Lucas said:
OK, tnx for info.
I have another problem with USB.
USB 1.1 is disabled in BIOS,( because I want USB 2.0 for external HD).
I have alrready tried Sweex PCI card fast speed 4 port USB 2.0,
and Mentor USB 2.0, 2 PORT usb Host adaptor PCI.
In the system hardware config I see:
'Standard enhanced PCI to USB-host controller' -Generic USB-Hub
'USB 2.0 Root Hub'
but I do not have the speed 2.0 for external HD (WD 500GB)
(with 'HDTune 2.53' test: transfer rate of 10-13 MB/sec average)
The sam ext HD plugged in a portable with USB 2.0 on board , has a
speed of 25-30 MB/sec.
what is wrong ?
I think many people have problems with PCI to USB 2.0 ?
Lucas
But you do have the speed.
If the interface was running at USB 1.1 rates, then the disk could
transfer no faster than 1MB/sec. You have achieved 10-13MB/sec,
which means it is running at USB 2.0 speeds.
The chipset is rather unique in computing, being an integrated
Northbridge/Southbridge in a single chip. The MuTIOL bandwidth
of 1.2GB/sec, eliminates the Northbridge/Southbridge bus as a
source of limitation in the architecture. (Some old chipsets,
used to connect the Northbridge and Southbridge, using the same
PCI bus as is used for the PCI slots. A bad design.)
http://www.sis.com/products/sis745.htm
The first thing I'd do, is test the enclosure on another computer,
to make sure it is actually a 30MB/sec design. Some USB enclosures
don't actually do that well. They never get 30MB/sec. You could be
putting yourself to a lot of trouble for nothing.
The problem could have something to do with the PCI bus. Try
moving the USB card to a different slot. For example, it could
be that PCI slot #1 shares an interrupt signal with the AGP
card, so each time the USB chip raises an interrupt, the AGP
card driver runs (to check whether the interrupt is for it or
not).
Another performance issue, could be the "PCI latency timer" setting,
which sets the maximum burst size on the PCI bus. Setting a
high value for that parameter in the BIOS, means an individual
card can have increased performance (due to less PCI overhead),
but at the expense of fairness (other cards not getting access
to the bus when needed - sound card gets upset etc). Older
boards might use 32, and you might occasionally see this set
to 64.
Another way to test your PCI bus, would be to use a PCI IDE
card, and connect the drive to that, install the driver,
then do the HDTach or HDTune benchmark run. For most disks,
you should see the classical curved benchmark result, with
higher transfer at the beginning of the disk (60MB/sec)
and lower transfer near theend (40MB/sec). If you could
get results like that, then you know the PCI IDE card is
not suffering from PCI bus starvation. (I.e. My example
result proves the bus can carry at least 60MB/sec, and
the burst transfer benchmark is an even better indicator
of performance.)
My guess right now is, that this is an enclosure problem. It
is going to take you a bit of work, to prove that the problem
is motherboard related.
Paul
