USB 3.0's New Jacks and Sockets

The Register has a brief look posted (with photos and diagrams) of "USB 3.0, the upcoming version of the universal add-on standard re-engineered for the HD era, made a small appearance at the Consumer Electronics Show (CES)." The posting explains that USB 3.0 "wasn't demonstrated in operation, but we did get to see what the new connectors look like." How does it handle backward compatibility? The extra pins needed for USB 3.0 "are placed behind the USB 1.1/2.0 ones. USB 3.0 connectors and receptacles will be deeper than the current ones."

Is it burst speed?

[danomac (1032160)]

I wonder about the new speed specification... in my experience even with no other devices on the USB bus getting 480mbit was impossible. I always had to resort to firewire for my drive caddy because I got consistent results with it.

I sure hope they've addressed this issue. The OS caching helped, unless you wanted to unplug the damn thing right away - then you had to wait 5 minutes for the cache to flush out.

Re:Is it burst speed?

[gillbates (106458)]

Yes, and no.

You see, 480 Mbs is the electrical interface speed. As in, 480 Million bits go across the wire every second. Not all of those bits are used for traffic.

However, some of those bits are used by the overhead of the transfer protocol. You've got USB packets in the stream which do nothing but reserve space for some psuedo-realtime device which might be connected to the bus at any second. Whether or not the OS/USB Controller allocates these blank packets even in cases where they aren't needed is a matter of programming.

As an aside, I've noticed that on the same computer, with the same flash drive, Linux does a much faster job with file transfers than Windows. I suspect Windows is just under-utilizing the bus, to make it easier for their engineers. But I could be wrong, as I haven't looked into it in detail.

Re:Is it burst speed?

[jubei (89485)]

Your windows file copy slowness could be because Windows does not use write caching for removable drives. This allows clueless users to just yank out the disk without unmounting properly. If you are getting slow reads, that is a different story.

Re:Is it burst speed?

[letxa2000 (215841)]

In case a "clueless user" yanks it "without unmounting properly?" Excuse me, but I don't think that's a matter of the user being clueless. If I have a removable drive, I don't think it's unreasonable to be able to remove it at any time--the OS should expect that. If the OS is still writing data to the drive and there's some kind of window open to that effect, then I'm stupid for disconnecting it in the middle of the process. If I "finished" copying three minutes ago, I don't think it's unreasonable for me to be able to disconnect the drive.

This is why Linux is a great OS for a server but not so hot for the desktop. Write-caching for a USB drive might make sense on a server, but not so much on the desktop.

Re:Is it burst speed?

[DaleGlass (1068434)]

You can disable caching on Linux with -o sync.

However, neither that, nor what Windows does will prevent damage on a FAT32 formatted device, because the filesystem isn't made to deal with that. And even for a filesystem like ext3, reiserfs or ntfs that will not corrupt itself in this case, you'll still lose data if you yank the drive while a file is being written. Windows will warn you if you yank the drive without telling it to disconnect the drive precisely for this reason.

Really the only way of dealing with this perfectly is making the media impossible to disconnect until the filesystem is dismounted orderly. This can be done with CD and tape drives, but isn't going to work with anything connected to an USB port.

Re:Is it burst speed?

[zobier (585066)]

Really the only way of dealing with this perfectly is making the media impossible to disconnect until the filesystem is dismounted orderly. This can be done with CD and tape drives, but isn't going to work with anything connected to an USB port.

It could do if you wanted it to, there's these two little holes on the USB connector that a latch could engage during transfer.

Re:

[MMC Monster (602931)]

Seriously annoying. Is it that difficult to have (name your distribution here) Linux do an automatic sync every couple seconds when a USB drive is connected.

Waiting 2 minutes to unmount without any progress meter is just broken UI design.

But...

[Junta (36770)]

syncing every few seconds isn't the good 'fix'. Using mount -o sync on obviously transient mounts would be the 'fix' to the problem described. If you sync ever so often, there *still* is no good way to track it/indicate it, while the fs being mounted with sync means the file write operation itself can be tracked more accurately.

Of course, the line to draw at what is 'obviously' transient may be hard, but I think 4GB and under and USB connected is a good rule of thumb today of transient sticks vs. persistantly attached usb storage. When you get into the realm of 'guessing' the intent of the user implicitly, things get hairy.

