Why the Apple M1 was different: three chips became one

hurraybro — Wed, 11 Nov 2026 18:03:06 +0000

In 2020, Apple's M1 changed how laptop chips are built, and Intel and AMD later started doing the same. Here is what actually changed. I also made an animated video of this (embedded below).

Three parts become one

A powerful 2020 laptop had the CPU in one place, its memory a few centimetres away, and graphics on a separate NVIDIA chip with its own memory. Every time the CPU talked to the RAM or the GPU, the data went over long wires.

The M1 put all three in one package: the CPU and GPU share a single silicon die, and the RAM sits right beside it.

What matters in a laptop

It was faster, but that is not what made it special. Day to day you feel weight, battery and fan noise. The M1 Air lasts up to 18 hours, weighs 1.29 kg and has no fan.

Distance

Fetching a number from memory costs hundreds of times more energy than doing math on it, so a big part of a laptop's power goes into moving data. The M1 cut that trip from centimetres to millimetres. A wire roughly ten times shorter takes roughly ten times less energy to drive. Less energy per trip means less heat, little enough that the Air does not need a fan.

Copying

With a separate GPU, the CPU and GPU each have their own memory, so data is copied across before the GPU can work, and the GPU waits. On the M1 there is one memory, so nothing moves.

Why didn't everyone do this?

A GPU needs a huge amount of data every second, and each wire carries one bit at a time. The M1 uses 128 data wires, and a fast GPU wants 512. With the clock, address and control wires, that is close to 800 wires. They all have to match in length and be spaced apart, which takes a lot of board that a thin laptop does not have.

So PC makers gave the GPU its own memory right around the GPU chip. That brings the copying back, adds a second chip with its own power and heat, and needs heat pipes, bigger fans and a bigger battery. That is why gaming laptops weigh two to three kilos.

Apple's memory was already millimetres from the chip, inside the package, so it just added more wires. A year later the M1 Max had 512, and the CPU and GPU still shared one memory.

Small cores

Big and small cores were not a new idea: phones had them for years, and Apple had used them in the iPhone since 2016. The M1 brought them to the laptop, with four big cores for heavy work and four small cores for everything else.

A performance core spends many transistors looking hundreds of instructions ahead and running them out of order. An efficiency core looks only a short way ahead, and it runs at a slower clock. A slower clock lets the voltage come down, and power falls with the square of the voltage. So macOS sends background work to the small cores and switches the big ones off until you need them.

I would like corrections if anything above is wrong. The full animated explanation is the video above: https://youtu.be/wGC2lMzVsqA