USB 3.0 slow with some USB 3.0 hard disks

[larrymoencurly]

I'm using Windows XP 32-bit Home with NEC/Renesas USB 3.0 PCI-E
cards equipped with the uPD702000A chip. I'm getting about 90
megabytes/sec sustained transfers with USB 3.0 enclosures based
on a Genesis chip but only 35 MB/s with Seagate GoFlex drives.
This is with both nVidia GeForce 6100 and Intel ICH5 motherboards.

It doesn't matter whether I use the newest NEC driver or the one
included with the Seagate GoFlex.

I've had no luck updating the NEC firmware because at the
start the updater software gives an error message:

W200FWUP.EXE - Entry Point Not Found

The procedure entry point_except_handler4_common could
not be loacated in the dynamic link library msvcrt.dll

But then the program asks if I accept the terms & conditions,
and then it seems to start updating, but in a few seconds
the progress bar stops midway, and the program just hangs
forever (I've waited 5-30 minutes). It can be terminated
from the task bar.

I've tried a new copy of msvcrt.dll, but it makes no difference.

 

[Arno]

I'm using Windows XP 32-bit Home with NEC/Renesas USB 3.0 PCI-E
cards equipped with the uPD702000A chip. I'm getting about 90
megabytes/sec sustained transfers with USB 3.0 enclosures based
on a Genesis chip but only 35 MB/s with Seagate GoFlex drives.
This is with both nVidia GeForce 6100 and Intel ICH5 motherboards.

It doesn't matter whether I use the newest NEC driver or the one
included with the Seagate GoFlex.

I've had no luck updating the NEC firmware because at the
start the updater software gives an error message:

W200FWUP.EXE - Entry Point Not Found

The procedure entry point_except_handler4_common could
not be loacated in the dynamic link library msvcrt.dll

But then the program asks if I accept the terms & conditions,
and then it seems to start updating, but in a few seconds
the progress bar stops midway, and the program just hangs
forever (I've waited 5-30 minutes). It can be terminated
from the task bar.

I've tried a new copy of msvcrt.dll, but it makes no difference.

I have no idea what is going on in detail, but 35MB/s is very
unusual for USB 3.0. It is however at the top of what
USB 2.0 can deliver. So I suspect that for some reason,
the USB 3.0 transfers produce to many errors and the interface
gets degraded to USB 2.0 or there is some incompatibility that
keeps the attachment at USB 2.0 from the start.

That could be a firmware issue. It could also be a subtle
hardware defect that reduces the margins normaly present
in the signals. Things you can try: Different cables
(preferrably shorter). Putting an USB 3.0 hub into the
connection. Trying with a different host-side interface.

As to the tropubles updating the firmware: Typically
you do not update the chip, but some external SPI
flash chip. That means the NEC firmware package may
fail becaue the card uses a Flash chip it is not
designed for. Can you get a firmware update from the
maker of the card?

Arno

 

[larrymoencurly]

I have no idea what is going on in detail, but 35MB/s is very
unusual for USB 3.0. It is however at the top of what
USB 2.0 can deliver. So I suspect that for some reason,
the USB 3.0 transfers produce to many errors and the interface
gets degraded to USB 2.0 or there is some incompatibility that
keeps the attachment at USB 2.0 from the start.

That could be a firmware issue. It could also be a subtle
hardware defect that reduces the margins normaly present
in the signals. Things you can try: Different cables
(preferrably shorter). Putting an USB 3.0 hub into the
connection. Trying with a different host-side interface.

As to the troubles updating the firmware: Typically
you do not update the chip, but some external SPI
flash chip. That means the NEC firmware package may
fail becaue the card uses a Flash chip it is not
designed for. Can you get a firmware update from the
maker of the card?

This is a generic USB 3.0 card, so tech support from its
manufacturer is unavailable, but I wouldn't be surprised if
the there's a problem with signal quality due to the cheap
design because I've seen that in generic sound cards, and
one of my Bytecc USB 3.0 enclosures has an interface board
with a 3-pin linear voltage regulator tweaked for 0.1V higher
voltage, through the addition of a couple of resistors that
were just slapped on the board on the board. Ironically, that
USB 3.0 enclosure works at over 100MB/s with that
troublesome USB 3.0 card.