Oh it'll go at full capacity

[SuperKendall (25149)]

You'll just need to dedicate six of your eight cores in the 3GHz Intel chip for the process overhead.

shades of future past

[pilgrim23 (716938)]

-Little fingers inside existing fingers to work with legacy USB devices... Does anyone rememeber the EISA slot standard designed to allow inserting a ISA card?
Now all we need is a MCA driver and we are in busienss for the new world of 1992.

Will it work on Linux?

[Marcion (876801)]

Is the software side of USB an open specification or some members only, pass the royalty thing that the open source world will have to take the next ten years reverse engineering?

Other Fixes

[MBCook (132727)]

Oooh. It's faster. Wow. Didn't see that happening.

Did they fix the CPU overhead? Did they make a P2P version so that I don't need a computer to connect a camera to a hard drive and have it work? Basically, did they do anything to improve it for high-bandwidth applications (which is obviously what they're targeting) compared to FireWire?

The cable worries me some. I understand the drive for backwards compatibility, but it seems like they should make the cable more obviously different. It just looks like it will be too easy to accidentally use a USB 2 cable, not realize it, and then wonder why the device is running so slow. Just a little nub on the bottom of the connector would do it.

Re:Other Fixes

[Svet-Am (413146)]

Did they make a P2P version so that I don't need a computer to connect a camera to a hard drive and have it work?

Yes, they did. Several years ago, in fact. It's called USB On the Go [usb.org]

And it doesn't work

[SuperKendall (25149)]

I bought a USB OTG external hard drive that is supposed to be able to copy files off a slave device, and a box that is supposed to support two master devices and initiate copies between them - neither work at all with any USB storage I have tried.

USB OTG is a farce.

Re:

[tlhIngan (30335)]

Did they make a P2P version so that I don't need a computer to connect a camera to a hard drive and have it work?

Yes, they did. Several years ago, in fact. It's called USB On the Go

Actually, it's not strictly P2P using USB OTG. One device is still the host, the other the client. It's just there's a complex protocol they can go through (Host Negotiation Protocol) to switch roles if necessary. Of course, both sides have to support OTG.

Also, there aren't many devices out there that are actually OTG complaint.

Re:

[dfn_deux (535506)]

Nothing complicated about it, ground out one of the wires and the port acts as host, let it float and it acts as a device. Only limitation it has that doesn't exist as part of the regular USB standard is the available current is only half (IIRC) of the regular usb standard. Some usb chipsets allow the switching to be down with software instead of using special cables even, something like 'echo "host" >/proc/usb/0' or somesuch, check the internet tablet forums to see how the Nokia IT users are already mak

Just goes to show...

[by Anonymous Coward]

... that longer male connectors are better.

Re:Just goes to show...

[4D6963 (933028)]

longer male connectors are better.

Nooo it's not how long they are but what they do. Besides if female connectors like long male connectors bigger that's because they themselves are *too* deep. A short male connector fits a "shallow" female connector as nicely as a long male connector fits a deep female connector.

So girls, quit complaining and laughing and get it worked out! Oh wait, oops..

Then what's the point of SATAII?

[Kickboy12 (913888)]

Funny how I just upgraded to a new computer that uses SATA 3.0Gb/s. If USB3 is faster than SATAII, then why not just use that for drives? Not that anyone ever really maxes out SATAII to begin with. So it's all kind of useless in the end.

Re:Then what's the point of SATAII?

[Midnight Thunder (17205)]

Funny how I just upgraded to a new computer that uses SATA 3.0Gb/s. If USB3 is faster than SATAII, then why not just use that for drives? Not that anyone ever really maxes out SATAII to begin with. So it's all kind of useless in the end.

The problem with SATA, IMHO, is that makes a shoddy external connector. There is no notion of hubs or even daisy-chaining. USB and Firewire both support hubs, whereas Firewire supports daisy-chaining. With SATA you need as many external SATA sockets on your computer as you have external SATA drives. If your main computer is a portable, then this is a poor solution.

Re:Then what's the point of SATAII?

[RobFlynn (127703)]

I've seen SATAII hubs before.

Here's one that I found with a couple seconds of googling: http://www.cooldrives.com/sahub5muussi.html [cooldrives.com]

The point of difference busses

[l2718 (514756)]

Because there's more to a bus than the bandwidth. USB has a lot of overhead (it can be branched, hook many devices etc). SATA is dedicated for controlling storage. That's why we put cameras on the USB, hard-drives on a SATA bus, the network card on the PCI bus, video card on the VESA bus ...

