Submitted by USB 3.0 isn’t even available to us mere mortals yet, but thanks to Sarah Sharp’s hard work, the Linux kernel is already underway towards having basic support for the new specification. “Now that the bus specification is public, I can finally talk about the code I’ve been developing at work. I’ve been writing a Linux driver for xHCI (the new USB 3.0 host controller), and changing the Linux kernel stack to support USB 3.0 devices.” Sharp got to demo her work at the USB 3.0 Superspeed Conference.
In her demo, Sharp showed Linux reading a movie off a USB 3.0 storage device, using a prototype xHCI host controller from Fresco Logic. With this prototype, she’s been able to achieve speeds 3.5 times higher than USB 2.0. Once the prototypes are replaced with actual silicon, she expects the speeds to go up, towards ten times as fast as USB 2.0.
With USB 3.0 devices reaching the market as soon as mid-2009, getting support for this new standard into your operating system is kind of important. Sharp explains that a few things need to be changed in the way the kernel handles USB, but these changes are relatively minor and will find their way into the kernel soon enough. The xHCI host controller driver is more problematic, since the xHCI specification is currently bound by an NDA, and as such, Sharp cannot release the code yet.
As for Windows, Microsoft has stated that Windows 7 will support USB 3.0, but they have not yet made any promises regardig Windows Vista and Windows XP. A safe guess would be that Vista will get support for USB 3.0 (seeing how similar Windows 7 and Vista are underneath), but I wouldn’t be surprised to see Windows XP missing out. Of course, vendors may release drivers and support despite what Microsoft does.
There is no word from Apple yet., but the company will probably be forced to adapt USB 3.0, seeing both Intel and NVIDIA, Apple’s prime silicon partners, will move to it. The fact that Apple is slowly but surely killing FireWire is another indication.
On slashdot you can read that Intel are developing USB 3.0 support for Linux, why does sharp duplicate the effort?
http://hardware.slashdot.org/article.pl?sid=08/12/13/2032211
Dude,
Sharp is the last name of the intel developer doing this work. Have a cup of coffee and it will all make sense.
Mea culpa. I really need to get used to drink coffee before I posting anythin 😉
This is pretty good information, I’m am much more comfortable now with Intel developing the USB3 driver internally for Linux and seeing it reach close to the practical 300MB/s plus limit .Its neat to know the people actually doing the work can get to the shows and talk about it.
I’d hope the work can be reused for the other open source OSes too (Haiku etc) and I sure hope it gets backported to Win XP and 2000. Would be a real bummer if Microsoft asumes older Windows can live without USB3 but I’d sort of understand if there are technical reasons it can’t be done given the comments in the article.
I saw that Kangaru has a USB Flash drive with a Sata port on the other end getting us ready for a big speed bump from these devices.
Now I wonder when we will see USB3 Flash drives.
And if Intel marketing reads this, please pressure the vendors that put this into products to be honest on the packaging about speeds, no more 480*10 MB/s nonsense when the vendors know perfectly well what the limits really are, bith USB2/3 and the devices own limits.
that the isochronous transfers interface get simplified since it seems to be a constant source of pain to BSDs and BSD friends (like Haiku, I do not know about QNX). Do you know if the spec is easier to implement than USB 2.0?
While I greatly welcome USB 3.0, I don’t understand why Apple want to abandon Firewire. It is pretty good. I’m not an Apple customer and I don’t plan to, so it’s not a problem for me but I’d like to know. Before buying a new computer, I want to see USB 3.0 and SSD wide-spread for most silence and speed.
There are a handful of reasons, I think. Most new video cameras use solid state storage, which makes Firewire unnecessary. And while USB2 isn’t great, it *has* been good enough to keep Firewire largely confined to DV cameras and Macs.
Then there’s eSATA, which appears to be a better option than either Firewire or USB2 for connecting external drives (it’s also much more likely that a computer can handle eSATA, given the ubiquity of SATA).
And with USB3 on its way, I suspect Apple weighed the cost of upgrading Firewire to make it competitive with newer/upcoming standards VS. the potential revenues and decided it wasn’t worth their while.
But…They still need connectivity with a computer as much as before, no?
