Another USB Quandary

[JohnO]

I'm playing with two things, an Intel DG965SS mobo and Microsoft's
UVCView.exe USB tool. UVCView shows that there are seven root hubs, two of
which contain the magic word: Enhanced, which means they are Hi-Speed USB.
The board has ten physical ports.

When I inventory the physical ports....plug in a device and see which root
hub/port is connected to each physical port, none of the physical ports
match up to the two Enhanced Root hubs shown in UVCView.

At first glance that would mean I don't have any hi-speed USB ports, right?
But, the Intel product specs claim hi-speed on all ten physical ports.

So, is it possible that UVCView isn't showing me the entire story, and that
the mere presence of an Enhanced root hub (anywhere) means I've got hi-speed
everywhere?

Do some motherboards have mixed USB...some full-speed and some hi-speed? If
so, how do I determine which is which?

-John O

 

[Paul]

I'm playing with two things, an Intel DG965SS mobo and Microsoft's
UVCView.exe USB tool. UVCView shows that there are seven root hubs, two of
which contain the magic word: Enhanced, which means they are Hi-Speed USB.
The board has ten physical ports.

When I inventory the physical ports....plug in a device and see which root
hub/port is connected to each physical port, none of the physical ports
match up to the two Enhanced Root hubs shown in UVCView.

At first glance that would mean I don't have any hi-speed USB ports, right?
But, the Intel product specs claim hi-speed on all ten physical ports.

So, is it possible that UVCView isn't showing me the entire story, and that
the mere presence of an Enhanced root hub (anywhere) means I've got hi-speed
everywhere?

Do some motherboards have mixed USB...some full-speed and some hi-speed? If
so, how do I determine which is which?

-John O

Using ICH8 datasheet from Intel - Table 13 mentions physical port mapping to
controller. Table title is "USB Interface Signals".

http://www.intel.com/design/chipsets/datashts/313056.htm

Six USB ports are on one controller, four USB ports on the other.
There should be two enhanced entries in UVCView. One should have
room for six entries, the other with four entries.

You should see two Enhanced entries in Device Manager. That
would tell you there is a USB2 driver in place.

Then, when you use a test device, which is USB2 capable, and
test each USB jack, it should show in the area in
UVCView, where there are the six entry and four entry sections.
(At least, if UVCView can handle all those entries.)

The era of "mixed" USB is long past. There was a time, when
USB2 just started appearing, where some chipsets only had USB1.1,
and a separate NEC chip or similar may have been bolted to the
motherboard, to give some USB2 ports. The stacks on the back
of the computer, may have mixed the two types together (but the
motherboard manual identified which was which).

With modern motherboards, that is no longer necessary.

You could take a look in the BIOS screens, to see if
some USB setting there is responsible.

Also, does your USB2 test device, get detected as USB2
on another computer ? Is the problem isolated to your
DG965SS ?

Paul

 

[JohnO]

Using ICH8 datasheet from Intel - Table 13 mentions physical port mapping
to
controller. Table title is "USB Interface Signals".

http://www.intel.com/design/chipsets/datashts/313056.htm

Six USB ports are on one controller, four USB ports on the other.
There should be two enhanced entries in UVCView. One should have
room for six entries, the other with four entries.

Yep, that's it. and ten other ports, too.

You should see two Enhanced entries in Device Manager. That
would tell you there is a USB2 driver in place.
Yes.


Then, when you use a test device, which is USB2 capable, and
test each USB jack, it should show in the area in
UVCView, where there are the six entry and four entry sections.
(At least, if UVCView can handle all those entries.)

Aha...if I connect a hi-speed device it *does* show up on one of the
Enhanced hub/ports. But if I use a slow device, it shows up on a different
hub altogether. That explains it: There are 20 ports listed, 10 for slow and
10 for hi-speed...I get a port based on the device.

I had been using a slow flash drive and a cheap hub, assuming that the
physical ports were tied to logical ports, but now I see that's not the
case.

But, just switching to a 2.0/hi-speed hub makes all the difference. This
gives me a lot to think about and a lot to work with.

The era of "mixed" USB is long past. There was a time, when
USB2 just started appearing, where some chipsets only had USB1.1,
and a separate NEC chip or similar may have been bolted to the
motherboard, to give some USB2 ports. The stacks on the back
of the computer, may have mixed the two types together (but the
motherboard manual identified which was which).

With modern motherboards, that is no longer necessary.

OK, that's good.

Thanks Paul, I appreciate the help on this.

-John O

 

[Paul]

Yep, that's it. and ten other ports, too.


