PDA

View Full Version : 2uestion about USB 2.0


alternate
05-13-2007, 06:04 PM
Is the 480-Mbps bandwidth distributed without restriction through the ports according to the demands by each device connected to a port, or is there a limit of bandwidth per port? How many ports can one have of USB 2.0? Typical modern computers have only one USB 2.0 controller, and all ports are controlled by this controller?

I suspect the latter part is affirmative of the first question because a WD RE 160 GB Caviar SATA HDD can transfer a large file of a size between 5 GB and 7 GB at a faster speed through a SATA port than through a USB 2.0 port (via connectivity devices), and both speeds are below 480 Mbps = 60 MBps. I find it an annoying discrepancy to what I wish of the USB connectivity.

I have more related questions, but I don't want to make people who might want to answer some questions no longer want to answer some questions. (They regard SCSI, SATA, and SAS and the same questions for them.)

Thanky wanky! :D

saphalline
05-16-2007, 01:33 AM
The 480MBps bandwidth is not fully achieved with USB 2.0 due to several aspects of overhead and signal integrity and wait states. But of course we all marvel over USB 2.0 because USB 1.0 & 1.1 had these same problems but with lower bandwidths. ;)

However, you are correct in your practicality measurements of bandwidth using both USB 2.0 and eSATA/SATA. That 60MBps just isn't realized. On the plus side, USB 2.0 is much less susceptible to bandwidth reduction from multiple devices vs other external busses (*cough* firewire *cough*). So the reduced bandwidth that is available is nevertheless robust.

The USB spec limits the number of root ports to two per root hub. This is why you often see USB ports offered in pairs. 10 USB 2.0 ports means 5 root hubs. It's so easy to add them that chipsets seem to add them in pairs as we go along. I remember my old Athlon 1.2 system had 4 of them. Now we've got 10 in modern chipsets. As for the number of devices, you can technically add up to 127 in a daisy-chain fashion per root hub, but OS limitations prevent this from happening. Apparently, root address coherency was not part of the USB spec, so we see an availability of only 127 devices total, rather than "per root hub".

Still, it's difficult to fault USB for its limitations vs eSATA/SATA. After all, the two specs were initialized 5 years apart! A lot can change in 5 years. For instance, 10 years after the fact, we're seeing WUSB (wireless USB) on the horizon. I have a feeling that BlueTooth will be greatly reduced in the near future. :D

alternate
05-16-2007, 10:17 AM
That 60MBps just isn't realized.

The USB spec limits the number of root ports to two per root hub. This is why you often see USB ports offered in pairs. 10 USB 2.0 ports means 5 root hubs. It's so easy to add them that chipsets seem to add them in pairs as we go along. I remember my old Athlon 1.2 system had 4 of them. Now we've got 10 in modern chipsets. As for the number of devices, you can technically add up to 127 in a daisy-chain fashion per root hub, but OS limitations prevent this from happening. Apparently, root address coherency was not part of the USB spec, so we see an availability of only 127 devices total, rather than "per root hub".


But I get about 58 MB/s with SATA and about 16 MB/s with USB 2.0 using the same device. I would hope to get higher bandwidth with USB 2.0 but instead get only 16 MB/s. Are we sure each port on a root hub can take the full 480 Mb/s bandwidth of the hub (minus overhead limitations)?

So physically which are root hubs? I have two pairs on my back panel and an additional three pairs of internal pins for USB 2.0. Each pair corresponds to a root hub, and each pair can support 480 Mb/s? So that overall the five USB 2.0 hubs can support 5*480 = 2000+400 = 2400 Mb/s?

Thanks, saph!

If you have an external USB hub, this is just adding more ports and not increasing bandwidth?

saphalline
05-16-2007, 11:40 AM
The root hub in its physical form is a collection of transistors built into the chipset. It is the arbiter and manager of its USB bus. It also communicates with the rest of the system on the external bus, whatever that may be. In the case of modern chipsets, the USB root hubs are strapped into the proprietary bus on the Southbridge (or Bridge) chip.

While you should be getting more than 16 MBps, there are plenty of other factors to consider. Just because the max theoretical bandwidth from a USB device to a USB root hub is 480 Mbps doesn't mean that's the only bottleneck! The USB bridge chip built into the HDD enclosure may not be as efficient. Or perhaps the chipset subsystem can't keep up (unlikely considering the chipset of your mobo). It could even be something as simple as old or improper chipset drivers. You have to look at the whole issue, not just USB's bandwidth alone.

As of right now, I'm pointing my finger at the USB enclosure. Those are often hit or miss with performance.

alternate
05-17-2007, 02:41 AM
As of right now, I'm pointing my finger at the USB enclosure. Those are often hit or miss with performance.

Hmm... in that case I'll have to blame the SATA-to-USB-2.0-converter device as well as the USB 2.0 WD My Book. They both have a similar USB 2.0 transfer rate. I am lead to wonder of the existence of a USB HDD enclosure that has a transfer rate around 58 MB/s... and why my device and My Books do not support this transfer rate...

So, to confirm there are 5 USB 2.0 hubs in the SB chip for a total of 5*480 = 2400 Mb/s? And my Kensington USB expander hub is just more ports to the same USB controller for a total of 480 Mb/s?