I wish I could answer that, but without seeing in print, exactly
what the failure mechanism is, it is hard to offer advice. The
advice I give, is intended to be as conservative as possible,
since I don't really know the answer to questions like yours.
Intel has not admitted there is a problem.
Using a PCI USB card means that the Southbridge USB ports no
longer need to be used. That workaround still cannot offer
protection for a case of induced ESD discharge, where an
electrical conductor near the USB ports receives ESD, and
the energy is coupled into the USB wires. So, even if you
use only the PCI USB card, I cannot actually guarantee a
long life for the motherboard.
There have been a couple of reported cases, where the user
was not plugging in a USB device at the time:
"Symptoms: Machine was running in standby mode. Hit the
keyboard to get her going, nothing. Stated to smell electronics
burning. Tried to kill the machine with the button but no
go. Then, the machine just shuts down on it's own."
What that should tell you, is no amount of being "careful" will
help. Simply stay away as far as possible, from the motherboard
USB headers and the rear USB ports in the I/O area
Another poster has suggested there could be a manufacturing date
dependency on the problem. But none of the posters so far have
reported the numbers printed on the top of their Southbridge
(or the serial number on the motherboard box, as the first
two characters are a date code), so we have no way of verifying
that observation.
Latchup is a condition where a phantom PNPN junction forms
between VCC and GND. PNPN is the same structure as an SCR
(two of which are used in a TRIAC or lamp dimmer). The PNPN
junction, once triggered by its gate input, continues to
conduct large amounts of current until the voltage across
it drops to near zero. In the case of the phantom PNPN structure,
it appears when more than a certain amount of current is forced
to flow into an input. Mechanisms that can cause this critical
amount of current to flow, are a static electric discharge,
but also an inrush current can do it too. ICs typically start
to have input latchup problems at 500mA to 1A of current
jammed into or out of an input lead. That amount of current
is much larger than the normal current levels used for most
logic functions.
Paul