It's interesting to see this done with so few ICs. Early Z80 machines had a much higher part count, because a lot of external support was needed to talk to anything. Today we have the "system on a chip", with lots of onboard peripherals, but a Z80 is just a CPU.
The predecessor to the "system on a chip" was the semi-custom "junk chip", which contained unrelated stuff the standard chips didn't have. He may be using some of those to get the parts count down. Can't tell; schematic too small.
Way back I used to hand solder Z80 embedded controller board prototypes, and they never needed much in the way of chippery.
In an S100 system the biggest extras were the video (formerly VDU) cards, and the memory cards.
Both were limited by the RAM of the day, because RAM came in tiny sizes like 1 bit x 16k dynamic or 2k x 8 static. So you needed a lot of chips on a separate board for what was really not much RAM at all.
Also, money. 64k boards cost over $1000, which is > $3k in modern money.
Now you can get 8 bit x 256k static RAM on a single chip, so you can park that next to the processor without a separate card. With a bit of glue you can even share it with the graphics hardware, eliminating any need for separate graphics RAM.
http://monster6502.com/ as seen on HN https://news.ycombinator.com/item?id=11703596
Good point but I didn't get it quite right. What is your definition of "things that compute"? I am curious.
A single transistor definitely doesn't compute. A single gate? Eh, not really.
A design with a couple latches and an adder that can perform SUBLEQ? Yeah that computes if you attach a ram.
A non-turing-complete pipeline of math functions? Yes, that computes, though why make that when it takes more gates than a microprocessor needs.
Point contact germanium BJT:
Etched silicon FET:
The rusty razorblade + pin point contact diode doesn't need anything special because it's a Schottky diode, with a metal/N-type semiconductor junction. If you probe lots of places you can find an N-type region where the dopant is random impurities. But the transistor needs both N and P-type regions, so I don't see any way to build one with common junk. Jeri's point contact example works by diffusing phosphorus from phosphor bronze into commercially produced N-type germanium using a high current to heat it. Phosphor bronze is arguably household junk but I don't think germanium diodes are.
Interesting to see a little bit of history and how game writers really were working from home on an idea even in the early days.