Rumor around here is that when it came time to get rid of the old Burroughs, Burroughs sold a new machine to the university that bid on our old machine for less than what they had bid on the old machine. Everyone here assumed it was because Burroughs was pissed that we had gone with big VAXen; but perhaps that was just the way Burroughs worked. http://www.ptsc.com/ describes ShBoom, a stack machine designed by Chuck Moore (aka the inventor of Forth). The theory behind ShBoom, and his more recent stack machine, is that if you use a wide word (in this case 32 bits) you can fetch a whole bunch of tiny instructions at once then blaze through them. In the case of ShBoom, the core runs at 100Mhz while the external bus runs slower. Since each opcode is a byte, he can fetch four opcodes at once with the slower bus rate yet keep the fast internal core fed. Chuck Moore also has some newer designs. F21 and P21, for instance. These are 21-bit machines with (IIRC) a 20-bit external bus. Each opcode is five bits, so (again) he's fetching four instructions with each external bus transaction. There was a big argument on comp.arch recently about just how hard it is to go superscalar on a stack machine. IIRC, a Unisys A series engineer was heavily involved in that discussion. Perhaps interested folks can dredge it up using http://www.dejanews.com/. I don't recall whether there was a conclusion to the argument (is there ever a conclusion on usenet?); a lot of the discussion was over my head. Roger ivie ivie(a)cc.usu.edu