ARM-ed Mac: Not Again Or For Real This Time?

by Jean-Louis Gassée

Once again, we look at what needs to happen for the Mac to shed its old x86 processors and make the tantalizing transition to Apple-designed processors based on the ARM architecture. It was easy in 2006, it wouldn’t be that simple in 2021.

I’m back at my writing station after a two week hiatus. I won’t yield to the temptation of remarking upon the Covid-19 situation other than to point interested readers to Max Roser’s always excellent ourworldindata.org site and, more pointedly to his detailed fact- and science-based discussion of the epidemic: Coronavirus Disease (COVID-19).

Today’s lighter fare will be another go at a now well-aged rumor — or prediction — that Apple’s home-grown ARM processors will replace x86 Intel chips as the main CPU of our Macs, a prophecy that was reinvigorated last week when notorious and prolific analyst Ming-Chi Kuo added a timeline to the prediction: Apple will make the move in the next 12 to 18 months.

This isn’t wild speculation. The iPhone and iPad have run on Apple-designed ARM chips since birth, and while early entries into the Axx line were underpowered, the latest chips created by Apple’s silicon design teams have yielded laptop- or even desktop-class performance. Couple that with Apple’s well-established drive towards vertical integration and control over its products’ key technologies, and the “rumor” seems inevitable.
To that mix, one should add power efficiency. For a roughly equivalent amount of computing power, ARM chips need (and dissipate) 40% to 60% less electric power. This is especially nice for laptops. And, as we’ll see when discussing Ampere ARM chips, a relief from the voracious power consumption of cloud servers.

I first discussed “ARM-ing” the Mac in 2018 (here and here), and went at it again in July 2019 (here and here). Last year, I concluded that a move to the ARM processor created two serious challenges for the Macintosh line, two forks in the product line.

First, ARM-ing the Mac would create what I called a “rolling fork” across the seven different main models that make up the Mac line, from the Mac mini to the Mac Pro. Despite possessing an energetic and infinitely-funded magic wand, Apple wouldn’t be able to instantly replace each member of the line with an equivalent ARM-powered model:

“If (or, more likely, when) the Mac switches to Axx chips, the change won’t be instantaneous. Some Macs will become powered by Apple’s home-grown CPU chips, others, like the Mac Pro, will remain on x86 processors. And thus we’ll have a fork of macOS.”

In Theory, that glorious country where everything works, this isn’t a deal-breaker. But here in Reality, what would customers think of buying a new x86 Mac when the company just made it clear that it’s now past its sell-by date?

That’s the rolling fork problem. One would like to hear what a haruspex of Ming-Chi Kuo’s powers says about Apple’s ability to manage the Herculean Mac-line transition.

I also saw another problem: What happens when the “roll” finally meets the Mac Pro?

The Pro is a monstrously powered machine that costs tens of thousands of dollars and is designed for a (relatively) small audience of content creator professionals and other high-end technical users running fluidics dilutions and the demanding calculations involved in machine-learning applications. Will these users be satisfied with a lesser CPU in the name of Mac line cohesion? Or will the “rolling fork” stop short of the Pro, thus creating a “fork ceiling”?

It’s one thing to (rightly) call an iPhone a pocket supercomputer; it’s another to run applications on a Mac Pro workstation that dissipates a kilowatt or more at peak utilization, using a 28-core Xeon processor at the top of Intel’s line. I had a hard time seeing ARM processors achieve such performance.

I was wrong.

Enter Ampere Computing, founded by ex-Intel President Renée James in 2017 and well-supported by investors such as The Carlyle Group, Oracle, and ARM Holdings (itself owned by Softbank). Ampere designs and sells high-powered ARM chips that compete with the Xeon processors used in cloud servers. Ampere top of the line chips consume less power, about 210 watts, than a competing Xeon CPU needing as much as 400 watts, for about the same amount of computing power — hence investors' interest in a device that could progressively supplant Intel products in tens of millions of servers around the world. Ampere shows us that the ARM architecture can yield the class of chips a Mac Pro would need. And, as it happens, the chips are manufactured by TSMC, the same company that makes Apple’s Axx processors.

So, again in Theory, an ARM-ed Mac Pro is technically possible. But there’s a final problem (again, from a previous MN):

“…the Pro’s sales volume is likely to be in the tens of thousands, not the iPhone’s hundreds of millions. For the Pro, Intel’s high-end designs will be economically more attractive, sharing the investment with other Intel clients.”

Why invest in the development of such a high-end chip for Mac Pro’s low volume? To which an armchair (spontaneous pun) product strategist might say that a lower-powered version of a hypothetical Ampere chip would soon find its way — and an economical justification — in the iMac or Mac mini. All will be well; Apple will take a “rolling fork” hit, but all Macs will be ARM-powered.

Today, the Mac line represents a little less than 8% of Apple total revenue. How much of a temporary revenue disturbance would Apple be willing to endure in order to secure an ARM future for its iconic personal computer? Could the iPad’s rising revenue (6.5% of total) help cover the hit once its user interface (and keyboard with trackpad) makes it more laptop-like?

I, of course, have no idea but this one: ARM-ing the Mac is easier said than done, regardless of its intuitive desirability.

— JLG@mondaynote.com