Feeds

back to article USB accelerates to 10 Gbps

Universal Serial Bus, the connectivity standard so ubiquitous the world has long stopped caring about the derivation of the USB acronym, has just been upgraded to 10 Gbps. The upgrade comes in the form of a .1 release, USB 3.1, that is backwards compatible with USB 3.0 and 2.0 kit. The new speed will only be achievable with kit …

COMMENTS

This topic is closed for new posts.
Anonymous Coward

SuperSpeed+, what is next, ultimate speed and then ultimax?

There have been many technologies that have been better than USB. The next question, when will Intel support USB 3.1? With Intel having the biggest share of the processor market, they can dictate when it becomes mainstream. If they decide to wait a few chipset generations, then many computers will not have USB 3.1

0
4

I agree that USB standards naming has been a mess for the longest time.

But the adoption of this standard by Intel is not that important. its rather easy to change the chip that provides (extra) USB3 ports on a lot of motherboards for a 3.1 chip.

maybe Intel is a contender for the firm that produces most processors, but intels share in the total processormarket is far from the biggest. if ARM adopts USB3.1 in their designs it will become mainstream rather fast.

4
1
Go

>SuperSpeed+, what is next, ultimate speed and then ultimax?

USB Plaid.

13
0

USB Sonic?

No screwdriver required.

It's so fast it hums?

0
0

Obvious marketing name

Plaid speed

1
1
Anonymous Coward

Which costs money that the manufacturer will need to recoup. Intel has USB 2.0 support in their latest chipsets. For a manufacturer to put USB 3.1 in, they will need to use a PCIe interface and install another chip. For many machines, they are made as cheap as possible and thus, they will wait for Intel to offer USB 3.1 support.

2
1
B-D

Cheap As Chips

Absolutely, the fiasco surrounding the 1st gen Marvell SATA3 controllers that were tacked on to X58 family motherboards because Intel hadn't gotten SATA3 integrated in time proves that point.

Sure, they had SATA3, but it performed only marginally better than the integrated Intel SATA2 it was meant to usurp.

There is also the mobile computing market where less is more and separate controller chips are very much undesirable, for that market it is SoC or GTFO.

0
0
Silver badge

There's no reason for ARM to adopt even USB3 support. Their processors would struggle to keep up - they aren't made for speed. They are made to be compact, require minimal supporting components, and achieve a very impressive level of energy efficiency.

1
8
Stop

Erm, some ARM procesors absolutely are made for speed. Quad core parts running at close to 2GHz are not exactly sluggish.

8
1
IT Angle

Even faster?

In Sweden, I saw a 3 telecom advert saying "Max Farts" (sic) - apparently farts in Swedish means something like speed... so maybe USB 3.x could be max farts USB....

3
0
Silver badge

what is next

USSB? then USSSB, U4SB...

0
0

Re: what is next

Reboot - then it will be:

USB ONE

0
0
Anonymous Coward

Yes, but USB doesn't do DMA, it requires the CPU to help it along. That means battery life and responsiveness of the device suffers.

2
0
Silver badge

Or borrow a page from the iPad and just drop the versions altogether.... 'The New USB'

1
0

Ludicrous Speed of course.

http://www.youtube.com/watch?v=ygE01sOhzz0

0
0
Anonymous Coward

DMA

Isn't the DMA aspect of Firewire a big security issue?

0
1
Silver badge

Re: DMA

>Isn't the DMA aspect of Firewire a big security issue?

That depends if you trust what you are connecting to.IP/Firewire or IP/TB is a bad plan for general networking, but would be good for building a high-speed cluster interconnect for firewalls, load balancers etc.

You just treat it as a PCIe extension. It would be quite nice to see a switch which emulates 10G ethernet cards, so TB (or better, Lightpeak) gives you your physical link to a virtual 10G ethernet interface which is in the switch. It would be interesting to see how that would stack up against "normal" 10G ethernet.

1
0

Potential

Pretty cool potential to replace CAT5-based networking, for those of us that still dabble with wiring.

0
3

Re: Potential

Erm, maybe not. Can you say latency? Overhead?

I'll stick with Ethernet and cat6, thanks.

7
0

Re: Potential

Admittedly those technical details are beyond my scope, that's why it's "potential". The bugs need to be figured out by somebody a bit more qualified. Theoretically a router/hub/switch should work all the same way, the only thing that's really changing is the type of wires and connectors involved, you just no longer need a dedicated network plug.

0
0
Gold badge

Re: Potential

It's harder than you think. USB is a master-slave protocol and if you want the slaves to be able to send traffic you need to either statically reserve bandwidth for them or poll them. It's not like firewire, which was a genuine bus (which is why Windows would happily offer network services over firewire, but not USB). I suppose it is just a simple matter of programming, and I suppose token ring networks demonstrate that you can achieve reasonable performance even if your protocol includes a "speaking stick", but I think it's quite a lot of programming.

6
0
Silver badge

Re: Potential

Even ethernet has too much latency for some applications. If you want low-latency, you use infiniband. Costs a fortune though. It's used for cluster interconnects, and I read that some high-speed trading operations are asking for infiniband connections now because 10gig ethernet just has too much latency for them.

