Forgot your password?
typodupeerror

Paint-on Laser Brings Optical Computing Closer 132

Posted by ScuttleMonkey
from the paint-by-number dept.
holy_calamity writes "New Scientist has a story about a laser made by painting a solution of semiconductor crystals onto glass. It could be used to break the interconnect barrier by having optical interconnects, the interconnect barrier threatens Moore's law unless a faster way of connecting chips is found."
This discussion has been archived. No new comments can be posted.

Paint-on Laser Brings Optical Computing Closer

Comments Filter:
  • by LiquidCoooled (634315) on Wednesday April 19, 2006 @03:58PM (#15159970) Homepage Journal
    Is there any chance this paint is waterproof.

    Sincerely,

    Dr Evil.
  • I've suddenly got a new, wonderful idea for a stupid "Pimp My Ride" car customization.
  • Google has failed me, turning up references to it but nothing about it.
    So can someone explain what it is... and what exactly the problem is?
    • by Enigma_Man (756516) on Wednesday April 19, 2006 @04:21PM (#15160177) Homepage
      Well, I don't have any direct sources, but: The interconnect barrier means that while chip size keeps decreasing, the interconnects between sections of chips, and even between individual chips themselves aren't able to be made much smaller, so things like resistance, capacitance, and inductance get in the way with the bigger interconnects. Basically, the interconnects can't keep up with the growth of the chips themselves.
    • by peragrin (659227) on Wednesday April 19, 2006 @04:27PM (#15160232)
      yep it's simple really.

      take your standard Network. Incoming ISP network, Local Router, cables, computers.

      Now take a Fiber optic version of that. Fiber from ISP, to Interconnect, to router, cables and computers. Sometimes they can even make the lines to the machines fiber as well but not always.

      Basically in order to have fiber optics everytime you hit a junction you have to convert the signal to electrical, sort it, and then convert it back to light. That process slows down the overall data transfer rate considerably.

      What they are trying to do is make it so that you can plug the fiber right into your computer and have the signal remain as light the entire distance it travels. This will increase bandwidth and speed of the networks signifcantly just be replacing routers.

      • What does network bandwidth have to do with Moore's law? Either you or the submitter has got something confused....
        I think I'll trust Enigma_Man and this has something to do with the connections between chips.
        • This, from the article, is probably where he's getting the assumption:

          The team's biggest hurdle was tuning the frequency of the beam to infrared, the frequency most used in optoelectronics and fibre optic cables.

          Obviously though, from the article yet again:

          "People are making all kinds of optoelectronic components, but the missing piece has been the laser source," Sargent says. "This is the first paint-on semiconductor laser - it could be used to connect microprocessors" in an optical computer.

          That s

      • I don't think this is completely new I remember hearing about some kind of fiber dipped in a liquid to amplify the signal on long fiber runs 2 years ago in school.
        The process of recreating the signal would still be needed to remove unwanted noise which means a computer needs to identify each 1 and each 0 then send those back out again the same as if they were being converted to electrical signals.
      • by anttik (689060)
        Whoa, your modders are on crack. Your answer is totally offtopic.
      • Er, how does having a laser that is painted on a surface prevent you from having to convert an incoming signal from light to electrical?? The innovation here is that you can create laser light in a small package, not that you can route already existing light from an incoming signal.
    • by peragrin (659227) on Wednesday April 19, 2006 @04:33PM (#15160282)
      this is what you get for NRFA

      Well in this case they have developed a way to create an infrared laser small enough to go inside a CPU.

      Optical computers Here we come. BSOD's will really turn your computer Blue.
  • by hkgroove (791170) on Wednesday April 19, 2006 @04:17PM (#15160144) Homepage
    But, will we have 5 megawatts by mid-May?
  • by iXiXi (659985)
    Now I will have to build a new Flux Capacitor!
  • Actual Moore's Law (Score:2, Insightful)

    by Anonymous Coward
    "The observation made in 1965 by Gordon Moore, co-founder of Intel, that the number of transistors per square inch on integrated circuits had doubled every year since the integrated circuit was invented. Moore predicted that this trend would continue for the foreseeable future. In subsequent years, the pace slowed down a bit, but data density has doubled approximately every 18 months, and this is the current definition of Moore's Law, which Moore himself has blessed. Most experts, including Moore himself, e
  • There are frequency limitations to the speed at which processors can be run. Something about crossover fields at frequency or some such.

