View Single Post
Old March 31st 14, 02:46 AM posted to alt.comp.periphs.mainboard.gigabyte
external usenet poster
Posts: 13,364
Default GA-78LMT-USB3, no USB3

John wrote:
A friend has a system based on a GA-78LMT-USB3 (mobo rev 5.0) and
can't get any of his USB3 ports to work. The USB2s are fine, but the
USB3s, both on the front panel and the back panel, are not.

The ports have power -- when an MP3 player is plugged in, it charges.
But there is no data connectivity. The system (Windows 7, 64-bit)
can't see any device that he plugs in.

I checked the BIOS settings and USB3 is enabled. I tried to reinstall
the USB3 driver (D/L'd from the Gigabyte site) and the installation
failed, saying it could not find the hardware.

I'm thinking he has a hardware problem and he's going to have to
return the box and probably have the mobo replaced. But before he
does that, is there anything else I can try?

Thanks for any suggestions

Reply-to address is real


Pop in a Linux LiveCD, and see if the USB3 chip is detected.
Linux has the "dmesg" scrolling log file, which is freshly created
on boot. The buffer is only so big, so if enough events happen,
some of the more interesting messages could be lost. Typing
"demsg" in a Terminal window, dumps the log.

Linux also has things like "hwinfo" to display detected hardware.

Windows, there are a number of utilities you could use. This is an
old utility, but I happen to like it. Even if at this date, a
number of things will show up as "unknown". You still have the
VEN/DEV or VID/PID numbers to use from the screen of this tool.

"EVEREST Free Edition 2.20"

For things not automatically identified, there is a PCI database,
filed by VEN and DEV numbers.

There is also a USB database, for USB peripherals connected
to USB ports. These are filed by VID and PID numbers.


This is a simplified view of a peripheral chip.

Bus ---- Logic --------- I/O Pad ---
block Signal Driver (D+, D-)

If static electricity hits the signal wire with the arrow on it,
that can blow out the hardware signal driver. The result is,
Windows continue to detect the logic block and tell you
everything is fine. But no amount of plugging and unplugging
peripherals elicits a response. It's like the chip is "deaf".

There is another reason for that failure to occur. Sometimes,
popular brand computer cases, the front panel cables are *miswired*.
This can cause the port to fail.

If you suspected that in this particular situation, you might
refuse to connect the front panel wiring when the new motherboard
comes back. I use a multimeter to "buzz" the wiring and ensure
the right things are connected on either end of the cable.
If you've experienced a peripheral failure, on a peripheral with
front panel wiring, don't hook up the wiring again! That's the
simplest prevention. I have two Antec cases with wiring errors,
but fortunately, they never hooked VCC to the wrong pin. If
I'd gone ahead and used the front cables, the port would have
refused to work. My back ports work, so I know my chip is not

Another form of failure, is when the bond wire for VCC or GND
on the driver pad, burns out. This can be caused by latchup,
which is caused by static discharge. An example of the bond
wire blowing, is shown here. The top of the IC in question
(Southbridge) has a burn mark.

(I'm surprised this site still works!)

The logic block itself, hardly ever fails.

The chip select to the logic block, can be
blocked by the BIOS. So disabling the chip in the BIOS,
can make the logic block "disappear" and refuse
to be detected. I'm not really sure of all the
ways that's implemented now, so I can't help with
theories at that level. PCI Express doesn't work
quite the same way as PCI, and I don't know the
ways that are guaranteed, or most of the time work,
to disable chip response. PCI Express is a hub bus, and
disabling an "arm" of the star wiring, could stop a chip
from working. Even disabling a clock signal for the chip
could be used to kill it, but that is not normally recommended
as a hardware control method (I've had hardware burn by doing
that :-) It seemed funny at the time).

Use the detection methods first (like Everest),
before sending the motherboard back.

Another thing you could try, is to remove all power
from the computer. Do the "clear CMOS" thing, which
causes all the BIOS settings to be reset. You could
even remove the CMOS battery, to help clear the
CMOS settings.

Sometimes, unexplained behavior is actually caused by a
CMOS variable you cannot observe. Removing PC power can
help with certain kinds of hardware failures, while removing
all power (unplug) and pop out the CMOS battery, helps with
errant settings. You should record (either with a camera or
write them down), all the BIOS screen settings before you do
this, so you can put them back later after clearing the CMOS.

Clearing the CMOS should only be done, with the computer
unplugged, as many motherboards have a purposeful design
flaw, where a diode on the motherboard gets burned if you
Clear CMOS with the power still on. Unplugging is a nice
safe policy for this.

If I knew of a web forum for the GA-78LMT-USB3, I'd go research it.
You could try Newegg reviews or Amazon reviews, to see if
this problem is a common one. I looked at a few reviews, and
blown USB3 isn't standing out there.

The motherboard manual says the USB3 chip is VIA VL805 quad port USB3.
The quad port means two USB3 on the back panel, as well as a dual
connector for front panel wiring (19 pin F_USB30 connector). The missing
pin is for keying, so the connector goes in the right way. Since the
pinout isn't particularly symmetric, maybe putting the connector in
the wrong way will blow it ? You would need to take the single sheet
of paper for the computer case documentation, to see if everything
matches. And check whether the computer case cable has a keying tab
to mate with the place for a tab on the motherboard connector. If no
keying mechanism is evident (computer case cable isn't keyed), it's up
to the user to put it in the right way :-( Now you know why the replacement
motherboard, you want to be careful with that F_USB30 thing (if it isn't
keyed to make pin 1 go to pin 1).