Via quizzed on USB 3.0 implementations, Light Peak potential

Thanks to USB 3.0, external storage is interesting again. The additional bandwidth offered by the SuperSpeed spec is sorely needed, and even budget external drives are more than fast enough to take advantage. Most TR readers probably associated USB 3.0 with the two-port NEC chip that’s dominated the motherboard market, but it’s not the only solution in town. Via has a number of different USB 3.0 devices, and project manager Terrance Shih recently sat down with the folks at SemiAccurate to discuss the technology.

To date, Via has revealed a number of different USB 3.0 devices, including 2- and 4-port host controllers, a 4-port hub controller, and also a 4-port NAND controller. The host controllers have yet to be certified by the USB-IF governing body, but Shih expects to get the group’s seal of approval in early 2011. Whether that’s early enough for the chips to appear on Sandy Bridge motherboards remains to be seen. Gigabyte’s P67A-UD7 motherboard does feature a pair of Via USB 3.0 chips, but they’re hubs connected to NEC host controllers.

Other USB host controllers have been slow to receive certification from the USB-IF, and some have blamed driver development for the delays—Windows 7 lacks native USB 3.0 support, so controller makers are on their own. According to Shih, Via did all of its driver and firmware development in-house, and it found maintaining backward compatibility with USB 1.1 and 2.0 devices more challenging than implementing the new spec. Shih also goes on to say that the physical layer made up of “connectors, cables, and the actual technologies involved in sending and receiving raw data bits” presented the greatest challenge on the hardware front.

When asked whether Light Peak is a threat to USB 3.0, Shih suggests that Intel’s next-gen optical interface could revolutionize docking stations, but that it has no answer USB 3.0’s backward compatibility with the “literally billions” of USB devices in the wild. That’s an interesting point, because while I can see a 10Gbps Light Peak link being an excellent conduit to a collection of expansion ports, I can’t think of a single one that would require that much bandwidth, at least for consumer applications.

Comments closed
    • Sahrin
    • 9 years ago

    Light Peak has all the drawbacks of a new high-tech networking standard with none of the advantages.

    A better solution is continued development of USB, and commoditizing Fibre Ethernet (which enables 10GBaseTX and up more easily than traditional UTP – though truth be told it’s not necessary). Fibre Ethernet for ridiculously high bandwidth or long-haul connection; USB for local high-bandwidth and powered connections; wireless for medium-bandwidth and unpowered connections.

    The only real ‘purpose’ LP serves is to lock the peripherial world under an Intel licensed-solution. Great for Intel. More expensive for everyone else.

    • GreatGooglyMoogly
    • 9 years ago

    Meh, eSATA > USB3 for storage.

    • Krogoth
    • 9 years ago

    Light-Peak already has some teething issues which is forcing Intel to go with copper. (Now Copper Peak? 😉 )

    Anyway, Light-Peak biggest problem is that it is an interface marketed towards the wrong audience. Average joe and majority of computer hardware enthusiast do not need ~10Gps interconnection for peripheral devices. The HPC and clustering crowd would love to have a cheap, high-bandwidth interface.

      • OneArmedScissor
      • 9 years ago

      Average Joe would surely appreciate a universal connector, or maybe the possibility of a little extra battery life for the stupidly overpowered laptop he will buy just to check his facebook. :p

        • Krogoth
        • 9 years ago

        Son, we already have that. It is called USB. 😉

      • MadManOriginal
      • 9 years ago

      Uh, how do you figure Light Peak is being marketed to ‘mainstream’ customers, just because Intel doesn’t have a chipset with built-in USB 3.0 yet? (as far as we know – Fudzilla said it’s rumored SB chipsets will have USB 3.0 and Intel is just holding it close to the vest.) Ask a mainstream buyer what Lightpeak is and I’d be surprised if they even heard of it, maybe they haven’t heard of USB 3.0 in particular but they know USB, and at the same time even USB 3.0 is niche-ish because the only use for which it has an advantage is for external storage. You are absolutely right that in the near term HPC, clustering uses, and other enterprise-level applications will eat Lightpeak up and that’s all it’s being marketed to atm. We know about it because we’re tech nerds 😉

      I agree with you that USB 3.0 will remain for a long time and likely won’t be replaced by Lightpeak especially when cost is taken in to consideration, I just don’t get the ‘wrong marketing’ statement.

    • dragmor
    • 9 years ago

    Light Peak = External Graphics (its the only thing that needs the bandwidth).

    The USB standards body dropped the ball, they should have bought USB3 out at 100MB/s 3 years ago. Actually all of the standards bodies move to slow. JEDEC hold back rams speed. We won’t get SATA4 until 2013 when 2011 drives will be maxing out SATA3.

      • indeego
      • 9 years ago

      SATA is also a dying standard, SSDs are already outstripping the latest iteration. PCIe / Onboard blade-style slots (like macbook air) will likely be the next waveg{<.<}g

        • shank15217
        • 9 years ago

        Wrong.. sata is going in lockstep with ssd development. On board pci storage is niche and will stay like that.

          • mutarasector
          • 9 years ago

          “Wrong.. sata is going in lockstep with ssd development. On board pci storage is niche and will stay like that.”

          I seriously doubt that. It may be niche now, however the RevoDrive X2 almost matches current SSD drive prices, but seriously outperforms them. If onboard PCI storage is niche, its only because current internal SSD drives aren’t quite mainstream either, but as even they’re becoming more popular as the pricing drops, they will likely become standard system/boot drives(at the very least) within the next 5 years.

        • Krogoth
        • 9 years ago

        In the server/workstation world, it will probably be true.

