Item — HN Reader

>> The problem seems much more software based than hardware based.

- During transition from Motorola's 68k series of chips to IBM/Motorola PowerPC processors, Apple produced classic MacOS versions that could run on either platform.

- During the transition from PowerPC to Intel, Apple produced Mac OS X versions that could run on either platform.

- In 2005, Apple shipped special Intel build machines for devs only.

I think there is a method to this madness and despite all the flaws we currently observe, which are orthogonal to the transition, I trust them to get it right, if and when it happens.

1 comments

The earlier transitions involved going from a less powerful cpu to a more powerful cpu so emulation was an option. Now, at best you are going to an equivalent cpu and at worst a less powerful cpu.

Apple has some pretty decent IC design engineers... And ARM has fewer legacy corner cases to support and a more dense simpler to decode instruction set. Apple also has high margins so can afford to have big silicon dies with lots of logic and cache.

Given all that, I think it's totally possible Apple could release a CPU with more raw performance and performance per watt than Intel/AMD.

That extra performance would allow possibly emulation of x86. The emulation world is much more advanced than it was 20 years ago - dynamic recompilation could easily lead to a 50% slowdown rather than a 5000% slowdown that interpreting might be.

Do you have a source for that? The thesis of this article is that it is a more powerful CPU, in terms of performance-per-watt, which is the same metric that spawned the PowerPC-to-Intel transition. The author pointed to specific ARM chips that exist today which "consume less power, about 210 watts, than a competing Xeon CPU needing as much as 400 wats, for about the same amount of computing".

Depends how seamlessly they can get devs to re-compile. ARM having to emulate x86 would be definitely slower.

There is no reason to believe the per-watt performance of actual ARM code won't outperform equivalent x86 chips. i.e. an A(x) Macbook Air replacement could easily outperform the Intel variant at the same power envelope, based on the observed performance of the iPad pro my wife uses.