Naming

[teslatug (543527)]

So they're going with a 3.0 instead of some crazy More Full Speed (TM) name this time?

Probably not

[XanC (644172)]

They'll come up with that later. USB 1 had two data rates: "low speed", 1.5Mbits/s, and "full speed", 12Mbits/s. USB 2.0 added "high speed" at 480Mbits/s. No idea what superlative they'll reach for this time.

Re:Probably not

[RobFlynn (127703)]

Super Mega Ultra USB Supreme Ranger... CONNECT!

Re:Probably not

[by Anonymous Coward]

Super Mega Ultra USB Supreme Ranger... CONNECT!

Private: Captain EHCI, unknown device in range!
Captain: Run level 9 enumeration protocol, stat!
Private: Error -123: device is not accepting our address!
Captain: Arm the compatibility layer and reset the host controller!
Private: Device is a Super Mega Ultra class AHCI master!
Captain: Increase port voltage to 480V and reenumerate!
*ZZAP*
Private: Reporting overcurrent condition on port 5, and Super Mega Ultra device running in low-speed compatibility mode!
Captain: Roger that, commence loading driver modules.

Re:Probably not

[phallstrom (69697)]

Ludicrous speed!

There's really no other option...

Re:Probably not

[sconeu (64226)]

But then they'd have to make all the connectors plaid.

Re:Naming

[Brobock (226116)]

So they're going with a 3.0 instead of some crazy More Full Speed (TM) name this time?

FTA:
Dubbed SuperSpeed USB, the third major incarnation of the serial bus standard is set to deliver data transfer speeds of around 4.7Gb/s - ten times today's 480Mb/s limit.

They haven't TM'd it yet though.

its like DB9 all over again...

[johnrpenner (40054)]

the more things change, the more they stay the same -- now
they're back to using 9 pins to implement the spec -- other than
making the connectors physically different so people don't end up
plugging in old RS-422 cables into it -- from the number of actual
pins needed to implement a spec -- we're physically back to using
9 pins that were available in the DB9 form factor, only this connector
is considerably more difficult to manufacture. :-^

Still using rectangular connectors, I see.

[croddy (659025)]

p>Whatever they come up with, in the end, I have only one wish for the USB3 hardware developers: that they be made to plug 1000 of them in upside down in the dark.

A serious question

[moosesocks (264553)]

What happened to firewire? All signs point to it going extinct in the very near future....

Wasn't it vastly superior to USB? It had a higher maximum throughput that could almost be realistically achieved, delivered useful amounts of power over the bus, and allowed devices to talk to each other. The audio/video features are pretty nice as well....

Both firewire and usb were well-supported on all platforms, so *that*'s not the issue. It's also robust, to the point of being found in many modern aircraft designs and the space shuttle.

IEEE1394c is even cooler, and uses CAT5e/RJ45 for wiring, allowing for automatic negotiation between other 1394 devices, and normal ethernet devices. Max speed is 800mbps, and it very nicely bridges the gap between "traditional" peripherals, and network-attached devices.

So what happened? Did I miss something? Who killed Firewire?

Re:A serious question

[Lumpy (12016)]

Going extinct? huh?? I dont see any pro cameras ditching firewire for usb.

I see sata taking over for external hard drives. I converted all my firewire 800 external drives on my powermac tower to SATA 3 drives last year and gained a crapload of performance at 1/3rd the price. but every HD camcorder that is more than a toy for the masses has firewire on it and will be there forever. Even the hard drive based cameras from panasonic that cost more than most guys' houses still have firewire on them.

Problem is SATA has a failure point. I can have 20 foot firewire cables.. good luck making sata work over 3 feet.

Re:A serious question

[appleguru (1030562)]

Nobody... The problem with firewire is its cost-- USB is, on the device side, dirt cheap to implement. This comes at the cost of needing a host controller (your computer) to do anything and that comes with CPU use overhead. Firewire requires these 'controllers' in every device, making it far more useful (allowing things such as communication without a computer!), robust, and fast without the overhead. But it costs more! And, as we know, price is what drives the marketplace. As a 'normal' uniformed consumer, would you buy a firewire 400 widget for $100 if the usb version cost $50 and both "did the same thing" and ran at a theoretical "480 mbps" (And we all (by all, I mean us on slashdot) know how well usb2 does that...). As a 'normal' consumer, of course not!