        For the mainstream world, SATA and USB are not going away. They are too ubiquitous to ignore.

        Intel is going to be disappointed with “Copper Peak” adoption. It is looking like a repeat of Firewire. It isn’t the first time Intel had tried to cram a standard/technology down the market and it flopped. Intel will soon be eating crow and be implementing USB3 to battle against AMD on mainstream arena/

    • derFunkenstein
    • 9 years ago

    I was not really impressed with either of the two VIA USB 2.0 cards I owned, and preferred cards with NEC controllers back in The Day when I was using a PPC Mac. I was not terribly impressed with either them in Windows, either; devices would drop out or quit responding until I unplugged and replugged them. So I personally would stick to motherboards that use the NEC USB3 chipsets.

    • thudo
    • 9 years ago

    USB3 should have been more widely adopted since it started rolling out over a year ago but its all about LightPeak thats holding USB3 due to Intel. As mentioned above, USB3 is “a threat” to LP even though LP likely is superior (unless it takes forever to be widely adopted AND they don’t allow devices to be powered by it like USB3 does so well).

    I’d like to see USB3 ramp up on boards taking over all the 10+ slots as right now its a puerile joke (only 2 mostly). We need to see what LP can truly do but USB3 is here for over a year and its being stifled.

      • Deanjo
      • 9 years ago

      USB 3 is pretty much a stillborn standard. The only thing that can really utilize the extra bandwidth is external storage. Unfortunately most external storage can easily be handled with esata. USB 3 didn’t bring a much higher current capability and by doing such it offers no real advantage over esata as external devices still usually need their powerbricks (unless it is a low powered 2.5 in drive).

        • Farting Bob
        • 9 years ago

        Are you kidding? eSATA is dead, USB still dominates the external storage market. Hell, even new multi-TB enclosures still will only have USB 2 on them. My nan knows how to plug any USB device into her computer. Give her an eSATA cable and she’ll just give up on computers all together.

        • clone
        • 9 years ago

        I disagree that USB 3.0 is dead or stillborn, it’s so far ahead that little has been offered to max it out.

        the backwards compatibility is also a huge bonus and that alone will force at worst USB 3.0 to coexist for at least a decade.

        flash memory isn’t fast enough for USB 3.0, external HDD’s aren’t quick enough and the rest don’t do anything requiring the throughput.

        Camera’s, video and photo’s…. if flash read speeds ever do a huge jump USB 3.0 will be the bomb or if SSD capacity actually starts to compete with HDD’s then USB 3.0 will be getting used proper.

        • OneArmedScissor
        • 9 years ago

        I haven’t seen any eSATA smartphones or MP3 players lol. Once those are too high capacity and fast enough to truly justify it, they’ll move to USB 3.0 and it will take off like a rocket.

        • jensend
        • 9 years ago

        What are you talking about? USB3 devices can depend on being able to draw 900mA; USB2 spec was limited to 500mA. That moves every 2.5″ magnetic hard drive out there (except the VelociRaptor) from being sometimes and undependably workable to being easily within the power budget.

        Yeah, it might be nice to power a 3TB Caviar Green off a single USB port, but providing that much power to each of ten ports gets to be more costly. With terabyte 2.5″ hard drives, many people don’t need 3.5″ external storage anymore, and SSDs and other flash storage continue to improve, further lessening the need for more power.

        • indeego
        • 9 years ago

        I agree. USB3 is DOA. If Intel ignores something that could sell chips, you pretty much know they want something else that will sell /[

          • Krogoth
          • 9 years ago

          Not necessary, Intel has made a number of flops when it try to going against market and later went with it.

          RDRAM, Netburst architecture, VLB, Itanium…….

    • Duck
    • 9 years ago

    Light Peak can carry a whole bunch of stuff like PCIe and HDMI and USB2 all down one pipe at the same time I thought. I’m sure it will be a thread to USB3. External storage can be connected with PCIe or SATA via Light Peak. While keyboards and mice have no need for USB3.

      • mutarasector
      • 9 years ago

      That’s the problem – while Light Peak may be able to host other (multiple) serial streams simultaneously, the disparity in bandwidth requirements and power supplying those various types of interface devices is going to be a bit problematic for one, and companies like SIIG, Koutech, StarTech will probably be the first ones to make LP ‘hubs’ or switches, and/or LP-to<insert favorite serial interface> devices. Now, they also have to provide the power supplies to power those devices, in effect just moving the current underpowered USB/SATA serial ports (currently *in* the chassis) to *outside* of the box, only with some of the same issues.

      PC technology seems to be coming full circle back to the Amiga/Atari ST days of composite computing devices with daisy chained or external peripheral devices. Factor in SFF/mini type of systems are outgrowing in popularity of desktop systems, the LP power problem becomes even more pronounced, as it would with, say, an Atom powered mini or slate/pad devices.

      Now if LP could be used to chain a bunch of low powered devices to provide a means for ad-hoc ‘hive’ clusters synching multiple Atom/Brazos based system CPUs for some live LAN party, or the weekly get together at a Panera’s Breads WiFi hotspot, that *might* be useful <g>.

    • Farting Bob
    • 9 years ago

    USB really needs a more elegant solution than 2 (or occasionally 4) port controller chips duct taped onto a motherboard. Also, just how busy is USB-IF? All they have to do is make sure that controllers meet all the specs that have been widely available for a good while now, and so far only 1 such chip seems to be in the wild, at least on consumer boards and exp cards. If they want to push the adoption of their standard they need to put a bit more effort into it.

Pin It on Pinterest

Share This