OT: how big would it be?

And how does one implment the Acc, DE, HL register pairs serially... Still needs the flipflops. Then there is the matter of the 7level stack x14bits in the 8008 or the smaller one of the 4004. That's a lot of memory on the chip in the form of registers/FFs. In the parallel version that stack with it's 7x14 and pointer register is some 1/5th the total chips area and ignoring data paths would be a lot of FFs. Older serial machines did their best to avoid this kind of trap as did the PDP8 series as it's register intensive. The ALU in reality is if anything the easy part as a result even though it's not a simple array. While the PDP-8/S is serial arithmetic there are large sections that are still parallel like the memory. I may add the 8/S was not so much size reduction but cost reduction with side effects of some incompatability with other -8s and signigficant speed reduction. Look at the MIT/lincoln labs TX-1 and TX-2 for some indications. Allison