ARM Mac: Piece of Cake Or Gas Refinery?

by Jean-Louis Gassée

Here we go again, probably “for real” this time. In 2020, Apple is to kick Intel out of its Macintosh line. This could merely be the third processor transition since the Mac’s 1984 birth, but this one will turn out more “interesting” than previous CPU switches.

1984: The first Macintosh comes out with a Motorola microprocessor.
Why: The 68K architecture is deemed more elegant that Intel’s 8086.

1994: The Mac switches to PowerPC microprocessors.
Motorola’s 68K line ran out of steam/time. And the Apple IBM Motorola (AIM) alliance made for politically correct sounds at a time when the Mac wasn’t doing well.

2006: Apple ships Macs using Intel x86 processors.
PowerPC chips designed by committee lost against Intel’s monomaniacal culture.

2020 [estimated]: A rumor, apparently more solid than previous scuttlebutt, tells us Apple will move the Macintosh line to home-grown processors. Codenamed NO86 by snarky insiders, future Mac chips would ship in 24 to 30 months. Successful 1994 and 2006 processor transitions and, since then, a stronger Apple software toolbox all converge to create confidence in the company’s drive to create a new generation of fully integrated NO86 Macs.

One question lingers, though: can Apple develop processors as or more powerful than Intel’s chips?

A bit of history might help.

In 2008, for $278M, Apple acquires Palo Alto Semiconductor, a semiconductor design company. Less than a year after the iPhone starts shipping, the intent is clear: instead of buying off-the-shelf ARM chips, Apple wants to design its own processors for future iPhones.
The move isn’t universally praised, to the contrary.
For perspective, 2007 iPhone shipments are not impressive: $367M in (calendar) 2007, less than 1.4% of Apple’s total $26.5B.
Compared to modest first-year iPhone sales, spending $278M to acquire Palo Alto Semiconductor looks out of proportion. And, of course, let’s not forget a context where Intel reigns supreme having just “captured” the Mac. As always, references to Apple’s “well-known” arrogance are at the ready: Who do these guys think they are, what do they know about microprocessor design and manufacturing?
Answers start to come out in 2010, with the A4 processor powering the first iPad and the iPhone 4. The first Apple-designed processors are deemed serviceable, a reluctant compliment. The 2011 A5 and 2012 A6 follow without drama. But, in 2013 Apple surprises everyone with an industry-first, a 64-bit mobile processor, the A7 — for the iPhone 5S.

At the time, as I recount in a snarky Monday Note titled 64 bits. It’s Nothing. You Don’t Need It. And We’ll Have It In 6 Months, semiconductor “industry observers” pretend to be unimpressed, calling Apple’s 64-bit processor mere markitecture. The pooh-poohing doesn’t resist the first serious benchmarks measuring how ahead of the pack the A7 is. Then, the “desktop-class” phrase appears following a comparison with entry-level x86 processors:

Right away, speculation starts: Apple will soon kick x86 processors out of the Mac line and replace it with its own processors. For the next five years, nothing of the sort happens. Instead, Apple follows up with one interesting home-grown chip after the other: iPhone A8, 9, 10 and A11 processors, S chips for the Watch, W processors for the company’s earphones and T chips for Mac Secure Enclaves.

Back to the question of Apple CPU chips matching Intels processors used by Macs, last Fall, an ExtremeTech benchmark favorably compares the 2017 A11 chip to the Intel Core i5 used in a 13” MacBook Pro:

This is a tilted comparison: the iPhone A11 chip is optimized for mobile, battery-conscious uses, but the MacBook Pro x86 CPU optimized for speed at the expense of watts. Imagine now an Apple chip designed with a different focus: speed versus battery conservation. If the 2017 A11 processor matches the x86 CPU inside a MacBook Pro, a 2020 no-holds-barred line of NO86 chips would easily power high-end Macs.

In 2020, Apple would finally achieve total control of the Mac ”stack”: Processor-OS-Apps. Consider it done.

Not quite. The “different” NO86 processor architecture is misdirection. Instead, what seems to be in the works is both more straightforward and more complicated: a desktop-optimized ARM derivative, one able to run a Mac OS X port and, we're now entering a danger zone, iOS.
We can see the iOS gravitation well lurking nearby, by which I mean the staggering imbalance between Mac and iOS devices: for the December 2017 quarter, comparing device categories, Mac achieved 5.5% of iOS units, and 10.2% in dollar terms:

For Mac app developers, this isn’t a great picture. A new processor, better battery life, lower weight perhaps, might not make a huge difference. Instead, with an iOS-compatible processor running inside new-generation Macs, why not build a new world where the same app would run on both Mac and iOS devices?

This is a dangerous topic. We know what happened with previous attempts to build environments where one app would run on different operating systems. Often referred to as Write Once Run Everywhere (WORE), these superficially pleasing constructs didn’t please the people who actually use and pay for the products. In reality, for an app to be competitive on a given platform, details, details and details need to be attended to under the surface. Such very OS-specific optimizations do not translate to the other platform and thus defeat the WORE theory. Speaking of translations and looking more specifically at Mac OS X versus iOS, one would be facing two languages where words in one have no equivalent in the other. Consider the trouble with wabi-sabi, dépaysement, fingerspitzengefühl or, if you’re really in the mood, Donaudampfschifffahrtsgesellschaftskapitänsmützennadel: the feather on the hat of the captain of a Danube steamship, obviously. You might get the translation by googling segments of the word one at time… Back to bits and bytes, consider iOS having no notion of a cursor, or the Mac not having a touch-screen, or a stylus, to name but a few transaltion challenges.
Recently, we’ve heard rumors of a Marzipan project, an Apple effort to get iOS apps to run on a Mac. As the saying goes, It’s A Mere Matter Of Software. Still, with Apple in control of both OS X and iOS anything’s possible — in theory. Actually, Xcode, Apple’s development system features an iOS simulator that runs your iPhone or iPad app on your MacBook Pro:

But this is a simulation, ARM code running on x86 hardware inside a heavy development environment carrying lots of scaffolding. We don’t know how an iOS app would feel running on an iOS process on ARM Mac hardware — if that’s what’s in the future.

This leaves us with more questions than answers. In particular, we don’t know how Mac hardware and MacOS will evolve when switching to an iOS-compatible ARM processor. In particular, would Apple give some Mac laptops a touch screen and a stylus? Most people I know swear Apple would never do that, or swear at me for even asking the question.
Speaking of strong words, various Apple execs spoke ill of styli or toaster-fridges, and we know what happened…
Thinking of future Macs would be simpler if its putative new processors weren’t iOS-compatible, but here we are. That being said, setting aside inopportune claims of courage, Apple is a cautious company, well aware of the risks in trading a relatively simple life of separate Mac and iOS product lines for a complicated hybrid platform. This coming transition will be interesting to watch.

—

JLG@mondaynote.com