Try Overclock.net because they seem to discuss this:
http://www.overclock.net/a/database-of-motherboard-vrm-failure-incidents
Apparently MSI is pretty bad, something mentioned at other websites, too.
I thought all voltage regulators for the CPU were supposed to safely shut
down in case of overload.
This was the generation, where a number of AMD motherboard makers
put in "weak" VCore regulators. They covered their tracks, by
building a CPUSupport page which didn't list any 125W or 140W
processors. (My suspicion, is AMD suggested this too them,
to make the motherboards undercut the Intel competition.)
Anandtech wandered into this by accident, blowing
boards by running too hungry a processor. This probably
happened, by them rushing to review product, without
looking carefully at the CPUSupport charts.
"Post Cards from the Edge - AMD 780G, NVIDIA 790i, Gigabyte 680i"
(Failing AM2+ boards due to limited VRM/VRD designs)
http://www.anandtech.com/show/2496/2
I thought it was interesting, that some boards actually popped
during POST. Implying even single threaded execution was
enough to trash the VCore regulator. You would have to
compare the TDP of the processor they used, to the CPUSupport
chart, to see the ratio of TDP (overload) to TDP (rated).
If a 125W processor pops a board that runs 95W processors
at max, then it implies no design margin at all.
And this is the marketing push in the generation following.
After the rash of blown stuff, the VCore got more attention.
Robust VCore became a "feature" :-( Sort of similar to
replacing electrolytics with polymer caps, to avoid the
"stink" of the capacitor plague era.
http://event.asus.com/mb/140w/
*******
The regulator designs do seem to have protection against
overcurrent, but perhaps the overcurrent trip level is set
way too high (to prevent accidental shutdowns for legit loads).
Sometimes, the current through the MOSFET, develops a voltage
the main regulator chip can monitor. There are various
tricks to reduce component count and still have protection
for the regulator. There should be shutdown to the phases
on the thing, if a problem is detected. I don't think it's
thermal as such - it is too hard to monitor temperature.
But current flow, and maximum safe current flow could be
monitored. If the protection resistor is set improperly
(on purpose), the regulator might not respond unless there
was a plane-to-plane short around the socket. A slight
overload of a 125W on a 95W regulator, might be enough to
fry it.
*******
To relate another story, I was shocked to open up a computer
I gave someone years ago (for a check). While the thing
was running, I put a finger on coils and MOSFETs, and the
thing was cooking. MOSFETs might have been around 50C or so.
It's not good for the MOSFETs to be hot, because they can have
a form of thermal runaway. It might not be considered classical
bipolar runaway, but the MOSFET has a safe operating area on
the datasheet, and getting the MOSFET to heat up, is a sure
way to prepare it for trouble. And at the time, the computer
in question was just idling. I'd hate to see the temperatures
when Prime95 was running. The CPU cooler in that case, stops
rotating when the CPU is below 55C, so you can't even rely
on side spill air from the CPU cooler, to help the VCore
regulator components. Since the machine is similar in design
to a laptop (it's an all-in-one), there really isn't much
I can do to modify it or replace the board in there. The
board design is custom form factor (fits behind the LCD panel).
On one of my older computers, I purposely reduced CPU clock,
in an effort to keep MOSFET temperature in the "didn't burn
my fingers" zone.
But if a motherboard is going to pop while the BIOS is posting,
you might not even get a chance to change the CPU operating
conditions.
Paul
