ian@cbosgd.ATT.COM (Neil Kirby) (04/27/87)
Larr McVoy asks dan@prarie "..what does optics buy you?" Optical computing has the potential to buy you a great deal in terms of speed. In optical computing, light switches light. Currently we switch electricity with electricity. If I understand things correctly, we are nearing the limits of electrical switching speeds. Already we are worrying about transmission effects in boards and in silicon. I've been told that the limits for light are magnitudes higher. Optical switching exists in the lab NOW. As far as when or if it makes good computers remains to be seen, but the potential for speed is incredible. AMD said a few years ago "Reach for Light Speed". I wonder if they knew that folks would take them up on it. Neil Kirby ...cbosgd!ian PS: Please include a clean path in any replies, I don't trust the 'reply' function of the mailer.
farren@hoptoad.uucp (Mike Farren) (04/28/87)
In article <3560@cbosgd.ATT.COM> ian@cbosgd.ATT.COM (Neil Kirby) writes: >Larr McVoy asks dan@prarie "..what does optics buy you?" > >Optical computing has the potential to buy you a great deal in terms of >speed. >[...] >Optical switching exists in the lab NOW. As far as when or if it makes >good computers remains to be seen, but the potential for speed is >incredible. While optical switching exists in the lab, and general-purpose optical computers are probably some time away, special-purpose equipment which uses electro-optic technology exists right now, and more is coming. One example of such technology is the Instantaneous Fast Fourier Transform box manufactured by Litton/Applied Technology. How about a FFT computed in the time it takes a light beam to move one foot? Also, several efforts are underway to manufacture more generalized array processors with electro-optics. One that I know about was set up to multiply a rather large pair of matrices (I forget just how large, but on the order of 1K X 1K) in 50ns. The potential for serious number-crunchers is immense. -- ---------------- "... if the church put in half the time on covetousness Mike Farren that it does on lust, this would be a better world ..." hoptoad!farren Garrison Keillor, "Lake Wobegon Day
farren@hoptoad.uucp (Mike Farren) (04/28/87)
In article <2063@hoptoad.uucp> farren@hoptoad.UUCP (Mike Farren) writes: >Also, several efforts are underway to manufacture more generalized >array processors with electro-optics. One that I know about was set >up to multiply a rather large pair of matrices (I forget just how >large, but on the order of 1K X 1K) in 50ns. The potential for >serious number-crunchers is immense. > Whoops! Not quite that big - my brain wasn't working well when I posted the preceding... I don't remember how big, but I do know that a lab prototype setup was actually doing 32 X 32 matrices in the same 50ns. Much bigger stuff was in the works. -- ---------------- "... if the church put in half the time on covetousness Mike Farren that it does on lust, this would be a better world ..." hoptoad!farren Garrison Keillor, "Lake Wobegon Days"
thompson@dalcs.UUCP (04/29/87)
Two or three years ago, I read a article in Scientific
American (I don't remember the exact issue but if anyone is
interested I can look it up) about these optical transistors that
someone had developed. The thing that impressed me is that not
only would you get the high speed of light in a computer built of
these things but you could also get parallel computing by using
different colors of light.
--
Michael A. Thompson, Dept. Math, Stats, & C.S., Dalhousie U., Halifax, N.S.
thompson@dalcs.uucp From Bitnet or Uucp
thompson@cs.dal.cdn From Bitnet or Cdn
thompson%dalcs.uucp@seismo.arpa From Arpagareth@comp.lancs.ac.uk (Gareth Husk) (05/01/87)
In article <2547@dalcs.UUCP> thompson@dalcs.UUCP (Michael A. Thompson) writes: > > Two or three years ago, I read a article in Scientific > American (I don't remember the exact issue but if anyone is > interested I can look it up) about these optical transistors that > someone had developed. The thing that impressed me is that not > only would you get the high speed of light in a computer built of > these things but you could also get parallel computing by using > different colors of light. >-- The edition of SciAm was December 198[12]. The group that developed these wee beasties is based at Heriot-Watt University in Edinburgh. I went to a talk by one of the group last year and he was a bit embarrassed by the article. He said it represented an awful lot of crystal ball gazing. And that the editors had removed the depressing stuff about what they actually had working rather than what *might* happen. What they have got are discrete components cooled by liquid Nitrogen and they are using a thermal effect in the switching, thus the speed of switching is v.low because they have to wait for the device to stabalise after each switching. So that although transmission is ~.9c switching is governed by fairly low speed thermodynamic effects. Dr. Abrahamson (sp) was talking of 15-20 yrs more work at least plus the need to have optically reactive materials or much smaller thermal effects to lift processing speeds to the pico second levels mentioned in the article. -- " I am a doughnut " JFK UUCP: ...!seismo!mcvax!ukc!dcl-cs!gareth DARPA: gareth%lancs.comp@ucl-cs JANET: gareth@uk.ac.lancs.comp