I'm thinking of buying a pc maybe this year, with a view to NEVER having to upgrade, just replace parts. My last 2 boxes were laptops, the next is a desktop. Does anyone have any insights on this?
https://www.digitaltrends.com/comput...nm-2019-zen-2/

To me, as a hardware head, 7nm fab is a fairly important development. Back in the earliest days, there was 10^-6m (=10,000nm) fab, and chips were struggling to do 1Mhz. The fabrication improved by leaps and bounds https://en.wikipedia.org/wiki/Semico...ce_fabrication (This list, incidentally, doesn't tally exactly with wafer fab sizes as I read them from some datasheets over those decades).

Now it seems AMD has the jump on Intel in wafer fab, having use of 7nm plant. AMD traditionally made better use if it's traditionally larger wafer fab size in an effort to be competitive. It seems the leader in that race is about to change. AMD will be 7nm this year, and apparently Intel aren't there yet. With that inhouse expertise (optimizing a given fab size), AMD should have the jump on Intel.

The limit of achievable miniaturization appears to be 5nm. Things like Quantum effects, & molecule size make smaller fab unlikely. 7nm is a huge achievement in itself. If someone sneezes nearby, the consequences could be dire.
Ireland is well away from earthquake zones. Intel built Fab 14, a 64nm plant here in Newbridge, Ireland, and it made trash, until it was discovered the plant was being upset by micro-earthquakes. The solution was to dig a deep trench all around the plant to isolate it from these tremors. Entrance is now by bridge.

There is no such thing as "never having to upgrade" as long as Moore's Law still holds. Eventually your CPU will be too old, and the CPU and motherboard are your computer; everything else is peripherals. Even if the process doesn't change for silicon, they could move to other materials or resort to more cores, larger dies, offloading certain functions to a quantum CPU, etc. to maintain Moore's Law as long as possible. Remember that the entire computer industry depends on being able to sell your faster and faster hardware forever. They will keep the gravy train rolling at all costs.

You're right. That statement is based on 2 factors

1. I don't believe Moore's Law still holds. I believe it's been slowing since 2016 or thereabouts, and that 5nm is a real limit on wafer fab. It may be breached, but the gain from each successive advance is diminishing. Logic dictates that there is a plateau to performance, and that we are approaching it.Further,
2. There is my age to consider. I'm not young, and not in condition to exercise sufficiently. So the odds are against my continuing indefinitely

I also feel quantum computing is another glorious non starter in the world of technology. I've seen the pattern before: Loads of hype, loads of visionary promises for years on end, greatly limited experimental 'proofs of concept', billions spent. It fits the Classic definition of a complex systemTherefore I foresee that a decent system bought this year or next might have 20 years in it. My computing requirements are not that complex, and I don't expect to have any great interest in proceedings beyond that date