The Continuing Progress of ARM
(and the destruction of Intel)
10 months ago I wrote about the incredible story of ARM. The first ‘desktop class’ ARM processor from Apple is called Apple M1.
And it’s absolutely stonking. I think Apple threw everything and the kitchen sink at this first Apple-designed CPU for their mainstream laptops and desktops. It’s a technical tour-de-force.
On top of that they have included another astonishing bit of software technology they call Rosetta 2. This enables the new Apple machines to run Intel x86 code, so these new computers can run software designed for older Intel machines.
Unlike Microsoft’s attempt at doing this (for their ARM based Windows tablets), which does an interpreted emulation of x86 instruction set, the Apple does a one-time cross compile of x86 code into ARM code. This is brilliant and ‘just works’ to use that hackneyed phrase. Combined with the shear speed of the Apple ARM M1 chip, there are benchmarks showing the M1 is faster running some x86 programs than actual x86 processors. Truly astounding.
For the hardcode geek there’s a great in-depth article — and I mean really in-depth at AnandTech. (Although a bit old now — pre M1.)
If you want to understand why Apple had no choice but to move to their own ARM CPU, there’s one graph from the AnandTech review that shows the progress of Apple’s ARM CPU over the last 6 generations, compared to Intel’s:
The purple line is the increase in Apple CPU performance, and blue line is the increase in Intel CPUs over the last 5 years or so (and it doesn’t include the new M1 in that either! That would, literally, be off the chart.)
It’s simple: Intel’s CPUs are simply not progressing as fast as Apple’s CPU. That’s utterly damning for Intel. This marks the clearest indication yet, in my opinion, of the end-of-days for the Intel architecture. This is not just the ‘beginning of the end’ of Intel. That was the first iPhone ten years ago (Intel completely missed that one), but this is the near the end-of-the end for Intel.
And the reason why this is so goes right back to the architecture. As my original article said — the ARM was designed from day one to be simple, but fast. As anyone who has written assembly language for both processors will know the Intel (and AMD) x86 architecture is a nightmare of complexity.
Less is more. Again.
Maybe this is the kick up the butt that Intel need. As a first move the old CEO is gone and now a technical engineer in charge — that’s a great step IMO.
But you can bet that Microsoft, and other people who develop ARM based systems are scrambling to compete with Apple. And the one way they can realistically do this is via ARM. Not just in terms of raw horse-power but, crucially, in terms of power-usage. I have an Apple MacBook Air, and it’s not just the shear speed and slick ‘smoothness’ but the that it can do this and never get warm, is silent with no fan, and has truly amazing battery life.
I’m no devoted Apple fan-boy. I’ve been developing Windows software for 25 years or more. MacOS sucks in some critical ways, compared to Windows. And most of it is around usability — the area most people (and certainly Apple) think they master. Well, they do not. Maybe I’ll describe why one day.
But in term of hardware, and in terms of CPU and chip design, what Apple has done is breathtaking.
