Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!mcgill-vision!bloom-beacon!mintaka!think!zaphod.mps.ohio-state.edu!uwm.edu!psuvax1!rutgers!aramis.rutgers.edu!athos.rutgers.edu!nanotech From: dennis@cpac.washington.edu (Dennis Gentry) Newsgroups: sci.nanotech Subject: SPECULATION AND QUESTIONS Message-ID: Date: 9 Feb 90 02:37:40 GMT Sender: nanotech@athos.rutgers.edu Lines: 70 Approved: nanotech@aramis.rutgers.edu From: well!dduke@lll-crg.llnl.gov (David Allen Duke) Newsgroups: sci.nanotech Keywords: gate operations, virus, mental state Date: 8 Feb 90 01:31:54 GMT 2. Recent articles in _Foresight Update_ speculate about incredibly powerfull computer systems possible in an advanced nanotech society. I guess these speculations are based on the size and speed of imagined mechanical/molecular gate operations. Has anybody guessed what the difficulty would be of making mechanical gate ops interconnected, and how these gate ops would be powered? [As a computer architecture person myself, I've looked at it. The connectivity isn't a problem--remember that existing computers are basically planar; if you allow yourself 3 dimensions it's a lot easier. Power is a bit more out of my field, but suggestions tend to run to electrostatic motors of some kind. No one, by the way, seriously expects that nanocomputers will be actually built this way. Drexler holds them up as existence proofs; real nanocomputers will probably use highly involved quantum effects and run at speeds of about 100 terahertz instead of the roughly one megahertz the mechanical design could do.] Uh, doesn't the Drexler-designed rod logic gate switch in a few picoseconds? Since current CMOS logic gates switch in a few nanoseconds and current CMOS computer speeds are roughly 20-30 megahertz, wouldn't a rod logic computer (switching in a few picoseconds) run at roughly 20-30 gigahertz? (rather than the one megahertz you mention above). As in electronic computers, the connections between each "layer" of logic could be the same medium as the logic itself, i.e., carbine rods. Power is not difficult--rod logic can be (synchronously) powered by some kind of oscillating rod, e.g., a rod connected to a cam which is driven by a nano scale rotary motor (which Drexler has also designed). I think the hard part about these computers right now is *building* them, not designing them. I believe I could easily design a CPU using rod logic that would outperform any current (single-processor) CPU by a factor of at least 10, if not 1000. Or am I missing something in the question? Dennis [It depends on what you want to call "switch". The lock knob, moving at 1000 m/s (speed of sound in carbyne) can move from not blocking the probe knob to blocking it in a picosecond. However, as mentioned, Eric designed the thing *very* conservatively, since it is intended as an existence proof. Thus, the rods are supposed to move at a small fraction of Mach 1, and wait for vibrations to settle out, before the next cycle. Consider implementing a macroscopic mechanical computer. A 68000-scale machine with coat-hanger sized rods would be at least a cubic foot. Can you build a 1kHz clock machine in that technology? Speed of sound in steel would allow it, but my intuition says that 10 Hz would be doing very well indeed. Another thing to note is that real mechanical controllers are almost never digital computers. They are generally special- purpose devices, gaining orders of maginitude in efficiency to the particular task. If nanocomputers are mechanical, my bet is that a lot of them will be special purpose devices like cams, linkages, Geneva mechanisms, etc. --JoSH]