Using an M2 SSD instead of soldering the chip on board has more implications: PCB gets physically larger, and takes more power or has less performance talking to the SSD. Heat transfer is also worse. I completely understand why they go for a soldered SSD chip.
One way of true environmentally-friendly innovation could have been to find a way to attach the SSD chip so that a user could safely replace it, though, with little additional space.
> PCB gets physically larger, and takes more power or has less performance talking to the SSD. Heat transfer is also worse.
I don't think these are real problems. The M.2 device would take up space you could have used for the PCB, but then you would have had to use that PCB space for the chips that are on the SSD.
The SSDs in current Macbooks do around 3GB/s. NVMe Gen5 does 14GB/s. The speed-of-light latency from any kind of connector is going to be totally irrelevant compared to the latency of the flash controller itself. There is no performance concern. Power is the same; when idle the link goes to sleep, when in use the connector is negligible compared to the device itself.
Heat transfer doesn't even seem related. If you want to improve heat transfer from the SSD then you put it into thermal contact with a heatsink or the chassis, which you can do regardless of whether it's M.2 or not.
> One way of true environmentally-friendly innovation could have been to find a way to attach the SSD chip so that a user could safely replace it, though, with little additional space.
The only real space requirement is the size of the connector itself, which is on the order of 50 square mm in a PCB which in a 12" laptop is some tens of thousands of square mm. <0.5% is "little additional space" to begin with.
Obviously you could design a connector which is even smaller, but the premise would have to be that that's even a real problem.
M.2 devices save space on a PCB: yes, the connector itself takes some room, but the alternative is sticking all those chips on the PCB itself, and those chips take up more space (just look at any M.2 NVMe drive). The M.2 form factor is moving those things off the main PCB, and onto a daughterboard that usually sits directly on top of it and parallel to it.
The idea that a PCB gets larger with an M.2 slot is truly insane.
Even on desktop motherboards, the space under a M.2 slot is usually nearly empty. On laptop motherboards, it is almost always completely empty save for possibly a thermal pad. Some laptops position the M.2 slot to have the SSD extend beyond the edge of the motherboard. But in either case, laptops are not reducing PCB footprint by using M.2 SSDs, because nothing gets stacked under the SSD; that space is reserved for the SSD.
For an iMac, though, they have a bigger thermal envelope and no battery. It seems more reasonable. Apple even did the software engineering necessary to support the Mac Pro.
That would be Apple’s counter-argument right there. Want Linux on your M1? Get a Mac, not an iPad. Want swappable storage? Get a Mac Pro, not an iMac.
That is some gorgeous PCB work. And goes to show that JLCPCB (you can tell from the order number) is perfectly usable for applications needing a controller impedance stackup.
One way of true environmentally-friendly innovation could have been to find a way to attach the SSD chip so that a user could safely replace it, though, with little additional space.