>>865347 (OP)
>systems in cars so laggy and their monitors have such awful refresh
They rely on USB 2.0 to move the data around. Some of the limitations of USB2.0 speeds have been overcome by using two ports per module - one for "upstream" and one for "downstream" - but you're still limited by USB2.0 speeds. Anybody that's tried to watch a 1080p movie saved on external drive connected to their laptop via a USB2.0 connection can tell you it doesn't work very well unless the video is one of these shitty 2GB DVD rips.
>>865357
>reliable computer that was barely acceptable performance 5 years ago in a car today
That's the main answer to the question. R&D on all these stupid little media modules lasts 3-5 years.
The initial design phase can last 6mos to one year, then a solid year of testing (design validation, product validation), then a year of in-car testing. Automakers are a year ahead in their production. E.g., it's the start of 2018; 2019 model cars are about to go into limited production - 3K or 10K vehicles that are given to select automaker employees. These are often the cars you see going down the road wrapped in zebra-striped vinyl wrap or those black pleather bra type deals. These employees log miles and keep records of all the quirks and bugs that they come across in operation, which, depending on severity, will either get ironed out by the R&D teams before the full production run, but sometimes get left for the customer to get annoyed by.
One fairly recent example was a bug in a certain brand of vehicle. Every so often, at startup, the head unit would not recognize the SD card in the media module that had all the GPS maps stored on it. So, when the vehicle was first started, the nav screen would be black with the message, "No map detected," or some such. Oddly enough, this bug has existed in this exact form since the 2015 model. Only until last year when the 2018 models were in limited production did anybody start to care about it. It ended up being the head unit's fault. It wasn't giving enough time for the media module to respond when queried on the size of the installed card, so it defaulted to "no card". It was allowing 16µs instead of 1.6ms for the response - a misplaced decimal point.
>>865405
>They get the cheapest parts they could find
That's not at all true. The parts are fine when they're purchased. As noted above, they're two or three or five years out of date by the time they end up in front of you. They're cheap in the sense that, the module maker sent RFQs to various IC manufacturers, and they got a reasonable price on a minimum purchase quantity over however many years. They got a part for 10¢ that you have to buy off Digikey or Mouser for $10, but they have to buy a half million of them by 2020.
ICs are developed at a phenomenal rate these days. Manufacturers do limited runs of each successive design. They may make a million parts, for example, and everybody and their brother wants 5k or 30k or 100k. The part is discontinued, listed as obsolete, and production of the next iteration begins almost immediately. Whatever doesn't get sold ends up going for pennies on the thousand to warehouses that specialize in selling obsolete/discontinues parts (who may or may not also end up with the schematics and the dies for making more if/when the time comes).
>>865473
>>865478
They all have the same hardware behind them. Generally speaking, the brains of the operation is a 32-bit microcomputer with integrated CAN, USB, UART, ADC and DAC, and LCD drivers. Different auto OEMs have different requirements as to form factor, overall size, current delivery (for usb charging), etc., but the units are all designed and created by the same people in the same place at the same time.
>>865830
>two VMs running on one piece of silicon
Lies.