Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?

Intel's Thunderbolt With Fiber Optics Years Away 69

CWmike writes "Intel's Thunderbolt high-speed interconnect technology could be years away from getting optical technology, an Intel executive said this week at IDF. Originally introduced in February on Macs, Thunderbolt was pitched as being optical technology but currently uses copper wires. Dadi Perlmutter of Intel's Architecture Group said copper wires are working much better than expected, and that fiber was expensive. 'It's going to be way out,' Perlmutter said. 'At the end of the day it's all about how much speed people need versus how much they would be willing to pay.'"
This discussion has been archived. No new comments can be posted.

Intel's Thunderbolt With Fiber Optics Years Away

Comments Filter:
  • by Anonymous Coward

    .. we will be the judge of whether we need the speed or not.

    • by DrgnDancer ( 137700 ) on Friday September 16, 2011 @08:21AM (#37418734) Homepage

      Except it doesn't work like that. Let's say Intel know that it will cost them 10 million dollars to create the optical version of the tech. They know that the optical version will (ignoring dev costs, just on parts) cost, say, three times as much as the copper, but only offer a 15% improvement in performance. They can make a reasonable guess that while a small subset of people will happily pay three times as much for a 15% performance gain, they aren't going to be able to make their 10 million back. If they can't make back their dev costs, they aren't going to dev. They'll wait till the economics make more sense.

      • by Amouth ( 879122 )

        from what i've seen in how they are doing copper cables vs using fiber.. this is nothing that new - it reminds me almost exactly of using SFP's/Mini-GBIC which for networking and the san world allow for both copper and fiber connections - and while not always cheap, that is mainly an attribute to the market they belong (enterprise class equipment).

        i might be wrong but while i'm sure developing the underlying protocol and controller isn't cheap - it should have a simple interface to the cable. the controlle

      • Apropos of your sig: Since Intel is already active in some areas of networking hardware(in addition to their longer term R&D about optical chip interconnects) they probably already have a very good idea of what various high speed optical interfaces would cost. Their 10 Gig-E optical interface modules are off the shelf items today, and I'm assuming that they are continuing work in that vein to be ready for the next faster round of ethernet standards. I imagine that there are some differences between what
      • Yeah, that's all and well, but Intel has over $20 billion in cash, short-term investments and trading assets. Meaning a huge lot of money that, given the current interest rates, is probably earning almost nothing. Why not put the money to work and sink it into advanced, though possibly not extremely profitable, ventures? There is just no way putting cutting edge technology on the market and expanding the company patent portfolio can be bad for them, and some of the newer technologies might turn out to be mu
        • This I wouldn't disagree with; and Intel does do some of this. For whatever reason they don't think this particular product is worth it. I pulled the 15% number out of my ass of course, but if they aren't expecting the optical interface to be a huge improvement over the copper, maybe they just aren't bothering for such small gains.

      • The issue is per unit costs, not startup. I also think there is also the profit angle.

        Right now, if adding optical cost $5 in parts, that's $45 at retail... There's no clear use, so the "race to the bottom" starts befoe the tech is highly profitable. There is also the matter of handing Fiber Optic cables to the general public... It's just too fragile and users would rebel.

        Also, 10Gb fiber cards are like $500 each right now... Why would intel kill that market?

  • I need the optical connection now so that my optical mouse can fully function. Without it, its way too slow!

    • I need optical so that we could send video without any conversions. I'm sure we already have strong enough lenses to scale the picture down to the fiber diameter.
      • by Anonymous Coward

        You need a coherent fibre bundle, not just any old fiber, but that's roughly how endoscopes work.

    • by Tarlus ( 1000874 )

      "Sup dawg, we made your optical mouse connection an optical connection so that your optical... will optical..."
      Meh, it was worth a shot.

  • Like FireWire... (Score:4, Interesting)

    by MarcQuadra ( 129430 ) on Friday September 16, 2011 @08:27AM (#37418762)

    FireWire was also supposed to 'go optical' at some point, but market forces kept it copper.

    • ISTR the original roadmap showing 1.6Gbps on copper by now, and 3.2Gbps "some time in the future" on fiber (with copper next to it for power.)

    • by Gilmoure ( 18428 )

      If the MPAA folks hadn't pushed for DRM on HDMI, we'd have nice Firewire connected AV equipment and everything'd be cool. Can't wait until I find a gateway to the universe where that happened. Although I imagine there's a bunch of slider versions of myself all hanging out in my living room over there.

  • I read that as "we don't own all the patents on the interconnect hardware, and to produce it would cost us more than using our in-house patent base and patent-free copper connections. Surprisingly, it turns out we're somewhat incompetent at modeling electrical connections and the results don't match our simulations but they're better than we planned, so we'll patent what we have and plan on taking that to the bank."

