markv@pixar.com (Mark VandeWettering) (04/19/91)
I had the distinct pleasure of attending Eric Drexler's talk at NASA Ames last week. It was most enlightening and enjoyable, and I thought I would mention some of the highlights. The context of his discussion was nanotechnology as applied to parallel computing. By manufacturing "mechanical" computers, he estimated that he could obtain a staggering 10^15 MIPS per cm^3 of matter. To back this up, he had the preliminary results of molecular dynamics calculations. He determined that a sequence of rods with knobs could be used to implement a simple PLA. The length of the rod must be bounded by the "play" in the system: the bond strengths, uncertainty and other factors. He found that you could implement a unit with 16 knobs on it, with an error rate which was on the order of 10^-12 or so. (That may not have been the exact magnitude he said, but it was of that order). Mechanical energy to move these carbon rods are generated from small electrostatic motors. More details were presented, which I am unfortunately unable to reproduce. It did address some of the shortcomings that I had seen, and I was pleased to see some hard "engineering" studies of the problems. By his determination, a typical 10 MIPS workstation could be reproduced within a cubic micron (which would make it much smaller than a single human cell). It would dissapate very little heat, and in general be incredibly non-intrusive. I prefer to think of them as the world's smartest virus. Anyway, perhaps others who attended the talk would like to comment. mark
kevin@maspar.com (Kevin S. Van Horn) (04/20/91)
In article <Apr.19.00.09.45.1991.25876@athos.rutgers.edu> markv@pixar.com (Mark VandeWettering) writes: >I had the distinct pleasure of attending Eric Drexler's talk at NASA >Ames last week... > >The context of his discussion was nanotechnology as applied to parallel >computing. By manufacturing "mechanical" computers, he estimated that >he could obtain a staggering 10^15 MIPS per cm^3 of matter. To back this >up, he had the preliminary results of molecular dynamics calculations. Has this talk, or its equivalent, been written up anywhere? I would be very much interested in seeing the details of his calculations and proposed design. ------------------------------------------------------------------------------ Kevin S. Van Horn | The means determine the ends. kevin@maspar.com |
merkle@parc.xerox.com (Ralph Merkle) (05/01/91)
Drexler's paper analyzing molecular logic is titled "Rod Logic and Thermal Noise in the Mechanical Nanocomputer," by K. Eric Drexler, in Molecular Electronic Devices, F.L. Carter, R. Siatkowski, H. Wohltjen, eds., North-Holland 1988. Copies are available from the Foresight Institute.
Howard.Landman@eng.sun.com (Howard A. Landman) (05/01/91)
In article <Apr.19.00.09.45.1991.25876@athos.rutgers.edu> markv@pixar.com (Mark VandeWettering) writes: >By manufacturing "mechanical" computers, he estimated that >he could obtain a staggering 10^15 MIPS per cm^3 of matter. >He found that >you could implement a unit with 16 knobs on it, with an error rate which >was on the order of 10^-12 or so. So assuming each unit executes one instruction (a very generous assumption), we get 10^21 IPS * 10^-12 errors/I = 10^9 errors per second for a cm^3 size supercomputer. Really reliable. :-) -- Howard A. Landman landman@eng.sun.com -or- sun!landman [Giga-EPS, a new computational measuring stick. Seriously, though, 10^-12 is a perfectly usable error rate for even the simplest error detecting and correcting codes. A SECDED code would give you an average of more than Avogadro's number of correct operations between detected but uncorrected errors, and something like a million years between undetected errors. --JoSH]
Howard.Landman@eng.sun.com (Howard A. Landman) (05/10/91)
In article <Apr.30.15.15.56.1991.11923@planchet.rutgers.edu> Howard.Landman@eng.sun.com (Howard A. Landman) writes: >So assuming each unit executes one instruction (a very generous assumption), >we get 10^21 IPS * 10^-12 errors/I = 10^9 errors per second for a cm^3 size >supercomputer. Really reliable. :-) Josh replies: >[Seriously, though, 10^-12 > is a perfectly usable error rate for even the simplest error detecting > and correcting codes. A SECDED code would give you an average of > more than Avogadro's number of correct operations between detected > but uncorrected errors, and something like a million years between > undetected errors. Yes. But the high total error rate means that error correction and fault tolerance are REQUIRED features of any computer in that class. Traditional architectures don't cut it. Of course, since the MTBF is long enough to do a million million instructions (roughly what a 3,000 MIP machine could do in an hour!), for most simple tasks you could just rerun the job a few times and compare results. Also, there are some applications where reliability is a very low priority, such as real-time graphics/animation (if one frame gets a little screwed up it's not usually a big deal). A (nano-)processor per pixel, anyone? -- Howard A. Landman landman@eng.sun.com until Friday May 3, 1991 then at Crosspoint Solutions, (408) 988-1584 probably as landman@xpoint.com or uunet.uu.net!xpoint.com!landman