Firewire is far from dead, however... Nearly all consumer/prosumer mini dv cameras use it (including hdv cameras), many set top boxes and HDTVs have 1394 links on them for connecting devices (DVHS decks, HDTVs, and cable boxes... this transport MPEG-2 transport streams), and every mac since the iMac debuted has shipped with firewire ports on it (Many, many external hard drives have firewire ports on them.. the good ones anyways ;))... Sony has been shipping 1394 on its vaio computers for ages (in the form of i.link), and all modern computer manufactures have followed suit.

So, to answer your question, consumers "killed" firewire by being... well... price conscious consumers. But in reality it's not going anywhere, and with any luck and all the cool networking capabilities the firewire spec has these days it will eventually catch on with the majority of consumers as a convenient way to interconnect devices and stick around for good.

Re:

[Pecisk (688001)]

No one. Simply Firewire is in pro market, so it gets less attention of mainstream press. Still, FW rules over Apple world (Hard disks, scanners, cameras), and is taking serious inroads on Windows and Linux platforms. Still, USB3 can deliver some blow to posibility that Firewire will come into casual computer user.

Many say that USB consorium is more organised and actually delivers. While Firewire has been promising, it's market has been difficulty to deliver actual results. And also there is simple reason wh

Patents killed it

[lennier (44736)]

"So what happened? Did I miss something? Who killed Firewire?"

Patent royalties, I believe, or at least that's the popular impression: this guy [teener.com] seems to be saying that Steve Jobs attempted to hike the royalty price and though he wasn't ultimately successful, perhaps the mere suggestion that he could was enough to sour third party implementors and move them to USB.

Like with Token Ring vs Ethernet and Objective-C vs C++, the answer seems to be that if there's a nearly-almost-good-enough open technology and a w

Re:A serious question

[DECS (891519)]

Microsoft actually supported Firewire before getting USB complete, particularly USB 2.0, where the two standards overlap in certain areas. While Firewire was invented by Apple, Microsoft also actually delivered support for IP over Firewire first (several years first), although I doubt many people used it on Windows because most PCs that have Firewire only have the 4-pin, non-powered version like Sony's iLink. Mini-to-mini Firewire cables are not too common.

Apple didn't support IP over Firewire networking until around 10.3.5 IIRC. Now that it's there, it is actually quite useful on Macs as a secondary network interface, since all modern Macs have FW400 and many now have FW800 too. Macs also have smart enough firmware to use Firewire in Target Mode, which is a significant feature other PCs won't match anytime soon.

The new FW3200 uses the same connector as FW800, an advantage over the different and more complex USB 3.0 connector.

Another advantage of Firewire is that it provides higher voltage for charging, so it can power more significant devices and can recharge devices faster. It's noticiably faster to charge iPods/iPhone over Firewire. The 30-pin Dock Connector has Firewire compatible pins for charging, even though modern iPods don't support Firewire for data exchange.

There's really no reason for Apple to drop Firewire, and it will be difficult for PC makers to match the features of Macs even when including Firewire ports on their PCs. Not only do BIOS PCs lack any firmware support for target mode use, but Microsoft dropped IP over Firewire in Vista (!). USB 3.0 might bump the speed for new devices, but it doesn't match the Firewire-related features that exist now, and doesn't match the throughput of FW3200, which is also in the pipeline.

Ten Big Predictions for Apple in 2008 [roughlydrafted.com]
What's Apple going to be up to in 2008? The previous article looked at clues from the Newton MessagePad to the iPhone. Here's a look at the potential future of the rest of Apple's businesses, from hardware to software to services.

Re:A serious question

[LWATCDR (28044)]

Well part of it is Firewire isn't a replacement for USB. I haven't seen a Firewire keyboard, mouse, printer, or joystick. Yes it is mediocre all the way around but it works well for some devices that Firewire doesn't work at all for. And works well just okay for many devices that Firewire works well for. Firewire will always be an port you have to get in addition to USB. so it will alway be less popular. But I would agree with you that it isn't dead.

