legacy PCs that Apple doesn’t need or care about. Forgetting some of those recent issues with security lapses in Intel CPUs, Apple’s A-series chips are optimized for its own operating systems. It doesn’t have to worry about the needs of someone else’s personal computers.

    Sure, it’s possible all those expectations of ARM-based Macs are overly optimistic, or just plain wrong. But the signs are activity are there. Apple is already providing ARM-based chips for low-level functions on many new Macs, a clear first step towards a long-term goal.

    I’m hesitant to make predictions, but I will not be at all surprised to see one or more Macs featuring ARM Inside, by this time next year — and likely earlier.

    That doesn’t mean I’m ready to buy one of the new models, but my current gear is really long in the tooth.

    THE FINAL WORD

    The Tech Night Owl Newsletter is a weekly information service of Making The Impossible.

    Publisher/Editor: Gene Steinberg
    Managing Editor: Grayson Steinberg
    Marketing and Public Relations: Barbara Kaplan



    Share
    | Print This Issue Print This Issue

    Leave Your Comment