Aha...if I connect a hi-speed device it *does* show up on one of the
Enhanced hub/ports. But if I use a slow device, it shows up on a different
hub altogether. That explains it: There are 20 ports listed, 10 for slow and
10 for hi-speed...I get a port based on the device.

I had been using a slow flash drive and a cheap hub, assuming that the
physical ports were tied to logical ports, but now I see that's not the
case.

But, just switching to a 2.0/hi-speed hub makes all the difference. This
gives me a lot to think about and a lot to work with.



OK, that's good.

Thanks Paul, I appreciate the help on this.

-John O

PDF page 218 of the Intel ICH8 document, has a picture of the USB details.
(See "ICH8-USB Port Connections".) There are a total of seven logic blocks.
Two USB2 and five USB1.1 . Each USB1.1 controller, handles a stack of two
ports. The USB2 ones are six ports and four ports respectively.

When you plug in a USB device to one of the ten physical ports,
during the setup, there is a binding between the physical port,
and one of the seven logic blocks. (That is what those little "switches"
in the diagram are meant to represent.) And that is why, a newly plugged
device, will either show up under a "group of two" USB1.1 block,
or a larger grouping used on the USB2 blocks. And that is also why
UVCView shows twenty entries, when the chip has ten ports.

The purpose of putting two USB2 blocks, in this chip, is to increase
the maximum simultaneous bandwidth on USB. A total of 120MB/sec max,
60MB/sec per USB2 controller, is shared over the ten ports. If you
had a couple high resolution web cams, or some other demanding
application, where normally you'd run out of bandwidth, and the
hardware devices would be unhappy (jerky etc), if you plug one
into each USB2 controller logic block, then things could improve.

My ICH5 has eight ports, on one USB2 block. So 60MB/sec is shared
8 ways on my older motherboard. Protocol overhead limits available
bandwidth, and 57MB/sec is a theoretical max, while with a USB2
external hard drive, I might see around 33-35MB/sec on a good day.

Paul

 

[JohnO]

PDF page 218 of the Intel ICH8 document, has a picture of the USB details.
(See "ICH8-USB Port Connections".) There are a total of seven logic
blocks.
Two USB2 and five USB1.1 . Each USB1.1 controller, handles a stack of two
ports. The USB2 ones are six ports and four ports respectively.

When you plug in a USB device to one of the ten physical ports,
during the setup, there is a binding between the physical port,
and one of the seven logic blocks. (That is what those little "switches"
in the diagram are meant to represent.) And that is why, a newly plugged
device, will either show up under a "group of two" USB1.1 block,
or a larger grouping used on the USB2 blocks. And that is also why
UVCView shows twenty entries, when the chip has ten ports.

Port routing, that makes sense.

The purpose of putting two USB2 blocks, in this chip, is to increase
the maximum simultaneous bandwidth on USB. A total of 120MB/sec max,
60MB/sec per USB2 controller, is shared over the ten ports. If you
had a couple high resolution web cams, or some other demanding
application, where normally you'd run out of bandwidth, and the
hardware devices would be unhappy (jerky etc), if you plug one
into each USB2 controller logic block, then things could improve.

Oddly enough, that's going to be useful. The machine I'm writing about has a
webcam and a WNIC, both on USB, along with a USB motor controller (slow), a
game controller pad, and who knows what else the customer might add.
Splitting the hi-speed devices to different blocks makes sense...if the
motherboard in that machine has them.

Off to look for the 945GM Specs....

My ICH5 has eight ports, on one USB2 block. So 60MB/sec is shared
8 ways on my older motherboard. Protocol overhead limits available
bandwidth, and 57MB/sec is a theoretical max, while with a USB2
external hard drive, I might see around 33-35MB/sec on a good day.

Paul

How do you measure that?

-John O

 

[Paul]

How do you measure that?

-John O

It's been so long since I had the USB2 enclosures, I've forgotten
I don't have it any more, so cannot rerun any tests on it.

Looking through a few results here, the best they can get, is 31.9MB/sec
on reads. So maybe my memory isn't so good

http://www.tomshardware.com/reviews/icydocks-mb559-happily-marries-esata-usb2-0,1347-5.html
http://www.tomshardware.com/reviews/western-digital-flashy-lighted-hard-drive,1148-5.html

For benchmarks, there is HDTach, HDTune, and you might be able to
find a copy of an Atto benchmarking tool as well. This is an
example of Atto testing a USB2 hard drive.

http://www.ocmodshop.com/images/reviews/storage/vantec_nexstar3/bench_01.jpg

Paul