700 tubes! That's daunting. It took whole departments a year to assemble machines like that, and with budgets. (Though they didnt have gigasample/100MHz software scopes :-) I'm awed that he got that much together. I figure I can get 2500 instructions/second for 70 tubes, or 100 active elements, not including the store (aka memory) which is just N x $ (eg. how much memory can you afford?). Lots of silicon diodes (not cheating at all, very 1958). 2500 not very good instructions, I might add.