Apple’s new M1 processors have shaken up the entire CPU landscape. Apple has used Intel for more than a decade, but these new in-house chips, based on their designs for iPad and iPhone processors, are powerful enough to run a laptop.
Well, that’s an understatement, they are in fact some of the fastest laptop processors we’ve ever seen, and they deliver this performance with incredibly low power consumption. The M1 chip is so impressive, it’s been referred to as “black magic”. But it’s not magic, it’s physics, architecture and electricity. There has to be a reason a chip is fast, and for this episode of our explainer show Upscaled, we set out to learn what it is.
One feature of the M1 is that it’s a RISC-based chip, designed using ARM’s instruction set. This means it has more in common with the chips commonly found in smartphones and tablets than those typically used in desktops or laptops. This already makes the M1 unusual, but it’s not unique. Microsoft’s Surface Pro X also uses a custom ARM chip, co-designed with Qualcomm, called the SQ2. On the face, these processors are similar: both the SQ2 and M1 use a 4+4 core design with a mix of high and low-powered cores, they both have 16GB RAM, and they’re designed as an SoC or “system-on-a-chip” meaning all the components you’d typically find on a computer are build together into one chip.
There the similarities end. Not only is the M1 much faster than the SQ2, Apple has pulled off a pretty amazing feat when it comes to emulation. Their Rosetta 2 emulator allows you to run pretty much any software that works on an Intel mac, for a slight loss of performance. Windows also has an emulator, but in our experience it’s slow, buggy, and restricted to 32-bit apps (for now).
Traditionally, there have been actual hardware reasons why running Intel or AMD software on an ARM chip is slow, which makes Apple’s achievement all the more impressive. In this episode we test the M1 versus the SQ2 and an Intel Macbook Pro, and dig into just how Apple achieved such good emulator performance. In the next episode, we’ll get into the architecture itself to break down how the M1 is just so dang fast.