news the 10gbps usb 3 1 specification has been completed
News

The 10Gbps USB 3.1 specification has been completed

In January, the USB 3.0 Promoter Group announced a version of the USB standard with twice the data rate of USB 3.0. Earlier this morning, the organization revealed that the specification for the new standard has been completed.

Officially dubbed USB 3.1 (called it!), the latest standard doubles the peak transfer rate to a cool 10Gbps. Also, in keeping with USB’s long and illustrious tradition, it retains backward compatibility with previous versions of the standard. Here’s the spiel from the press release:

SuperSpeed USB 10 Gbps uses a more efficient data encoding and will deliver more than twice the effective data through-put [sic] performance of existing SuperSpeed USB over enhanced, fully backward compatible USB connectors and cables. Compatibility is assured with existing USB 3.0 software stacks and device class protocols as well as with existing 5 Gbps hubs and devices and USB 2.0 products.

At 10Gbps, USB 3.1’s peak transfer rate will be comparable to that of Thunderbolt—the current version of Thunderbolt, anyway. Intel unveiled the even faster Thunderbolt 2 in April, announced its official name in June, and expects the first Thunderbolt 2 products to be out by the end of the year.

The USB 3.0 Promoter Group’s announcement doesn’t say anything about when the first USB 3.1 products will turn up. For the record, though, the USB 3.0 spec was finalized in November 2008, and we got our hands on some of the first USB 3.0 hardware about a year later.

0 responses to “The 10Gbps USB 3.1 specification has been completed

  1. Using less wires in cat5 isn’t the same as increasing the frequency or bandwidth. The only time the wires don’t function is if they aren’t present or are physically cut. Essentially it already uses multiple links (even if they’re physical).

    So back to my original point, there is definitely a lot they could do to make ethernet faster.

  2. I like the idea of having a laptop that I can take to work/class that doesn’t sacrifice size/weight/portability/battery life for something like a GPU that I’m not even using,

  3. USB 4.0 with passive cabels with optical fibers for the data and copper to deliver massive power would be sweet. And the cables would probably still be cheaper than today’s TB cables.

  4. TB has nothing to do with USB. It is an extension for the pcix bus and a conduit for display port data.

  5. Yes, its done with DisplayPort and HDMI automatically. They change the frequency depending on the link quality. It could be done with Ethernet too. It would enable a bit of a speed boost, maybe double the speed. They just don’t seem to have many speed variances. Its basically 1/4, 1/2, 1 speed.

    Ethernet kinda does it but they’re a bit more extreme going 1/10, 1, 10 speed.

  6. Aye, that would be most analogous of how wireless functions, but I was really referring to any sort of adaptive technology as a way of dealing with distance. Turning the bandwidth up or down based on link strength and interference isn’t a terrible way to do things. You aren’t limited to running multiple signals at the same time, but it’s entirely possible for them to do that in this case too. It would just need to make sure the data is in order before feeding it to the programs. If it can be done on wireless, it can be done on ethernet too.

  7. Yeah, they change the frequency up and down on these single frequency links like TMDS or Displayport etc. You were referring to wireless or DSL links where there is multiple frequencies that just get disconnected when they get too noisy. This isn’t suitable for single frequency links to be changing the frequency constantly depending on line noise. That’s why long distance, single frequency links don’t really exist.

  8. You should read up on ConnectX VPI by Mellanox. Supports Ethernet and Infiniband. The good thing about Infiniband is you can just talk pretty much directly to the card. Send data completely bypassing TCP because Infiniband doesn’t use it and is superior to it 😀

    So it has extremely low latency and low CPU usage. Just no software is coded to send packets that way — only TCP or UDP packets.

  9. Yep, I just moved a 4.4gig file to USB3 ext HDD in 16seconds.
    I have a Asus mobo with the 170% USB 3 boost though…..

  10. That cant happen forever, the only reason tb didn’t take off is because its expensive to implement. I would much rather see a usb 4.0 spec at 20 gbps that requires passive cables but isn’t hampered by backward compatibility. I’m sure it can succeed, especially if it can provide some power over the cable, like 60 watts perhaps.

  11. Thats because most 10GbE cards provide a lot more than 10GbE, they have iscsi acceleration, sriov support, toe, and a host of other technologies. They are also meant to run 24/7 transferring data all day and all night. There is more to a high speed network card than a usb controller.

  12. Not in non-fuck-off expensive ones.

    The price is coming down, slower than I’d like to, but it is going down. Why? I have nothing I can point to as a concrete answer but I somehow doubt its related to the tech itself.

  13. Ah and you haven’t heard of anything that changes frequency up or down…? I know of one extremely advanced part of the computer that has been doing it for quite a few years.