Re:

[Jeffrey Baker (6191)]

This is the stupidest and most-wrong, and also the most commonly believed, reason for FireWire's relative obscurity. Sure, there was a royalty for the FireWire name, but that hardly mattered in the great scheme of things. Early FireWire ASICs cost $50. Even today, a FireWire interface will set you back $20. USB interfaces are in the vicinity of $1. On top of that, your device must implement a complicated peer-to-peer software protocol. This costs even more. So the $1 royalty was only a tiny part of F

One suggestion

[edwardpickman (965122)]

Can they start color coding USB cable types? Some of us old timers have been around since 1.0 was popular. I've got a box full and it's always fun trying to find the 2.0 cable hiding among the 1.0 cables.I hate to toss them but I really haven't any use for 1.0 cables. I'd just love to see some kind of coding system since they all use the same connectors. At least with hard drives every time they change them we get new connectors. It may make them backwardly compatible but it does cause confusion.

Re:One suggestion

[Teilo (91279)]

Umm, you do realize that USB 1.0 and 2.0 use the exact same cables and connectors [usb.org], don't you?

Just asking, because you sound too serious to be joking.

Re:One suggestion

[ffflala (793437)]

That's not true! USB 2.0 requires gold-plated contacts for maximum bus fidelity. Monster makes a good USB 2.0 cable, and it goes for a steal at $79.99 per cable.

If you put the 2.0 cables in the freezer to align the molecules before you use them you get even better bus response. All of my devices have this warmer, more human feel when I'm using properly-designed cables.

Re:

[Jesus_666 (702802)]

Yeah, but you only get a really warm feeing if your USB controller uses vacuum tubes. Seriously, after using vacuum tube-driven USB once, I'm not going back to that silicone crap.

Re:One suggestion

[Feanturi (99866)]

They use the same connectors, but USB 2.0-rated cables are better shielded. A cable made when USB 1.1 was all there was did not have to be capable of carrying as much data. USB 2.0 is 40 times faster than "full speed" USB 1.1, so if you want to ensure you're getting the most out of your device, you want a higher grade of cable.

Still half vertically symmetric

[Trogre (513942)]

Aargh, this connector is *still* symmetrical vertically in form factor but not electrically. Which means you'll have people fumbling behind computers/laptops turning the connectors upside-down until the cable is twisted trying to plug in their camera/mouse/hdd/coffee maker.

Either change the shape of the connector (something like RJ11 would be fine) or make the pins such that it can be inserted right-way up or upside down (figure-eight power cable connectors for example).

Re:Still half vertically symmetric

[MattHawk (215818)]

One good example is the Apple Macbook power connectors. They're palindromic, so even though they're a plain rectangle, they plug in either way (and have a power LED on both top and bottom to accommodate such). They need either that, or a nub to indicate alignment - I HATE fumbling around with USB cables to get them plugged in.

Of course, this would require abandoning backwards compatibility... but seriously, by the time that there are only USB3 ports on a device, I'm pretty sure we'll be past needing to plug 2.0 devices into it, and if we need to use an old device that badly, it would be easy enough to make them electrically compatible such that a simple dumb cable adapter can fit it. Old device standards are passed by for new ones all the time, and clinging to backwards compatibility at the costs of advancement can be a serious mistake - clinging to backwards compatibility at all costs is a significant amount of what's hampering Windows right now, for example.

more pins: yuck

[nguy (1207026)]

This seems like a step backwards. Four pins and shielding was a good number; more makes the cables big and requires more connections on the circuit board.

(o-o-omfg ponies) EXTRA PINS??!?!?

[vectra14 (470008)]

Too bad they're adding the 5 new pins (given, 1 of them is in theory good old GND but still). As an EE, the one thing I liked about USB over Firewire is its physical simplicity... power, ground, and a differential bus. With 1394c heading towards RJ45 it's like USB and 1394 have traded places in terms of physical convenience (I'm sure a number of people have had the pleasure of dealing with ultra-over-engineered (and consequently overpriced) 1394a/b cables and repeaters.. oh the repeaters).

..Not like the host-heavy USB stack made it a much-liked protocol for me in the first place.

What about XP drivers?

[PingXao (153057)]

My gut tells me there will not be any MS-written USB 3.0 device drivers for Windows XP. Artificially making an OS "obsolete" by not providing drivers for new hardware is one way to accelerate the adoption of Vista. The code words that surround this new standard vis-a-vis Microsoft Windows reveal the inclusion of Vista-style DRM (e.g. "the HD era"). With that in mind I see MS declaring that USB 3.0 drivers for XP are technically "impossible" for reasons that will prove bogus. They may have legitimate business reasons for not providing drivers, but those won't be the reasons they trot out in public.

nice reference

[BitterAndDrunk (799378)]

Comedy isn't pretty.