Originally Posted by Blastdoor
A challenge with replacing Intel with an A# chip in Macs is that backwards compatibility with Intel programs will require emulation, and emulation speeds might not be that great.
You are assuming that Apple would even consider emulation of an Intel processor. They are far more likely to offer up a ARM based laptop, tablet or something in between, that can run iOS apps in a Window. This way you get access to a huge library of apps yet can still maintain a conventional operation of the device. I would imagine that this machine would be unlike anything currently shipping from Apple or anybody else.
You may wonder who would be willing to pitch out hard earned cash for such a machine. Well I'm about to raise my hand because if Apple did it right I'd be very interested. Why because a great deal of what I want to do is covered buy current apps. As long as I can run Python and a CLang based development environment on the machine I'd be happy. Well it is a little more than that as I'd want access to some of the UNIX goodies.
In the past when Apple has changed CPU architectures (680x0 to PPC; PPC to Intel), the architecture Apple was moving to was substantially more powerful than the one it was leaving, meaning that the emulation penalty felt by end-users wasn't too horrible. But if Apple moves from Intel to their own chips in the Mac, it will be more for reasons of cost than performance, meaning that there won't be much of a performance gain to hide the emulation penalty.
Performance gain? I'm not expecting that either though Apple could surprise us. ARMs 64 bit architecture isn't really all that bad and combined with other acceleration hardware will deliver much of what many users want. As to Apple surprising us, people need to remember that when they picker up PA Semi they picked up a well know team of designers that that did excellent work combining low power and performance. If you look at the current A7 as a proof of concept that doesn't drift far away from the ARM design then they have plenty of room for a high performance A8.
Given all that, I think that if Apple were to use A# chips for something other than iDevices, the first use might be in Apple's own data centers. Presumably Apple is using Xeons in its servers, and Xeons are very expensive. Replacing Xeons with a server-oriented version of the A# chips (say, 8 cyclone cores on a die; no GPU; bigger cache; support for ECC RAM) could be a good way to save money, and there wouldn't be an emulation penalty, since presumably Apple would be in a position to recompile all relevant software for the new chip.
Apple doesn't have the volume to justify building its own ARM based server chip.
It is interesting that you mention recompiles. If Apple wanted to offer up an ARM based machine, running straight Mac OS all they would have to do is change the App stores terms to demand a recompile of all apps submitted. I think you would be surprised at the speed with which developers would adjust.
Once Apple has A# chips running on its servers, perhaps a new high-end service for Pro users could be introduced that would allow users to run software on Apple's servers. Initially the only people who would use such a service would be people who can compile their own code. But in time, the service could be expanded as more software is ported to A# chips. Gradually, we might reach a point where so much software has been recompiled to run on A# chips, that Apple could make the switch in Macs.
You seem to think that this would take a long time to do, I see just the opposite. I see developers falling all over each other in the rush to recompile for ARM and test. In less that two months I could see 80% of the Mac App Store supporting ARM.