  14. They’ve been on server motherboards for awhile…

    Per my post, it’s been out for over a decade and been in a copper form for half that. Two years in the future could’ve been two years ago and it would’ve been just as relevant, he’s questioning why it hasn’t happened and I’d like to know as well.

  15. Most of your list is already excluded, from the mini-ITX mobos that is. About 10 years (or even more?) after VIA created these tiny boards, they started gaining use, to the point that nothing large makes sense for the most people.

  16. IIRC PCIe is also fairly high latency and GPU’s tend to be relatively insensitive to latency as well. GPU RAM tends to have hundreds of cycles of latency for instance.

    I also do not have any hard numbers either though and I’m too lazy to bother trying to compare them ATM but my “gut” reaction is to say latency isn’t the issue.

    I would point out though that USB 2.0 to DVI and other such connectors have been around for a while:
    [url<]http://www.amazon.com/Plugable-UGA-2K-A-2048x1152-1920x1080-DisplayLink/dp/B0038P1TP4/ref=sr_1_1?ie=UTF8&qid=1375420819&sr=8-1&keywords=usb+to+dvi[/url<] I've never used one but for office work they're probably OK. Wouldn't want to try and game over one though.

  17. My old nForce4 SLI board can write large files to my external HDD at 35MB/s and read at 36-38MB/s. It’s actually a bit faster than my x79 board.

  18. Thunderbolt cables require built in transceivers, 1 at both ends of the cable.

    They’re there to improve the SNR through the cable, and work well in that respect, but also cause the cable to be very expensive. These transceivers are also part of the spec so all Thunderbolt cables use and require them which means Thunderbolt cables will always be fairly expensive.

    Personally I’m still concerned about their durability too. I haven’t seen any public reports on the failure rates but I’ve used “active” cables before in the past and they were a nightmare due to being so delicate. Perhaps that experience has made me paranoid though.

  19. Sure right now. The whole point of my post was to point out the price is dropping over time. Give it a year or 2 and the price will be half that or less.

    Also most wireless add in cards chew up plenty of CPU cycles as well. There is no getting around the fact that higher bandwidth devices will demand more CPU time even if what they’re doing isn’t all that computationally intensive.

  20. Who actually cares about external GPUs? If you want a fast GPU for your laptop, just buy one integrated?

  21. 10Gbe is still $260 for an adapter card. The cheapest I could find. Its quite expensive considering the TCP stack is still processed by the CPU and not the card.

    Can get a USB 3.0 card for $20 with two ports on ebay.

  22. It doesn’t change bandwidth like that because it runs at a clock signal frequency with only that one frequency on the line. Multiple frequencies on the same line like DSL, which uses OFDM Orthogonal Frequency Division Multiplexing, or wireless is generally for slower speeds. Where reliability over a noisy transmission medium is required. They transmit multiple channels over a carrier frequency … the channels are basically lower frequency transmissions that are reliable.

  23. 10GbE is starting to become common place on server mobo’s. Even multiple 10GbE ports. Its working its way down to the consumer and small business its just taking longer than “normal” to get cheap enough for some reason.

    Also its worth pointing out that the wireless standards typically get nowhere near their rated specs in actual use for various reasons. They may be good enough for the avg. web browsing user but if you actually need 1Gbps out of your wireless 802.11ac router you will be sorely disappointed. Particularly if you want that 1Gbps at the lowest possible latency.

    I’m still using a “old” 802.11n Netgear wndr3700 which is supposed to get around 600Mbps over its wireless connection. I don’t think I’ve ever broken 60Mbps in real world use in my house.

  24. It’s not the bandwidth. It’s the latency. TB, if I remember rightly, has relatively low latency; USB has relatively high latency (note the use of the word “relatively”.) The difference between a built in (or connected to PCIe) graphics adapter, and one on the USB link, would be very noticeable, at least for games. At least, that’s my expectation; happy to be proven wrong.

    For typical office type work, it probably wouldn’t be an issue, but then, why would you need external graphics adapters for office type work? (Blatantly ignoring CAD/CAM, as that isn’t typical.)

  25. No, no, no. TB 1.0 has two 10Gbps channels, each with full duplex. The problem is that you can’t bond the channels, so any one device will only see 10Gbps instead of 20. Couple that with the fact that usually one channel is dedicated to DisplayPort traffic, and the whole setup is limited to 10Gbps excluding monitors. TB 2.0 allows using both channels at once for a single device.

  26. I’ve rarely ever seen a hard drive come to the upper-end of that 42MB/sec via USB2.0 you quote, but still that’s about half as fast as I write to my external USB 3.0 drive and about 1/3 of the read speed, assuming sequential transfers.

  27. It’s what external HDDs did on USB2, and that’s all that matter isn’t it? I don’t care about the theoretical throughput, only what’s left once I’m transferring photos or something and I’m having to wait on my computer before I can leave. Same goes for 802.11G for example, 54Mbps effectively boiled down to 2.5MBps plus change, which is why we were hurting for -N for so long.

