On Tue, 2004-05-04 at 13:53, Lamar Owen wrote: > > What you probably saw was a 16*K* byte memory, the boards from the late > > 60s were typically around 16K bits, 8 boards to a byte, and about as big > > on a side as a tower case with the small scale integration driver chips > > around the sides of the core fabric. A 16MB machine would have been > > gargantuan, indeed; I would guess 100 or more 19inch racks, 8 feet tall, > > and probably costing on the order of $50M in current dollars. > > I have an 8Kx12 core plane here from a PDP-8. Fairchild 7446 drivers and a > discrete sense amplifier driving a 74128. The core plane itself is about 5x5 > inches, and the card is 15 1/2 x 7 3/4 inches. I don't have the omnibus card > (H212) the plane card attached to; I threw it away long ago, just keeping the > core board itself. The plane is DEC H219-2779, dated 6/2/1975. For those > PDP-8 buffs out there, no, I don't have the CPU any more; I sold it back in > 1990 for a good price... > > As a trivia data point, elderly seamtresses were employed in the manufacture > of some core stacks, since these stacks were hand-sewn. Elderly seamtresses > were preferred since they could 'stitch' a stack with fewer errors than a > machine could. 'Has anyone seen my knitting?' The accuracy of experienced > seamstresses in doing crochet or knitting (or quilting, etc) is legendary. > > Another interesting thing about core: it is nonvolatile (but dynamic). While > every read of the core required the data to be written back (the read was > destructive), core doesn't lose its mind at power-off. This core plane still > has a program loaded: you can see the pattern due to magnetic attraction and > replusion of adjacent bits. And writing was just a special case of reading. > > Hmmm, 16 MB of H219 stacks would occupy 16x8x2x2048x2 cubic inches, just for > the cards. That would be 1,048,576 cubic inches. The standard 19 inch EIA > rack at 8 feet tall that is 26 inches deep (a fairly deep rack) can contain > roughly 47,424 cubic inches, yielding 22.1 racks. 23 racks would be enough > for the core and the CPU, but then you'd need another 2 racks for the power > supplies. So about one quarter of your estimate. > > For more information on PDP8 stuff, see both pdp8.org and pdp8.net, where you > can even download the manuals... See > http://highgate.comm.sfu.ca/~djg/htdocs/cgi-bin/tifftopdf.cgi/mm8aa.pdf?pages=1-5&loc=newstuff > for a pdf of the schematics for this module. > -- > Lamar Owen > Director of Information Technology > Pisgah Astronomical Research Institute > 1 PARI Drive > Rosman, NC 28772 > (828)862-5554 > www.pari.edu My father worked at Goddard Space Flight Center in the 60's and 70's. According to him and some other sources the people who removed the "old IBM 360" made more money on the gold plate (from connectors and wire wrap) and scrap copper (from the cooling system) than on the contract to remove the system. -- jludwig <wralphie@xxxxxxxxxxx>