0
0

Re: Potential

It turns out that running network protocols over USB is already used. Many 3G USB "modems" are actually routers, and also run a small web server on the device to control the router functionality. However I agree with the general point that replacing physical Ethernet cables with USB is not obviously a good idea.

3
0
Silver badge

Re: Potential

And USB wireless adaptors were popular in the early days of wifi.

0
0
Bronze badge

Re: Potential

Pretty cool potential to replace CAT5-based networking, for those of us that still dabble with wiring.

Hardly. USB3.0 is already limited in practice to around 3m cable length. Hardly practical for a LAN. Then you have the issue of the differing device types - i.e. host and peripheral - that a simple cable can't compensate for. That's without even considering the lack of true interrupts (no, high speed polling doesn't substitute), DMA, and all the other niceties that make a PCI/PCIe ethernet adapter work so well without the huge CPU hit USB implies.

0
0
Silver badge

Re: Infiniband

20Gig Infiniband PCI-E Controllers are around £30 on Ebay. Hardly expensive. 40G are a lot more expensive.

Current state of the art is 56Gig. That is mega expensive but still less than 10G Ethernet.

0
0

interference problems

I'm left wondering if the new standard does away with USB 3.0's interference with 2.4 GHz and 5 GHz frequency bands.

As it is, I can barely use my Wi-Fi and wireless mouse whenever I connect a Superspeed USB drive into one of the blue ports.

1
0
Silver badge
Unhappy

Re: interference problems

I doubt it :(

In a world where everything is build down to a price, and the likes of Ofcom don't care about end user or public good but only licensing fees, we should not expect any radio gear to work at all well.

0
0
Anonymous Coward

10 Gbps

... while in real world, when I'll plug a USB 3.1 stick into my laptop...

buy hey, who says all planes must fly like a blackbird?

1
0

Ludicrous USB speeds

5
0
Happy

Cables

At least it wont cost £25 for a cable like Thunderbolt

5
1
Silver badge

Re: Cables

It will if you buy it in PC World/Currys.

2
0

Hopefully being a point release the work to support it in existing USB 3.0 designs is very little, and we will see USB 3.1 supported in a lot of next year's SoCs, APUs and chipsets.

0
0
Silver badge
Joke

10Gbps, wow that's fast. But, how fast is it in classified documents per second?

3
0
Silver badge
Boffin

Tsk, you should know El Reg's official measure of speed is the kilowrist of pr0n movies (at least, until UK censorship is implemented):

http://www.theregister.co.uk/2008/11/12/arizona_boffins_grasp_fat_pipes/

1
0

Aaah

.... My wireless mouse stops working when plugging in usb3, assumed was driver issue between usb3 and windows 8

All new kit and i7, couldn't find much on it... Thanks for info!

0
0
Holmes

Aaah

.... My wireless mouse stops working when plugging in usb3, assumed was driver issue between usb3 and windows 8

All new kit and i7, couldn't find much on it... Thanks for info!

0
0
Anonymous Coward

Re: Aaah

http://social.technet.microsoft.com/Forums/windows/en-US/e43de6f5-4ad5-4f09-8021-92aefb00480d/windows-8-usb-30-problems

0
0

Too soon

10Gbps isn't enough of a leap.

0
0
Bronze badge

excessive

It might be just me, but off the top of my head, I can't think of anything that would actually be able to use that sort of throughput. Surely it's just moving the data bottleneck to somewhere else.

Can any disks write that fast? Even if they can, if you have an external disk, the connection to the host PC might be USB, but the actual connection to the disk is still likely to be SATA, which is slower, right?

I might be missing something really obvious, but other than external disks (and networking, bust as mentioned above networking over usb has it's own issues), I cant really see the need for that sort of throughput on an external bus.

Printing? Mouse? Keyboard? Web cam?

Just because everyone loves a car analogy. To me, it seems like having a car that can go 500mph. Great, you have the fastest car in the world, but there is nowhere you can drive it at that speed.

0
0
Bronze badge

Re: excessive

SATA 600 is the current de facto standard for SSDs, but there are faster interfaces around, such as PCIe.

For comparison:

USB 3.0 = 5 Gbps = 640 MBps

USB 3.1 = 10 Gbps = 1250 MBps

A current individual SATA SSD is limited to 600MBps due to the SATA 600 interface. So yes doesn't really need USB 3.1.

But an PCIe SSD is a lot faster, as an example, the OCZ RevoDrive 3, hits 1,000MB/s read, and 925MB/s write. So that type of speed fits nicely within USB 3.1.

So we just need someone to make a USB 3.1 to PCIe caddy :-)

0
0

Re: excessive

Graphics maybe - is there any reason monitors couldn't be made to connect to USB ports?

0
0

Re: excessive

Those already exist

0
0
Anonymous Coward

Super speed and super

Wow! USB's almost caught up with much longer haul fully peered Ethernet! What an achievement!

2
1
Anonymous Coward

Does it have DMA or will it be a CPU hog?

0
0

This post has been deleted by its author

This topic is closed for new posts.