    Light would have the ability to be switched much more quickly, but if you're going to switch it with electricity based circuits.....
  • Speed increase (Score:5, Informative)

    by centie (911828) on Wednesday April 19, 2006 @04:35PM (#15160313)

    The article and summary seem to be a bit misleading and vauge about how the speed increase arrises. The great benefit of optical computing is that it allows the signals to get much much closer together than electronic circuits, and as such allow more compact circuits, which as we know generally means faster. Interestingly, electronic signals in wires and optical signals in fibers have roughly identical upper speed limits (light in free-space optical computers is faster, but also almost impossible to do anything useful with), so its the density which is the major factor.

    Electrons are charged, so as you squeeze transistors closer together, the wires get thinner and closer together, and you get cross-talk and interference between them. Photons however hardly interact at all, so you can have many beams in the same space, and theres very little heat to be dissipated. Multiplw frequencies can also be used, resulting in massivly parallel computing (another GoodThing).

    There are downsides with optical computing still, photons cannot be stopped and stored (easily), meaning any kind of useful computer in the near term is likely to be some sort of electro-optical hybrid, with photons carrying signals and electrons storing them

    • Well, I guess consumer computing will just remain an "electrifying" experience.

      But, can't they make "paint-on" CPUs? I mean, the CPUs and the interconnects are like hand and wrist, right? Well, can they make them of similar "DNA" and part? Or, are they trying not to "kill off" some sacred part of the CPU industry?
  • by Ancient_Hacker (751168) on Wednesday April 19, 2006 @04:38PM (#15160333)
    A few quibbles:
    • How does a paint-on "laser" supercede a regular junction laser?
    • How do you solve the diffraction problem?
    • How does light communication solve the Moore's law problem?
    • Moore's law may be peering out, but mainly due to leakage and noise issues.
    • Usually the denser a chip, the less need for wide paths (to cache, RAM).
    • How does a paint-on "laser" supercede a regular junction laser?

      Quantum.

      How do you solve the diffraction problem?

      Quantum.

      How does light communication solve the Moore's law problem?

      Quantum.

      Moore's law may be peering out, but mainly due to leakage and noise issues.

      Quantum. Also, Bell's inequality. Quantum.

      Usually the denser a chip, the less need for wide paths (to cache, RAM).

      Quantum!

      Any questions? (Give ya one guess what my answer is...)
    • >* How does a paint-on "laser" supercede a regular junction laser?
      Nobody said that it was superior, just easier to make.

      >* Usually the denser a chip, the less need for wide paths (to cache, RAM).
      You're awfully wrong here: if you look at the history of x86 CPU, the chip density has increased a lot and at the same time, the datapath have been improved (increased frequency) *and* widened.
      The increased number of transistor and increased frequency allow the CPU to do more things, that's why it needs better
  • Bottleneck Slide (Score:3, Interesting)

    by Doc Ruby (173196) on Wednesday April 19, 2006 @05:10PM (#15160558) Homepage Journal
    The bottleneck in computing isn't Moore's Law of transistor density. It's programming paradigms. We're wasting the vast majority of processing/memory/transmission capacity with linear programming, rather than parallel programs. Procedural programs are based entirely on the bottleneck paradigm, with the entire system reduced to a single boolean operation at any given time. Any parallelism is exceptional, and difficult to express in the symbols humans send to computers.

    Parallel dataflow and distributed control are long overdue to the mainstream. Compilable UML is a slow, crude path to it. When I can draw a flowchart of primitive objects, any of which are packaged procedures or other flowed objects, and watch it run, I'll have a much better shot at exploiting all the compute/storage/transmit capacity available at that time. When "compilers" can distribute my data among the resources according to topology and analytical prediction, I'll finally get full use of the machines I'm using. Until then, I'm doubling my HW capacity every year or two so it can use half the efficiency gain running inefficient software.
    • We're wasting the vast majority of processing/memory/transmission capacity with linear programming, rather than parallel programs
      You probably don't mean linear programming [wikipedia.org] but sequential programs.
      • You are correct. I was thinking of a term to contrast the "nonlinear" parallel programs, and came up with the term that already means something else. Thanks for the correction.
    • would that not lead to a lot more task switching?