  28. He wasn’t talking about shortcomings of 10GigE, rather the fact that it’s been around forever and not been adopted by motherboards. You know the whole part where 10Gb was ratified in 02 and put on copper in 06, all the while through the infancy of 802.11 and USB.

    If run distances are a problem why don’t they make like a super version of gigabit? Where it senses run distances and shifts bandwidth up or down. So if it’s like a 10ft run, it would run at like 1Tbps, where as if it’s a normal run it would drop down to 1Gbps to prevent cross talk and interference. You know the adaptive technology that’s been put into wireless for the last decade.

  29. But the cable isn’t the issue is it? And TB is going to do it with optical cables. its basically a signal processor applied to a port, and even Ethernet over optical would be an option (even with RJ45 to optical transceivers at each end), except its freaking expensive for the ports themselves. Inifiniband can be gotten for cheap (I’m looking at a pair of cards and cable for under $250). But given the move to NAS devices in the home, there is an argument to be made that 10GbE would alleviate more bottlenecks and improve perceived performance (less concurrent connections) even if it was only the switch-to-NAS link.

    Just seems like it shouldn’t be so wildly out of skew with other high speed connectivity technologies.

  30. I’d be OK with the reduced profile of the thunderbolt port for internal items, too.

  31. Higher bandwidth = very good, few will complain. What about the latency? One of TBs advantages is lower latency (making it suitable for video), is that addressed at all with USB 3.1?

  32. Another step forward and you get to keep all your devices you know and love. This is the way it should be done.

    I don’t think USB 3 needs to worry about competing with TB though. Intel shot itself in the foot when they made a peripheral protocol that was overly expensive to implement and not compatible with everything else on the planet.

  33. Well, most of old sata cables are perfectly fine for new sata, but not all. I expect the same situation here.
    The problem will be in controllers. Bandwidth + 128b/132b encoding.

  34. The conspiracy continues… At least we have wireless that goes faster then 1Gbps now… >>;

  35. Posts like this seem silly. Even if there is no reason for extra bandwidth (which there is), I don’t understand why you would look down your nose at advancements in technology. Like that’s the whole premises of technology in general, advancements. The whole ‘get off my lawn’ and ‘What I have is good enough’ doesn’t play into progress and innovation in the tech industry. It’s actually a good way to stagnate things (see 1080p, consolization).

  36. Hmm. Even switching to USB 3.0 can cut Windows install time by half, ~6 minutes to a little over 3…

    Transferring multi-GB files to a 16GB USB 2.0 thumb drive can take a considerable amount of time too, and those capacities are fairly commonplace these days.

  37. Why is it possible to put 10Gb ports for USB and Thunderbolt onto motherboards for a relatively inexpensive amount, but 10Gb Ethernet is still ridiculously expensive despite having been out for years already? I guess dealing with latency might be an issue, but jeez, isn’t networking protocols stuff already well solved, so its just media and controller issues?

  38. So I guess now would be a good time to revamp the layout of the antiquated motherboard? No more usb1/2/3, no more vga, no more ps2, no more pci slots……….?? plenty of room for decent on board sound with a good amp built on it.
    Wonder what the effects will be on SSD performance?

  39. Reducing the length of the cable might not be enough. The connectors themselves could have bad enough discontinuities that reflections may prevent running the link at 10Gb/s.

  40. 480 Mbps = 60 MBps.

    Is there that much overhead that effective rate is only 25 MBps? Oh ew, I guess so:

    According to a USB-IF chairman, “at least 10 to 15 percent of the stated peak 60 MB/s (480 Mbit/s) of Hi-Speed USB goes to overhead—the communication protocol between the card and the peripheral.”

    Typical hi-speed USB hard drives can be written to at rates around 25–30 MB/s, and read from at rates of 30–42 MB/s, according to routine testing done by CNet.

  41. Ehhh, pretty much every piece of cat5e I’ve ever looked at could do 1Gbps. Did you mean non-e, regular cat5?

  42. Look into a USB 3.0 to SATA/PATA adapter. I’ve got one that does both 3.5″ and 2.5″ connectors on the PATA side.

  43. Car analogies aren’t even tolerated ’round these parts…

    And you go and throw an [i<]anime reference[/i<] into the fray!? Madness.

  44. It should but motherboard manufacturers are notorious for keeping new capabilities for new product so they can use it as a selling feature. Several features for example found on the 990FX chipset were never enabled by motherboard manufacturers on 890FX boards despite them being an identical rebranded chipset.

  45. I don’t remember firewire being “better” in the past because to this day, it is still a pain in the butt to get it work in applications that supposedly benefited from it. Specific devices often require specific controllers, and it still can take numerous firmware updates just to catch up to generic USB 2.0 equivalents.