    • I agree. Mod up. I don't know what the truth is, but I think Intel's explanation is BS.

      SPDIF has been around for years, and it isn't terribly expensive. I can get a 6 ft. cable for $2.99.

      It's so "expensive" that's it's built into the headphone and line in jacks on my Mac, and most people don't even realize it's there.
      • by Fjandr ( 66656 )

        Yeah, I've got an Asus laptop that's like ... 7 years old, and it has a hybrid copper-optical SPDIF port on it. Wasn't even a particularly expensive laptop at the time.

      • Afaict it's not the fiber itself that is expensive It's the laser diodes, photodiodes, precision connections to the fiber, protection against fractured fibers from kinked cables and so on that make a high speed fiber system expensive. Especially if it has to be made "idiot-proof".

        TOSLINK (optical version of S/PDIF) is indeed cheap but that is because the low speeds let them get away with REALLY low grade optical components (including polymer fiber that is less prone to damage than glass fiber.

        • It's the end connections I presumed were expensive, and an area where their patent portfolio was thin. I can agree somewhat on the cable front, and I want thinking about TOSLINK when I thought of cheap cables, but I'd never really considered what the bandwidth was.

      • Re: (Score:3, Informative)

        by JDG1980 ( 2438906 )
        Toslink (optical SPDIF) is very low in bandwidth compared to modern technologies. Most implementations don't go above ~1.5 mbps, which is enough for uncompressed redbook CD audio or DTS (and more than enough for Dolby Digital / AC-3). Just because this can be done cheaply doesn't mean that the same is true of optical connections that need to handle several orders of magnitude more bandwidth. And Toslink never made much sense in the first place. Coaxial SPDIF transmits exactly the same data, the cables are
      • by Anonymous Coward

        I agree. Mod up. I don't know what the truth is, but I think Intel's explanation is BS.
        SPDIF has been around for years, and it isn't terribly expensive. I can get a 6 ft. cable for $2.99.
        It's so "expensive" that's it's built into the headphone and line in jacks on my Mac, and most people don't even realize it's there.

        Optical SPDIF aka TOSlink operates at a few megabits per second (~5 max) using cheapass red LEDs (not even lasers! And truly generic off-the-shelf, any random 650nm red LED can switch on and off at a few MHz), the lowest grade plastic optical fiber, cheap plastic connectors with extremely poor alignment (important in the optical world), and so forth.

        You can't get away with that kind of low-grade hardware if you hope to send and receive 10 gigabits per second (as in Thunderbolt). That requires high performa

    • by tlhIngan ( 30335 )

      I read that as "we don't own all the patents on the interconnect hardware, and to produce it would cost us more than using our in-house patent base and patent-free copper connections. Surprisingly, it turns out we're somewhat incompetent at modeling electrical connections and the results don't match our simulations but they're better than we planned, so we'll patent what we have and plan on taking that to the bank."

      Given Thunderbolt copper cables rely on active cables (the cables actually have circuitry in

  • by Anonymous Coward

    I've never been a fan of optical cables, they have really poor flexibility for typical use in a desk/office environment. Considering how capable copper thunderbolt is, seems unnecessary.

  • by fuzzyfuzzyfungus ( 1223518 ) on Friday September 16, 2011 @08:54AM (#37418928) Journal
    While "Thunderbolt" is essentially a PCIe 4x external cabling mechanism, rather than a more typical external interface like ethernet, it seems reasonable to assume(for the sake of getting some rough numbers) that the challenges of getting a Thunderbolt 10Gb/s optical connection working would not be less than the challenges of getting other optical 10Gb/s connections working(might be slightly more, if, say, PCIe is touchier about latency or something, might be slightly less if Thunderbolt never promised to support a cable more than 10 meters long; but ballpark here).

    Conveniently, there exists just such a 10Gb optical interface: 10GigE. Even better, the optical portion is frequently broken out into a separate module(to allow for multiple different grades of tranceiver, depending on distance and fiber requirements), making it possible to price the optics package separately from the switch to which it attaches.

    10GB/s optical XFP or SFP+ modules are, indeed, not all that cheap. Much cheaper than they were; but (at least the Intel ones that some rough retail-pricing showed) still easily as costly as some of the smaller planned "thunderbolt" peripherals...
    • Yeah, going to optical when copper will do doesn't make economic sense. There's no need for the bigger bandwidth of optical as long as the transceivers are fixed and non-upgradeable, as they would be in consumer equipment. Even in the high-end space, there seems to be lots of 10GigE over copper these days.

      • As you note, with Thunderbolt showing up in places where swappable tranceivers are very unlikely to happen, it seems particularly unlikely that the interface will be going optical any time soon.

