Slashdot Mirror
USB 3.0 Is Ten Times Faster; Get It In 2010
thefickler writes "Seagate and Symwave are jointly demonstrating the first consumer applications of USB 3.0 at CES, showing a Seagate FreeAgent drive running through a Symwave USB 3.0-compatible storage controller device. According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.' Which means, if you understand PR-write, it will be much faster."
Most of the replies so far show a glaring lack of knowledge of what USB3 really is. Honestly, it only bears a passing resemblance to its predecessors, and is a closer relative to PCIe. If you want more technical information, Denali has a good whitepaper (registration required):
http://www.denali.com/en/events/usb3_whitepaper/?EB20090105
USB is PIO and not DMA? You understand that PIO/DMA transfer modes only meaningful for Parallel ATA devices?
That's because the ATA interface was originally THE 16-bit system bus (AT bus) and the disk controller was onboard on the disks (thus the name IDE - Integrated Drive Electronics).
So, the CPU accessed the disk controller through the AT bus which was originally programmed I/O and then started using DMA.
Of course, the system bus quicky become different and faster (FSB + PCI) thus the ATA interface became a disk attachment interface instead of the system bus and an additional controller was put between the system and the disk.
USB is a completely different beast. The "bus" actually transfer packets (URBs) and all USB controllers use DMA to transfer URBs to the main memory... So, no PIO/DMA stuff is involved....
You understand that PIO/DMA transfer modes only meaningful for Parallel ATA devices?
That's the meaning in traditional sense. But you can also use this distinction in a wider sense:
How much of an advantage this is, depends on how complex the initial parameter setup is, how much of the work is done by hardware vs. CPU, transfer speed, how large transferred blocks are, how often transfer occur, etc. etc. Besides overall speed, a big advantage is that the CPU can do other things (like decode a video stream, respond to keyboard / mouse input) while a tranfer continues in the background. This allows a system to feel much more responsive.
You state that USB controllers use DMA, parents says not. I don't know which is true. Perhaps there is DMA support for USB controllers, but the packets are small enough and flowing at a high enough rate that it feels like the CPU is doing all the work?
One other problem Firewire had in becoming widespread was that it required a more beefy, dedicated chip. As far as I understand it, Firewire is implemented mostly in the chip, cutting out the CPU, and creates a more-or-less guaranteed bandwidth. This is why it was popular with camcorders - you could always be sure that you could transfer video in realtime. Same for high-end sound equipment.
USB, on the other hand, while it has its own controller chip, is moderated largely by the CPU and memory bus. If the computer is under heavy load, the USB throughput suffers. Peripheral devices are at the mercy of the host to control things. This is fine for things like mice and keyboards, which transfer relatively little data semi-asynchronously. You don't need such a robust high speed bus for such lightweight peripherals. But for hard drives and other devices, USB has some catching up to do.
Royalties aside, the Firewire chipset that could implement a high-speed bus robustly cost more than the more lightweight USB controller. In the particular case of the iPod, a portable device, having the Firewire controller eventually took more board space than Apple was willing to provide. While every computer had a mouse and keyboard, not every computer was made with the ability for connecting external drives. So there, too, the economics played in USB's favor.