Soba said:
Good luck deciphering this if you're not a fly on the wall at Apple's hardware team meetings. All we can do is speculate based on past actions.
That would be hard if Apple was extremely inconsistent but they aren't.
First, hardware support they are explicit.
Obtaining service for your Apple product after an expired warranty
Learn about your options for getting service and parts for Apple devices that are past their warranty period.
support.apple.com
Strongly related to that Apple largely views the system as a hardware+software unit. macOS is licensed to the hardware. ( isn't sold loose/decoupled). So if stopping on one "half" a unified whole why would they continue on the other "half" and vice versa.
The Mac Pro 2009 , 2010 , 2012 went on the vintage obsolete lists at different times. Just like the Vintrage/Obsolete policy lays out it would. And the official macOS de-support basically the same way. [ There was lots of self congratlatory "We hacked our why around that de-support with firmware hackery" in these Mac Pro forums. However, that is missing the forest for the tree. The software de-support policy is very consistently tied to the hardware long term support policy ]
Soba said:
The first Mac Pro was released in August 2006, at which point Apple's transition from PowerPC processors to Intel was complete and the Power Mac G5 was discontinued. The G5 was supported by then-current Mac OS X 10.4 (Tiger) and the following release, 10.5 (Leopard). Then, 10.6 (Snow Leopard) dropped all PowerPC support when it was released in August 2009, exactly 3 years after PowerPC systems were completely retired. Leopard received its last security update in (I think) May 2012, almost 6 years after the retirement of PowerPC.
The PowerPC->x86 is a skewed point in time. Super deep inferencing off of that will likely get decouple from what Apple is out to do. There are four major differences between now and the x86 transition.
First, Apple owns the x86->arm translator tech. For PPC->x86 Apple slaps their name "Rosetta" on top of someone else's tech that they had to license and pay for. So it is really not all that surprising that they wanted to stop paying sooner rather than later.
[ In 2008, IBM bought transitive. Any future licensing pricing renewal discussions were not going to be with some struggling, small start up. Even if Apple wanted to keep paying ... they wouldn't after that. ]
There is no 'per unit' costs for this limited x86_64 -> arm translator. (minus whatever couldn't duck on Intel/AMD patents. ). A small, relatively fixed cost team to keep the lights on for the translator is sufficient.
Second, the installed base inertia is completely different. The installed base is around 2-4x as big now. Yes they are selling M1's a pretty fast rate , but it is also a much bigger mountain to climb to displace most of the old stuff in the hands of active users (who are more prone to buy stuff).
Piled on top of that Apple completed the transition in about 18 months. 18 months in this time and Apple isn't even close to be finishing off the product transition. ( Pandemic and other factors out of their control contribute to that. But probably also wasn't the original plan either. Apple probably meant two years when they said two years this time given the inertia hill they needed to climb and that were not selecting a SoC vendor that had a fully complete spectrum of offerings on day 0. Intel had more CPU models to pick from on day 0 than Apple even needed. )
Here probably closer to 2.25-2.5 years to actually finishing. That means going to be selling Intel models longer window than sold "new" PPC models (that inertia hill at least not shrinking , if not technically growing larger. )
Third, you are missing dates on that 10.4 (2005) , 10.5 ( 2007) . 10.6 (2009) releases. Some significant stuff happened around that time. iPhone OS v1 ( 2007) , v2 (2008) , v3 ( 2009) the first iPad was on "iPhone OS 3". [ and iOS 4 was coming 2010 ] . So these little "small" things called the iPhone and iPad showed up. Apple pulled resources off of macOS X and push them onto iPhone/iOS around the same time stripping PPC port of resources. Porting to Arm took on higher priority than even to evolving x86_64. The shared code Arm work did just as much, if not likely more, to kill off the PPC port as the x86_64 one did.
This transition is exactly the opposite. macOS is moving to where Apple already speeding most of the kernel, GPU driver, and special fix function unit library money anyway. Many of the left over Intel models have a T2 chip in them ( MP 2013 doesn't , but also not large base burning money either). Instead of going 3-wide on the port teams they are at the same two they were before the transition. There are "more savings" at some point by dumping Intel CPU, AMD GPU, etc at point, but it clearly isn't costing more to go wider/broader in platform coverage.
The performance upgrades to moving over are large enough that many are moving over voluntarily rather there being some deep seated requirement to forcibly herd them over to buying upgrades they don't necessarily want.
Fourth, Apple has a major subscription revenue flow now. If a Intel Mac OS users is steadily paying several app store subscriptions then Apple is still making money. Why run the risk of prematurely cutting them off then respond by cancelling the subscription as a "revenge" move?
Also the way Apple does accounting for upgrades has changed. Apple charges for upgrades up front when buy the system. They recognize the revenue when they ship the future upgrade ( one reason always eager to ship some half-baked macOS/iOS/iPadOS v0 version out at the end of calendar Q3 ). It is completely unlike trying to elicit new money out of customers for an OS upgrade when they know their platform is deprecated. (the number of people buying upgrades was already relative poor and slow. It would have been even more slower and poorer for "end of the road" PPC systems. ) Under the current system, Apple already has to money and can just miserly spend it on a "just enough" effort to ship something to keep people happy in the ecosystem.
Apple got no where near the time to complete upgrade transition times they do now back on 10.4-10.6 era.
Soba said:
The 2013 Mac Pro design was sold as the current model for an extremely long time and wasn't discontinued until December 2019. I don't see Apple dropping current macOS support for this system less than 3 years after it was discontinued—especially considering some of these systems will still be supported by the standard 3 years of AppleCare.
3 years is way, way , way off any substantive track record that Apple has put together over the last two decades.
Soba said:
Mini-rant: Apple serves no one's interests but their own with such a lack of communication on this topic. They should create and publish formal support lifecycle policies so customers can actually plan their business decisions.
Apple could do better, but also really not much of an incentive to do better when they do publish explicit support lifecycle information and people largely ignore it. Apple clearly documents the hardware support lifecycle. Apple also clearly communicates that they don't see a large chasm between 'hardware' and 'software' . They are primarily out to sell 'systems' as opposed to modular , largely decoupled components.
The large disconnect is mostly driven by folks outside of Apple insisting that the software has to be on some radically different track than the hardware. It isn't. Apple basically says it isn't. They pretty consistently don't treat it that way either.
The major problem with Apple's policy is that is built on "doing" rather than "talking". Intel trotting out 10nm roadmaps for years. Any business who bet the farm on those early roadmaps probably ran into trouble. It isn't about "talking the talking"... "walking the walking" ( backing up talk is what matters more if trying to coordinate multiple entities in trying to get something done).
The high end Mac workstation class products are problem more so because the disappear into a "do nothing" mode for long stretches. If 'actions' are the base of talking/communicating then doing nothing means saying nothing.
Apple could write up a chart to explain the n , n-1 , n-2 trail off on security upgrades but if consistently do it for decades ... what is better the consistent action or lots of promises may or may not kept?
Does Apple have wiggle room to not do software they have not promised to do? Yes. Is there not enough info to construct a plausible 2-6 year plan? No.
