Newsletter Issue #988
April 20th, 2020SO IS APPLE PREPARING ANOTHER PROCESSOR SWITCH?
As much as some people complain that Apple is spending too much time resting on its laurels — particularly with the iPhone and certainly the Mac — change has been a constant for the company over the years. In article in AppleInsider about how the Apple II saved Apple, Steve Jobs is quoted as saying that “all computer architectures have about a ten-year life,” but that was largely about boosting the sale of NeXT gear.
Now the ten-year lifecycle comment has in sone respects been true for Apple. So while the Apple II was quite a successful computer in the 1970s, the Mac ultimately supplanted it although it took a while.
Ten years after the Mac debuted in 1984, Apple made a huge change in the processor architecture with the arrival of the PowerPC. Looking back, it wasn’t such an impressive upgrade at first. It took quite a while for apps to be upgraded to the new architecture, so they ran in emulation, and thus much slower than regular Macs of that era.
It didn’t take too many years for the PowerPC to speed up enough — and for developers, the ones that were still left — to get with the program. So the Power Mac at the end of the decade was much faster than the original models.
That trend continued through the early 2000s. Apple wouldn’t run bake-offs pitting the PowerPC against the Intel Pentium. As much as the critics complained that Apple cooked the books, the tests were pretty accurate; the PowerPC was a lot faster.
Except that PowerPC technology soon hit the wall. To get the Power Mac G5 to become the powerhouse it was, for its time, some high-end models had to be equipped with liquid cooling. Just imagine what might happen if there was a leak. Sure, Pentiums ran hot too, but Apple’s arguments of superior performance were no longer true.
Now it’s hard to know how far the PowerPC might have developed had IBM and Motorola been willing to focus heavily on desktop computing platforms. But with basically one customer — plus a bunch of auto makers and other companies who used them as embedded processors — there was little or no incentive to tame the CPU for notebooks.
Apple kept its options open and, at the WWDC in 2005, the switch to Intel Core processors was announced. It was completed the following summer with the arrival of the Mac Pro, which was essentially a rejiggered Power Mac G5 modified to work with Intel along with compatible support hardware.
Now Apple’s move to Intel, while likely a desperate move at the time, was a genius move too. It allowed for the development of virtual machine apps, such as Parallels Desktop, to run Windows and other x86 operating systems with pretty decent performances. It allowed Apple to create Boot Camp, which runs Windows natively on a separately partition on your Mac.
In the old days, running Windows on a Mac was a painful chore because of the emulation overhead.
I dare say most Mac users never know — or cared — about the transition from PowerPC to “Intel Inside.”
And that takes us to ARM.
Apple’s home-designed ARM-based CPUs rule the roost in the mobile world. They run far faster than Android CPUs, even the ones that have more RAM, cores and high clock speeds. Putting the current A13 Bionic CPU in the $399 iPhone SE was another genius, and practical, move. It means that users who can put up with the loss of some features and a smaller device — and some prefer smaller — can get pretty much the performance of an iPhone 11 for hundreds less.
Apple claims that its A-series silicon can match the performance of most desktop PCs on the market, and that may be true in some respects. I’ve seen benchmarks pitting the 2020 iPad Pro against some 2020 MacBook Air configurations. The new iPad actually uses a modified version of the 2018 A12X, dubbed A12Z. The Air benchmarks somewhat faster in single-core tasks, and is noticeably slower at multi-core.
While canned benchmarks don’t paint the entire picture, if you assume a MacBook Air is basically faster than many PC laptops (most are actually cheaper with lower-grade internals), Apple’s claim is largely true. The A13 would have fared better, and only Apple knows why it wasn’t modified for an iPad. Or maybe it will be later this year. How the pandemic has impacted Apple’s internal development of new gear and services can only be guessed at.
But with the iPad Pro and the MacBook Air converging as nearly comparable notebook alternatives when it comes to productivity — assuming the former has the apps you need — speculation about Macs-on-ARM has only intensified.
With Intel’s development hitting the wall in recent years — most CPU upgrades usually have only marginal performance improvements — it would make sense for Apple to want to look for other options. Having its own chip design department is encouraging, but can an A-series processor beat Intel?
Just matching a mid-priced notebook doesn’t appear to be much of an incentive to go through the drudgery of another processor switch, even though it is far easier nowadays to build multi-platform apps because of Catalyst. That’s the feature that lets a developer produce iPadOS and macOS apps as universal binaries.
Now the potential of Apple’s silicon is only known to Apple. CPUs in an iPad are operating in a resource-constrained environment, with low RAM and the need to restrict power use to allow the units to run all day with a decent-sized battery.
In a notebook, there will be fewer constraints, and maybe these CPUs can be scaled up to deliver far greater performance. We also don’t know just what the forthcoming A14 and future silicon can manage, but it’s a sure thing Apple has been running ARM-based Macs for a while.
At some point, performance may well reach the point where it does best anything Intel can deliver. But that’s just part of the equation. What about running Intel-based apps? What about x86 virtual machines?
Sure, Apple knows emulation, but how much loss would it involve running Intel apps on ARM? Will Apple manage to speed up its processors to compensate without a noticeable performance dip? I don’t think Apple wants to repeat the problems in the early days of the PowerPC.
I suppose Apple could use Metal 2 to assume more of the processing chores to compensate for the performance loss of emulation. Or there are other tricks up its sleeve that we won’t know about until there’s a product to release.
Right now, the speculation has it that there will be a MacBook of some sort with ARM, a way to introduce the technology. You can be assured the user experience will be identical to any Mac, and thus most customers won’t really be aware of a difference.
The real question, however, is whether a full-scale migration is possible. Would Apple be able to deliver an iMac Pro and a Mac Pro on ARM and have it crunch numbers in the same range as those 18-core and 28-core Intel Xeons?
Even if it came close, it would make those uber-priced powerhouses far cheaper.
Apple’s advantage is that its processors are not constrained by backwards compatibility for thousands of potential system and hardware configurations, just its own. Considering how far the A-series silicon has gone so far, it’s very possible we will, as some speculation claims, see the first Macs on Intel late this year, or the next.
If that happens, a full transition won’t take too long.
That is, unless Intel makes a major breakthrough in processor technology that Apple cannot approach. I can’t see the tea leaves, but I do think Apple wants to make the switch, and it’s working very hard to press the “engage” button.
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
Worldwide Licensing: Sharon Jarvis
Print This Issue
