I agree Charles - we have been hearing of this technology for years - transmitting light to carry data in computers. It's nice to see someone is working on a solution and it is starting to become something that may be marketable in the near future...
Nancy, I first heard of switching photons instead of electrons about 15 years ago in a venture capitalist meeting about a new tech, I forget which one or from which company. I was electrified--or perhaps I should say, photofied.
I remember hearing about a computer that was built completely with fiber optics back in the late 90's. I think it was designed by the R&D department of one of the major tech universities. We were "photofied" as well!
Thanks, Chuck. Looks to me like some patient, careful R&D on the part of two big companies that know how to do patient, careful R&D and have the deep pockets for it. Plus how to come up with a practical solution that addresses all the challenges. I don't see that very often.
Ann, there is no doubt that light can carry more information at a higher speed. Moreover, I think signal losses are also very less and what about the cost factor when compare with the conventional method of data transfer.
Jack, I don't mean to pry, but would you tell us what decade that was, or how many years ago? I'm curious because I first heard of this idea in the mid-late 90s. Did you encounter this idea before or after that?
@ mydesign, my point is that for those very short distances on a backplane the optical connection woul;d be way more bother than it would be worth. For larger distances, several feet or more, it makes some sense, but for board-to-board interconnect I see it as a waste of effort and materials.
I think there's some confusion about the nature of optical connections being proposed. The whole point of this new technology is that, if you used optical *fibers*, you need so many at the board-to-board, chip-to-board and chip-to-chip levels that you can't physically get them into such a small space: they don't scale like transistors do. But if you implement them in waveguides--i.e., embed optical connections in thin sheets of silicone via standard semiconductor manufacturing methods for laying down circuits--you can get the size down for placing on, or integrating with, boards. The researchers specifically targeted this smaller, chip-to-board scale first, and are also proposing to extend it to board-to-board connections.
"those very short distances on a backplane the optical connection woul;d be way more bother than it would be worth. For larger distances, several feet or more, it makes some sense, but for board-to-board interconnect I see it as a waste of effort and materials."
Willam, am agreeing with you. For shorter distance coaxial cables are better than optical fiber because attenuation losses are less in shorter distance. Moreover inter connectivity is difficult with fiber optics.
Mydesign, the whole point of using this new optical technology--which does not rely on bulky fibers--is that communication speeds are far outpacing coax in high-end computing, and will do so relatively soon, at the current speed increases, in computers used by the rest of us.
It still does not sound like a useful technology, just because of having to convert from and then back to an electrical signal. That takes up both space and power, and then the losses are greater in any sort of optical medium than they are in an equal run of copper, correctly designed and built. Besides that, it certainly seems that the accuracy of alignment needed for these optial interconnects is going to be a bit tighter than for electrcal connections.
It still looks a whole lot like somebodie's neat solution searching intensely for a suitabl problem.
OF course, if it does turn out to be a useful idea, then we will see it in industry, someplace. If not, then we won't. But I am NOT going to invest in that company.
William, sounds like you're unclear on the waveguide concept. I suggest you read up on them, and specifically on this board-level one. Wikipedia has a good article on the general subject. The "suitable problem" already exists and the R&D has been going on for some time to implement these at the board level. There's also a fair amount of detail in the links we gave in this article.
I am a bit familiar with radio frequency waveguides, but not optical ones, other than fiber optics. So thanks for the links.
But it still seems that there are very few applications that would benefit from the very high speed multiple line connections, since it appears that the concept is for these links to carry at least 8-bits wide, and probably 32 0r 64-bit wide data. For single-line interconnections they would probably not be cost effective.
My guess is that the main application would be in medical image processing, which is a growing but rather narrow field. If there are others it would be interesting to hear about them. But computer gaming does not count as a valid application, at least in my book.
William, the article mentions supercomputing and datacenters, not computer gaming, for right-now apps. As a long-time student of comms technologies, I know that what starts at the high end--such as those two apps--ends up in the PC, at least as far as data transfer speeds are concerned. Remember when 100 MBps was fast? In the datacenter?
Ann, you may be correct about data centerapplications, but i suspect that the market for computer gaming toys is much larger. Of course, it will all get down to the ratio of cost-to benefits, won't it?
William, I'm not sure where computer gaming as an app for this technology comes from. The "apps" are more in classes of hardware--supercomputers, routers/switches, PCs--than in uses of the hardware. If you mean the technology may eventually come to gaming platforms, I agree--but then, it will also come to PCs and other consumer computing devices. And of course, price/performance tradeoffs will be one determining factor. But, in comms and connectors, that's not enough: market saturation will rule the day.
Ann, sorry if I caused confusion by using the word "application". My usage relates to what the word was taken to mean in the past, instead of a corrupted abbreviation for the phrase "application program." Lazy-mouthed slang terms seldom are able to convey a specific meaning accurately, it seems.
I am aware that language does indeed change with usage, but hearing a slang term used to reference somethoing that many people really don't have any understanding of what they are talking about does become rather boring. At least I find repetition without understanding to be boring.
How right you are about the uses of the term "application." And slang, too, although it's fun, often is confusing. However, I was not using "application" to mean a program, or class of programs, but a class of uses--for instance if I write "aerospace applications," this does not mean aerospace software, but various uses in the aerospace industry.
OK, Ann. It still seems that the largest market for high performance computing is in the games area. Of course I realize that the level of power for games is less than that of the high level scientific computers, but thye sales ratio is quite large.
But your comment about servers does indeed point to an area that I had not considered. So now for a question about the "wide" optical interconnect: would it be point to point, or would it be more like a bus? Point to point between adjacent boards could still be done by some other means, while a bus with multiple sources and multiple listners would be an entirely different realm. Very demanding of precise construction and alignment, and probably susceptible to the same problems that lead to hard drives now using the sata interconnect format.
Games may be a large market for high-performance computing, and thus of this technology. But I think what keeps getting lost in this discussion is that high-performance computing will not remain the only stratum where this technology is needed/useful. That's why I said "Remember Ethernet?" to remind us of how those speeds have continued to increase while the need for them has migrated down the performance spectrum. Faster data transfer is and will be needed everywhere, including personal computing devices.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.