USB 3.0 is built upon, and is backwards-compatible with, the USB 2.0 "High Speed" spec. It would be generous to even call the specifications "early" at this stage, but the group still had lots of information about how USB 3.0 will work and what features it will provide. The spec should be finalized sometime in the middle of 2008, with initial devices available in '09, and broad deployment by 2010.
The main two goals of SuperSpeed USB are to provide a 10X boost in transfer rate (from 480-Mbits/s in USB 2.0 to 4.8 Gbits/s in USB 3.0), while dramatically lowering power consumption. One example of their speed goals is to transfer a 27GB HD movie to a portable device in 70 seconds. The same thing would take 15 minutes or more with HighSpeed USB (2.0). The SuperSpeed devices will use the same connectors and the same programming and device models as existing devices.
Consider that it takes maybe one or two tenths of a second to transfer a typical 4 Mbyte song to a portable music player with today's USB 2.0. That's "fast enough" for some users, but just try to fill up that 80-Gbyte iPod and you might as well walk away and cook dinner while you wait: 3,000 songs would take perhaps 400 to 600 seconds – up to 10 minutes.
Nobody wants to sit at their computer for that long, but if you could get it to under a minute, you're on to something. That's the performance aim of USB 3.0. There will be dedicated "in" and "out" lanes for data instead of the multiplexed solution common to USB today. That means devices can send and receive data simultaneously without bogging down.
The host schedules all transactions, but there is no continuous polling of devices. A new device is asked only once if it has new data to send. If not, it is never asked again; the device informs the host when it has new data. This alone should dramatically reduce power consumption of USB 3.0 devices, especially when "idle." Consider a USB mouse attached to a notebook. It's not constantly in motion, but the USB port polls it for info over and over anyway, draining battery life. That situation should be much improved with SuperSpeed USB. Other features, like per-link power management, will further reduce power consumption.
Device virtualization will be an important part of the new USB spec as well. The Promotions Group wants to make sure that virtual machines can access USB 3.0 devices without software intervention. A final concern is over Mass Storage Device drivers. Currently, MSD drivers for USB 2.0 devices have enough overhead to limit performance to around 32 Mbytes/sec. The overhead in this driver model would just kill the great potential for SuperSpeed USB hard drives and flash storage performance, so a new, more efficient driver model is needed. This is out of the purview of the USB spec itself, but the teams involved want to work hand in hand with the driver specifications groups to make sure USB 3.0 Mass Storage Devices can live up to their potential.
So SuperSpeed USB (aka USB 3.0) still has a long way to go, but the development pace seems pretty rapid. The spec will probably be done, or nearly so, by the time Intel Development Forum rolls around next year, with initial device development underway. In two years' time, we might even have the first USB 3.0 devices on the market.
0 comments:
Post a Comment