This is a relative comparison not absolute, and trying to be nontechnical about a technical topic is a minefield, but vaguely speaking for many decades ARM generally didn't worry as much about backwards compatibility and backwards binary compatibility and was pretty chill about every tiny little revision having tiny little improvements such that like V8.3-A added the "famous" FJCVTZS opcode (google it, its funny).

Intel in comparison likes backwards compatibility. Your new motherboard probably could boot MSDOS 1.0 if not CP/M, plus or minus the need for a floppy controller LOL. Also Intel stereotypically likes major upgrade "sets", so Intel likes to release a giant pack of new features like Streaming SIMD Extensions (aka SSE) rather than how ARM will toss out literally one new opcode in a release.

Over the long term (decades) it seems the ARM strategy has been VASTLY more successful.

Thanks for that really interesting summary!

Regarding backwards compatibility, is this basically that certain companies developed new products so quickly that there was no point (or value) to supporting old hardware (for the company, or desired by users)?

And therefore delay + overheads of waiting for compatible new CPUs was worse than the value of what they did release each time? Plus the generally worse heat/power utilizing designs?

It is of course a general stereotype in that its useful for prediction but its not as solid as, say, the law of gravity, but ARM comes from a software as a project world where you write the code for a specific piece of hardware with a defined and short production life, whereas Intel desktop PC comes from a software as a product world where you expect your word processor to keep working even if you replace your motherboard or swap in a faster CPU. One has a world of permanently sealed cases and firmware thats never updated (with unfortunate security results) whereas the other at least in theory has user-serviceable parts inside a screwed together case. Ironically that sounds a lot like MAC vs PC so it makes sense that MAC would go ARM.

Basically the only advantage is simplified instruction decoding, which it has in common with other RISCs. It has some other nice properties too, I think it really shows that it has taken lessons learned from other RISC architectures and struck a good balance. But the reason for the switch isn't how nice ARM is, the reason is that Intel is struggling with its manufacturing process and Apple wants to be in control of its own destiny.

Apple also want to add their own narrow-purpose chips (like media encoders and what have you).

It's more than instruction decoding, though admittedly that's a big part. There's also memory ordering (tho it's debatable whether that's problematic), the number of logical registers (particularly important for JITs) and messy dependencies like partial flag updates.

ARM doesn't sell chips, they sell IP. E.g.: their ISA (instruction set) or soft cores (designs that you can integrate in your own chip and then fabricate).

There's nothing special in the chips themselves, it's the business model what's quite different from Intel's.

Historically, ARM (and their licensees) have designed their cores targeting power efficiency over everything else, but there's nothing that prevents Apple or any other licensee from designing a power hungry ARM CPU.

From what I remember biggest difference is that Intel designs and produces chips. Where ARM only designs chips and everyone can build their own chips based on the design of course paying license fees. So Broadcom, Qualcomm and others are building ARM chips and would not be surprised if Samsung and Apple soon build their own.

Edit: Yes Samsung has something called Exynos, Cortex is from Broadcom, Qualcomm has Snapdragon. Basically there is much more competition in ARM.