Re: It isn't enough for the ARM to match i5 perofrmance
As others already said, Apple could quite trivially have the Mac compiler build fat binaries - x86 and ARM64 - just like they did with the PPC transition. You'd only need to emulate old code, anyone producing Mac applications that were performance sensitive like Photoshop would be first on the block with those fat binaries.
The problem isn't Mac software, that's easy. The problem is Windows. A lot of Mac buyers run Windows at least occasionally, and the A10 is simply not an option there because when they are running Windows they're running x86 Windows and old school x86 Windows apps, not the ARM version (is that even supported on Windows 10 or did Microsoft give that up as the abject failure it was?)
It does seem like Apple has something in mind for their SoCs beyond phones/tablets, because they are really pushing the performance. The A9 is still faster than any SoC you can get in the Android world, and the A11 will probably be around the corner by the time they finally exceed A9 performance let alone A10. I'm leaning towards Apple having some performance target in mind for VR that they want to hit before releasing something. Sort of like how they started development on the iPad in 2002. They could have released it years earlier - in fact they did release a small version of it called the iPhone in 2007 - but waited until 2010 when they reached the performance and battery life targets they had set in an acceptable size/weight.
Based on TSMC's roadmap they should be able to get another 30-40% with the A11 next year, and probably an overall doubling (i.e. Geekbench 4 single threaded score of ~7000) by the A13 in 2019. That's sure not something anyone needs for their smartphone.
Biting the hand that feeds IT