      a cpu is still a calculator at heart, numbers and actions in, results out.

      to realy be able to do proper paralell prosessing you would need to go for cpu's like the ibm cell or similar i think. there you can give each SPU a diffrent part of the "flowchart" and then use the power based core as a general traffic cop.

      flowchart programming on top of a cpu designed for proper paralell prosessing, now that would be interesting.
      • Most CPUs these days are largely parallel. They decompose sequential code into parallel pipelines, even to the point of speculatively computing a pipeline to keep it full, even if those results are not necessarily used by the later code - when the later code decides to execute the precomputed results, the CPU just supplies them - or discards them if the later code decides not to execute that sequence. All that sequential/parallel translation is fairly crude (compared to the elegant logic humans like to thin
  • Moore's Law is only an observation, not a performance goal. Of course it'll go away at some point. Maybe the slowing of density increases points to a maturing of one part of the industry.

  • I recall that a semiconductor engineer mentioned "optical computing" to me at least 20 years ago when I was a kid, and I was thrilled by it. Will this involve the interconnects only, or the whole CPU? Maybe the whole system could be built into an optical chip?
  • A few paragraphs of vague details, then a sophomoric rehash of the last 30 years of semiconductor design. Did New Scientist need some filler material to meet a publishing page quota or something?

  • by rew (6140) <r.e.wolff@BitWizard.nl> on Thursday April 20, 2006 @01:08AM (#15162720) Homepage
    [i]the interconnect barrier threatens Moore's law[/i]

    Terribly sorry to rain on your parade, but the fact that we live in a 3D world with a speedlimit limits computing speed eventually.

    Electrical signals in wires travel (according to rough measurements I did about two decades ago) at about 0.3c (a third of the lightspeed). Light travels at 0.6c (in glass).

    So you win about a factor of two by moving to light, provided you use fibreglass to channel the communications to the right place.

    If you Aim lasers through normal air, you can win a factor of three. Wow. That might extend Moore another 2 years, but it does not solve the fact that physics limits Moore eventually.

    In theory, "computing nodes" can be connected using for example hypercubes. 4 nodes form a square with max communications distance of 2, 8 nodes form a cube, with max distance of 3. And so on.

    Wether these "computing nodes" are complete computers, elements of a parallel system, or just elements of a CPU, doesn't matter.

    As the dimension of the hypercube increases, the physical placement of the nodes in 3D-space means that the communications links between the nodes starts to increase. The Lightspeed limits theoretical computation speed to what you might expect of a 3D structure.
    • "Electrical signals in wires travel (according to rough measurements I did about two decades ago) at about 0.3c (a third of the lightspeed). Light travels at 0.6c (in glass)."

      The fact that the bits would get were they are going in an optical processor twice as fast as in a conventional processor (latency) is beside the point when you realize that photonic bits can be fit MUCH closer together than conventional electronic bits (more gigahertz) because photonic bits as opposed to conventional bits are hardl
      • Fully agreed: The theoretical bandwidth of optical links is unbelievably high compared to electrical links (Note that 100MHz plus electrical busses on motherboards were thought impossible about a decade ago).

        We still live in a 3D space. Moore predicts exponential growth. We'll hit the limit that 3D space imposes within the next century or so.

        Visible light can handle a bandwidth of up to about 600 thousand gigagherz. If you can handle that optically, fine. However currently we need to generate the informatio
        • "However currently we need to generate the information stream for a communications link electrically, and we're still limited by the electronics."

          True

          "Can you point me to an optical computer having more than 10^3 gates? 10^4? 10^5? Sure, Optical routers exist. But that's just switching. Not really computing."

          Granted

          "So, for something that's optical already, having a couple of optical gates comes in handy. But for computing it has yet to become useful."

          Yes, "yet to become useful", my point exa
  • I had been wondering how to create lasers small enough to make a palm-mounted version of a laser-induced-plasma-channel http://en.wikipedia.org/wiki/Laser-Induced_Plasma_ Channel [wikipedia.org] Now I have it.
  • The title said it all! Hardware stores will be the battlefields of the future! re:
    http://www.legalwarfare.com/index.cfm?attributes.f useaction=showTemplate&calendar=2006-4-20&cy=2006& cm=4 [legalwarfare.com]

    Warning - link contains strong language =)

The sooner you fall behind, the more time you have to catch up.

Working...