        In various niches, there is a demand, backed up by actual money, for all kinds of typically copper interfaces over optical, VGA, PS/2, USB, serial, etc. However, because the markets are so small, you can't really buy any devices with optical interfaces for that, you just make do with proprietary adapters with a co
    • I thought 4 channel Light Ridge Thunderbolt was capable of 4x10 gbps bidirectionally or (8/10 * 40 = ) 32 gbps of pure data. Isn't that PCIe 2.0 8x? Why does everyone say it's 4x? 32 gbps is 8x!
  • This is the latest in a very very long series of failures of optical interconnect in multiprocessors, in the computer room, and at the desktop. Since the '80s people have told us that wires will never keep working, and optics is the only solution. They have been wrong, and continue to be wrong. I was even blasted by a respected physicist that told me that there was an inherent power advantage for optics. That was wrong also. Optics is great if you need to go across the ocean, but don't tell me you wan
    • by Amouth ( 879122 )

      well there is one very really advantage to optics over wires.. and that is that there is an unlimited potential bandwidth - single mode fiber > any electrical conductor (excluding supper conductors)

  • TFA reads like an Onion [] article:

    "Copper will continue to improve, which happens. There have been many technologies that had been predicted dead 20 years ago that are still making good progress. We'll see," Perlmutter said.

    Aren't optical signals processed via devices connected with copper wires at the end of the day?

    • by Fjandr ( 66656 )

      Yes. The main benefit of optics is that you don't have the same sort of signal loss as a result of resistance. The longer the distance the more important a consideration that becomes.

      Short-run cabling probably isn't going to see much difference between the two transport mechanisms. Copper still has a lot of headroom.

      That processing that's done at each end is done by very efficient short-run copper. Until they start making optical traces and transistors in IC-size, that's going to be a limitation which is no

  • While shorter distances of copper are cheap for communications long distance is expensive and lossy. I wanted a technology that I could run across the house. and still utilize full speed. Cat5e will do for now. I know Cat6 exists but I can get gigabit ethernet for cheaper with cat5e and It works even across the house. Cat6 doesn't do significantly better. Finding adapters to utilize the capabilities of cat5/cat6 as high speed usb 2.0 480Mbits/second cables looks to be a $300 endeavor. Again I know I can buy
    • by Fjandr ( 66656 )


      Cable length is an issue for certain things, amongst them USB. I am currently at the spec limit for cable lengths. Since USB timing can get hairy when you exceed the limit on cabling, I have to make accommodations for that limit (chaining hubs is not something I will consider just to exceed it). I would like an external interconnect where this is not an issue. 15 feet is not a whole lot in many situations.

      Powered extenders are expensive, because they have to integrate timing corrections to compensate f

  • About a year ago, when I was reading about LightPeak(Thunderbolt), Intel claimed it was going to be 10-40gbit and was going to take a few years *after* 22nm became standard.

    Just based on Intel claiming LightPeak was meant to come out after 22nm, means it was released early. I am not surprised that the optical version is still some odd years off.

    Personally, I think this early release was a mix of Apple and Intel. Apple wanting the fastest and unique, and Intel wanting to make at least some money on their tec

  • Whats the point of it if we can use USB3? USB3 is compatible with my current USB devices Thunderbolt is not. Thunderbolt cost of implementing is higher than USB3. Its not that much faster there are almost no devices for thunderbolt. How many copper wires does it need? How would it scale for distances of say 50 feet? The real distance of going around the room if you don't go under the rug. If they stuck with fiber and had copper for power and sold adapters like they had set out to do. Even though its not tha
  • Now that graphics cards are powerful enough to drive 3 or more full HD displays all that's missing is a way to connect them across the home. LightPeak looked like a perfect fit: across one cable you could connect a display, a USB infra-red remote, and even USB drives or an SD reader. And with the fiber optic cable there was no range issue. You could just go through the attic and to the other side of the house, tens of meters away if you so wanted.

    But then all we got is Thunderbolt with a measly 3m maximum l

  • 4 channel Light Ridge Thunderbolt controllers can do 4x10 = 40 gbps bidirectionally. Since it's PCIe (8/10 is data), that's 32 gbps. That's equivalent to PCIe 2.0 8x (16x is 8 GB/s = 64 gbps). So it should be enough for even really high end external GPUs. And IMO that's the most relevant application. Intel wants to go the Ultrabook route, and the only way to make that viable is to enable the ability for people to connect high-end desktop class GPUs to their high-end CPU packing ultrabooks.
  • What about the Sony Vaio Z docking station Power Media Dock? It was advertised June 28 to be using ”an optical cable” and ‘Light Peak’. []

    It is available now, at $499.99, []

e-credibility: the non-guaranteeable likelihood that the electronic data you're seeing is genuine rather than somebody's made-up crap. - Karl Lehenbauer