    Thank gourd “inferior” USB killed it. As USB 3.1+ controllers are integrated into more and more CPUs over time, more devices which have historically been latency bound will be able to drop firewire, Thunderbolt, Ethernet, specialized networking, and even PCIe.

    More inferiority, please!

  46. Agreed. Last night, I wrote a backup image out to a USB 2.0 drive (because I don’t have 2.5″ IDE on my mobo for a direct connect). I just started it and went to bed for the night. Then the race was on to see if it’d get done before I had to go to work today!

    Would have much preferred to come back in an hour, put it in the fire safe, and then go to bed.

  47. Your comment is so depressing. It reminds me of a superhero who has reached the pinnacle of power and has nothing left to do except more of the same, or die a hero’s death. If CPU’s were like Dragon Ball Z, then I guess the Sandy Bridge architecture was Goku turning super saiyan, and every upgrade after that is just some malarkey created to keep the cash cow mooing.

  48. Lol, and you really think that motherboard vendors are not going to take advantage of a new revision of the spec and make it exclusive to their new product?

  49. Wasn’t the initial statement all current USB 3 devices(motherboards especially) should just require a BIOS update to support 3.1?
    If this is not the case I am gonna be VERY unhappy…

  50. Really? I find external drives to be far too slow on USB 2.0 to be of any use day to day. Shoot, I still get rankled that my camera and tablet don’t have USB 3.0.

  51. IMHO, at this point I’m still perfectly ok with USB 2.0.

    Edit – This is my personal opinion, as I’ve said. I don’t own any USB 3.0 external drives yet and my 1.5TB WD Eseentials still has a lot of space left. I connect only my keyboard, mouse, printer and wifi adapter. You guys with USB 3.0 devices go on ahead without me for now.

  52. Yeah, the only thing worse would be going through two PCI Express bridges… :/

    Every protocol has overhead. USB wouldn’t work because its latency is too high (1-2 ms), which is about 50 times too high.

  53. I like the consolidation of external and internal i/o. PCIe for everything internal (even SSDs) and a universal external i/o such as USB and a higher speed more expensive i/o for niche applications (TB2).

  54. Pretty much the reason I would rather they NOT made the new spec a subset of 3.0 instead of just moving on up to 4.0.

  55. Or probably just need shorter 3.0 cables which is more probable, same thing as with ethernet, reduce the length of the cable and you can get a cat5e cable to support 1Gbps.

  56. They already have, it was called the 1990’s and 2000’s. This is just peripherals doing their best to keep up.

  57. You need new cables for usb3 too, i don’t get your point. You potentially need new (and by that i don’t mean expensive) cables for full hdmi 1.4 compliance if your cables are really really really old.

    New cables may just mean the necessity of more rigorous testing standards and verification of the older cables to make sure they fully adhere to the new specification. Same kind of crap cat5e and up should be.

  58. TB 2.0 unfortunately is just a rehash of TB 1.0, same total speeds but now it’s fully duplexed or something.

  59. Probably won’t be enough if you add in USB over head. Real world bandwidth would be less than half of a PCIe 2.0 4x lane connector I think.

    Thunderbolt 2.0 would probably do the job just fine but will also probably not ever be cheap enough to be used.

  60. That’s why ford discontinued the the thunderbird. Fortunately the Corvette is still going strong.

  61. It won’t. Having to go through two bridge adapters would be a nightmare for a high performance hardware.

  62. I don’t know for sure, but “uses a more efficient data encoding” should reduce CPU/controller usage as well as how many bits are actually pushed down the wire, right?

  63. Alright, nobody’s asked it in this thread yet so I will:

    When will USB start making sense for eGPU/high performance applications? USB has historically operated on a much higher level than PCIe/Thunderbolt, with plenty of overhead and CPU usage, but at what point can the raw transfer speed overcome all that?

    Everyone heralded Thunderbolt as the end-all eGPU interface, but given its expense, OEM implementation difficulties, and lack of general availability/acceptance, can USB fill the void?

    I don’t know how much overhead USB 3.0/3.1 has compared to USB 2.0, so I’m hoping someone can help me answer any part of the above.

  64. Here comes the stream of poorly implemented controllers that are plagued with stupid issues…..

  65. USB really has come a long way. Now can we get CPU’s to increase in performance at this rate?

  66. [quote<]I think [i<]Thunderbird[/i<] is facing a real problem now. [/quote<] Thunderbirds are GO!

  67. [quote<]over enhanced, fully backward compatible USB connectors and cables. [/quote<] In other words, you need new cables for USB 3.1.

  68. I think Thunderbolt is facing a real problem now. Remember when Firewire 400 and 800 were better, but nobody used them? Unless TB goes from expensive to free, TB will go the same route that FW went.

    Sorry about that. I meant Thunderbolt, not Thunderbird.