I found a command line Windows flasher for the USB 3.0
card. It didn't work in a couple of Intel-based motherboards,
but in a Nvidia motherboard it complained that my NEC USB
3.0 chip was version 3 and therefore not compatible with the
version 4 firmware, yet it also said the chip had version 4
firmware.

I tried another USB 3.0 cable and even a whole other Seagate
GoFlex drive of the same vintage but got the same results.

 

[larrymoencurly]

On Wed, 27 Mar 2013 23:04:26 -0700 (PDT), (e-mail address removed)


I have a couple of the GoFlex bases and I have had them running
consistently at 114MB - 115MB/sec with a variety of drives
transferring large files. So it is not the GoFlex causing the problem.

It could be something else, but I don't understand why the GoFlex
drives are the only ones that have been slow with that controller
card. I've since tested with a WD drive (Prolific interface chip?),
and it ran at over 100MB/sec.

 

[larrymoencurly]

I took the Seagate USB cable and plugged and unplugged it several
times to clean it off, and I noticed it scratched up the plastic ribs
between the USB 3.0 contacts in the socket. IOW the USB 3.0
contacts weren't touching, and that could explain why Windows
warned that the GoFlex was operating only in USB 2.0 mode.
When I did the same thing with the other USB socket and one of
the cables that worked at USB 3.0 speeds, it didn't scratch those ribs.
So I trimmed the ribs to make each contact slightly wider, and the
Seagate GoFlex finally operated at USB 3.0 speed, about 125 MB/s.

 

[Arno]

I took the Seagate USB cable and plugged and unplugged it several
times to clean it off, and I noticed it scratched up the plastic ribs
between the USB 3.0 contacts in the socket. IOW the USB 3.0
contacts weren't touching, and that could explain why Windows
warned that the GoFlex was operating only in USB 2.0 mode.
When I did the same thing with the other USB socket and one of
the cables that worked at USB 3.0 speeds, it didn't scratch those ribs.
So I trimmed the ribs to make each contact slightly wider, and the
Seagate GoFlex finally operated at USB 3.0 speed, about 125 MB/s.

Thanks for the update! Seems my dark suspicions were on
the mark. But contact problems from wron contact
geometry is a new one for USB3.0, at least for me.

Arno

 

[larrymoencurly]

(e-mail address removed) wrote:

Thanks for the update! Seems my dark suspicions were on
the mark. But contact problems from wrong contact
geometry is a new one for USB3.0, at least for me.

I had a feeling all along that the problem couldn't have been the
NEC USB chip because I have a motherboard with this chip built-in,
and it aways gives at least 90 MB/s speed. OTOH its USB 3.0
connectors are name brand and have noticeably wider contacts.

This USB 3.0 card is a cheapened version of the ones selling for
twice as much in stores. It's the NEC reference design (even the
circuit board markings for the resistors and chips are identical),
only with some components missing, including the bulk capacitors,
it's at least 1/8" too short in the back and much of the soldering was
done by hand (badly -- reinforcement lugs on connectors not
soldered, many balled joints). At least it didn't have any solder
snot jumpers, where an arc of solder spans over copper traces.

 

[Arno]

I had a feeling all along that the problem couldn't have been the
NEC USB chip because I have a motherboard with this chip built-in,
and it aways gives at least 90 MB/s speed. OTOH its USB 3.0
connectors are name brand and have noticeably wider contacts.

This USB 3.0 card is a cheapened version of the ones selling for
twice as much in stores. It's the NEC reference design (even the
circuit board markings for the resistors and chips are identical),
only with some components missing, including the bulk capacitors,
it's at least 1/8" too short in the back and much of the soldering was
done by hand (badly -- reinforcement lugs on connectors not
soldered, many balled joints). At least it didn't have any solder
snot jumpers, where an arc of solder spans over copper traces.

Urgh. Cheapest possible manufacturing from somebody
out for a quick buck. With soldering that bad, my
suspicion would have been on some electrical problem
as well.

Arno