There is no one computer or chip that controls a car. There are hundreds of chips involved from your USB charger to your radio to your engine contol module and so on which contains numerous chips. And yes, many are carryover from year to year but new models always update to newer chips when they can. Old tech doesn’t drive new sales.
I had completely forgotten about the Olympics !
Watching tear down videos of Teslas it was eye opening how many computers are involved. The body control stuff is pretty much just off the shelf parts from regular auto industry suppliers, so applies to many new cars
If I remember correctly the steering wheel computer was based on a PowerPC core roughly equivalent to a 90s Mac. Of course an entire Quadra 605 has been shrunk to a single chip, which mostly just tells other computers about the steering wheel’s position. There were other computers in the steering wheel to handle the buttons and airbag.
Multiply that one circuit board with multiple components by the brake pedal, accelerator pedal, window controls, door controls, seat controls, etc. and that’s a whole bunch of silicon in any new car.
I’m seriously considering re-engining my 2000 Saturn because of this. It doesn’t even have an air mass sensor. Manual Windows, manual air/heat controls. No PMC computer.
It’s about as basic a fuel injected engine as you can get in this century.
EVs, as I understand it, have even more chips than ICEVs. There’s a whole bunch involved in battery/power control which is more than what’s needed for fuel injection and other internal combustion stuff. That doesn’t really sound right, but that’s what I’ve read.
As a point of comparison here, Renesas is a major manufacturer of automotive microcontrollers. Their highest-end line is the RH850:
They brag about being manufactured on 40nm:
The RH850 Family is offered in a Renesas 40nm process, an industry first
40nm is roughly six generations behind the state of the art. Process technology has basically two generations for each halving of process dimensions, and the nodes are roughly 40->28->20->14->10->7->5 (as an aside, the “nanometer” lengths have become disconnected from any physical dimensions, and instead reflect the equivalent increase in density) (as a further aside, I’m mixing up standard nodes with half-nodes here, but that’s because the ITRS itself squeezed a half-node offset in there…).
So not only is the process ancient and limited to a handful of old fabs, the fact that it has such poor density means that they get far fewer chips per wafer than they would otherwise. The latest process fits about 64x the number of transistors per square millimeter as the 40nm process Renesas is using! Ok, it’s not quite that simple since analog components and some other things don’t scale at the same rate, but nevertheless they’d produce way more chips per wafer with a denser process.