After Apple’s M1, are the days of x86 over?


The M1. It’s freaking insane.

That single-threaded performance don’t lie.


Could be better in multi-threading.

Mind you the XEON E5–2687W v3 is a 10-core server CPU with 20 threads. The Ryzen 7 1700—which I bought on release—has 8 cores and 16 threads. These are old CPUs though.

A “fairer” comparison would be against a Ryzen 5600X, that has 91% of the M1’s single-threaded score and is within 0.07% of the M1’s multi-threaded score per thread (not core).

—Oh great, it’s just a Ryzen with Apple branding. One would say.

But no! That’s not the point!


This is part of a trend we’ve seen coming for a while. Apple’s not even the first company to make a good ARM chip for a computer.

Anyone remember the AMD Opteron A1100 from 2016?

The logic has always been the same. ARM is very efficient, but not very powerful. But certain server processes require just that. Amazon is not stupid, so they did the same thing with the Graviton chips—though the first ones were kind of shit.

—But hey, it’s coming! ARM is just around the corner. It’ll be awesome. We said.

And then this M1 came.

It’s Ryzen-level performance on four times less TDP and roughly a quarter of the juice.

The M1 goes about 30W on a multi-threaded load. A 5600X will go up to 140[1]W fully loaded.

It’s no longer around the corner. This is here.

I’d be hugely surprised if AMD is not working on a revival of their Opteron ARM ambitions.

It will take some time for compatibility to catch up and SDKs/frameworks to be mature enough, but the incentives are now here. This performance per watt is insane. Much longer battery lives, much cheaper servers (Graviton), more competition in the CPU space. Win, win, win.

It’s not the end for x86/64 but it surely is a new beginning as things heat up.