But why choose USB 2.0 over Firewire if it is much slower? (USB 2.0 goes up to 480Mbps; Firewire up to 3.2Gbps)
Nowadays, I agree, but Firewire was launched in 1995, and it should have been ubiquitous, for its technical advantages and speed, but this hasn’t happened.
Most of the newer cameras (that I’ve used, at least) record to removable memory sticks – and then the video is transferred by plugging the memory stick into a card reader of some sort. I don’t have extensive experience with newer cameras by any means, but that was the case with one of the newer Sony HD cams that we had the use of at work recently. IIRC, Macbook Pros have a built-in card reader that is compatible with the memory sticks used by Sony cameras.
That said, I certainly don’t think that Firewire will lose its usefulness for video work over night. E.g., the main camera we usually use at work is a Panasonic DVX100B (a plain ‘ol SD DV, tape-based camera) – and for stationary single-camera shoots, we normally hook it up to first-gen Macbook via firewire and record directly to a hard drive.
The main reason (the sole reason, IMO) is the ubiquity of USB 2.0. With a USB 2.0 device, I can go to any PC made in the past 4-6 years and reasonably assume that it will work.
There are also related factors – E.g., I’ve found that external hard drive enclosures which support Firewire are generally more expensive than those with USB2/eSATA support.
Totally agreed – if technical merit were the only deciding factor, Firewire would have won out (just from the amount of time that it took for USB to *somewhat* catch up in terms of transfer speeds).
I’ve always lamented that we never (to my knowledge, at least) saw real Firewire drives – in other words, hard drives with built-in Firewire controllers, rather than ATA drives connected via IDE-to-Firewire adapters. And there’s also the sad / amusing irony that Apple themselves helped popularize USB with the first-gen iMac.
Edited 2008-12-15 21:40 UTC
It’s all about price. It is much cheaper to use USB than FireWire. USB is mostly implemented in software. The connection itself is dumb. FireWire requires much more hardware to function. That’s also a big reason why FireWire is faster and better for things like streaming video. That doesn’t really matter anymore now that new camcorders are completely digital. There is no need to stream video when you can just copy files.
This is really one of those “My, my, hasn’t the world changed!” moments for me.
I remember not all that many years ago when Windows had been supporting USB for *years*. USB was becoming popular and regular folks were buying USB devices off the shelf, plugging them into their Windows98 machines, and using them. Meanwhile, the effort to add USB support to the Linux kernel… churned on… and on. There was a group of devs working on it. They had a mailing list. And I presume a CVS server, and all the other trappings of an OSS project. But no actual working code ever seemed to appear.
Meanwhile, I would have customers go out and innocently buy a USB mouse or keyboard, expecting to use it. And when it didn’t work, they’d call me. And I’d have to hmmm and hawww in explaining that Linux couldn’t do that. (What’s worse, I couldn’t even blame it on some proprietary Microsoft dirty trick!)
Finally, one evening, Linus decided to look into it. The next day, he posted on LKML, and in pretty much these words, “I looked at the code and I couldn’t understand it. And I figured that if I couldn’t understand it, then it must be bad. So I whipped up this patch last night.”
Well, the code he had written from scratch “last night” was the firm foundation that the previous USB guys had failed to achieve in the years they’d been working on the problem. And real USB support in Linux took off pretty quickly from there.
Shortly following Linus’ LKML post, the original team published a statement saying that they had discussed the matter, and had decided to cease development on their project. To this day I feel embarrassed for them, having all that happen so publicly.
But anyway, we sure as hell didn’t have any Intel sponsored USB efforts back then. The things we just take for granted these days…
Edited 2008-12-15 18:32 UTC
But why choose USB 2.0 over Firewire if it is much slower? (USB 2.0 goes up to 480Mbps; Firewire up to 3.2Gbps
This is why I bought the Seagate external drives with USB2/eSATA/Firewire. If you are using large files it would be a mistake to just go with USB2. You would regret it. My Firewire drive connected to a dedicated Firewire contoller card operates about twice as fast when moving large video files around. USB3 sounds nice but until it is available I would definitely stick to a USB2/Firewire combo. In a pinch I can connect my latop to the drives and copy via slower USB2.