Last Week on My Mac: Next phase of the Apple Silicon transition

If you follow any Mac developers on social media, you’ll know that the topic of foremost importance last week marked the end of the first phase of the transition to Apple Silicon: the return to Apple of its hardware Developer Transition Kits (DTK).

The DTK was announced at Apple’s delayed WWDC last June, and started shipping to developers in early July. Together with beta-releases of Big Sur and Xcode, they enabled developers to port their code to run in Universal Apps. The hope was that by the time the first Apple Silicon Macs shipped to the general public, there’d be plenty of apps which could make best use of their ARM processors. And it paid off.

You shouldn’t have read too much about the DTK as, being pre-release hardware, it was supplied under a non-disclosure agreement. A few did release information about it, which gave us all tantalising glimpses of what was to come later in the year. The DTK has even earned itself an entry in Wikipedia, although it lacks an image of this very special Mac which never was.

Apple and longstanding Mac users are now veterans of this process, having changed architectures twice before, starting in 1993 and 2005. But no matter how experienced or wealthy a company might be, it’s high risk. Pulling it off requires obsessively thorough planning with an eye for every last detail. Headline measures such as Rosetta 2, Xcode build support and the DTK itself are among the most obvious, but underneath them lies a myriad of minutiae which only those with OCD might notice.

Neither has Apple made the transition any easier, as it coincides with the introduction of macOS 11 Big Sur and its potentially disruptive new features such as the Sealed System Volume. Just at the time when it might need to make repeated changes in firmware, kernel and extensions to fix problems which become apparent with the new M1 Macs, Apple is delivering software updates using a completely new mechanism, which so far has been unable to squeeze the most trivial change into less than 3 GB.

As a developer, documenter and user, from where I sit this first phase has gone unbelievably smoothly. There are three outstanding questions which Apple needs to address in the next phase, which will progressively replace all Intel Macs with Apple Silicon equivalents.

First is resolution of remaining incompatibilities, particularly those with USB-C and Thunderbolt devices. There are still abundant reports of M1 Mac users struggling with USB-C and Thunderbolt docks, displays and storage. Although I haven’t encountered an issue with my external display, external SSDs have been no end of trouble here. This is worryingly reminiscent of Classic Macs, which worked with very few generic peripherals, even printers, severely limiting the choice of Mac users. Hopefully these will be worked out as the M1’s firmware evolves.

Second are macOS updates, which need to be re-engineered to be more affordable for all users, and to work reliably with external bootable disks. Whatever Apple’s original expectations, few users will keep pace with frequent large patch updates, which rather makes them pointless. Apple has been justly proud of high uptake for system updates, and should be worried at the number of Big Sur users who already seem to be falling by the wayside before macOS 11.3 has been released.

Third is potentially the most serious problem, those worrying reports that M1 Macs are ‘wearing out’ their internal SSDs at an alarming rate, because they appear to be writing prodigious amounts of data to them. To some degree, this may be the result of overenthusiastic early adopters using what are intended to be low-end models for high-end workloads. Once Apple has a wider range of Apple Silicon Macs available, many of those users are likely to switch to those with more memory, for example.

One of the reasons advanced for this excessive writing to internal storage is extensive use of virtual memory, although that clearly doesn’t apply to all who have reported high usage. Assuming that Apple will be offering models with 32 GB and more memory, those with heavier demands will be able to move up to those more expensive models.

I think that there’s also more to this problem than first meets the eye. One of the new sub-systems which Apple introduced in Catalina was RunningBoard, which takes a more active role in managing resource use, including GPU and memory, a job which that sub-system already does in iOS/iPadOS. Few apps currently make much use of RunningBoard’s features, although it has progressed further in Big Sur. I therefore wonder whether RunningBoard will be used increasingly to minimise demands on virtual memory, so reducing its use and prolonging the life of storage.

During this second phase of the transition, we should start to see the first apps which offer not just better performance when run on Apple Silicon, but enhanced features, such as use of its neural engine. The first phase has lasted almost nine months; the second may be a little longer, before we enter the final phase in which support for Intel models slowly wanes and ends.