Sydney said:
Hi everyboby
Machine with Asus A7PRO motherbord and intel proccessor bought in 2002 .
Since one week the machine stops after a variable time from 30 min to 5
min. The power telltale lamp goes from solid green to a blinking green.
Speedfan tool software says Via VT82C686 device is at a temperature of
144 °C.
Asus Probe tool shows 26 to 29 °C for the motherboard and 41 à 45 ° C
for the processor . The 12 volt supply has an erractic drop to 3.5 for
microseconds .
The 2 fans (box and processor ) are running.
I am out of ideas and puzzled by the discrepancy between the temperatures.
I don't know what to do next.
The A7PRO manual is dated November 2000, quite a while ago.
The manual claims the motherboard has an S462 socket, and those
normally accept AMD processors ( "Athlon/Duron" ).
The Northbridge on the motherboard, has a small green heatsink, sized
to cool a 3 or 4 watt load. The Southbridge power dissipation is
low enough, it has no heatsink on it.
Now, the first peculiar thing I notice about that board, is the
Vcore regulator is a VRM module that plugs into a high current
socket. You should check that it is secure. Sometimes, the contacts
on such power devices, either degrade (corrosion) and fail to
make good contact. Sometimes, wiggling re-seating the module is enough
to provide temporary relief. Such a module may have some form
of protection (overcurrent), but it's hard to tell anything
because the picture of it isn't very good.
I doubt your motherboard has hardware-based CPU thermal protection.
Some of the last of the S462 motherboards, had a small 8 pin DIP to
sense the CPU diode, and remove power in the event the CPU overheats.
Your overheat protection on the other hand, could use a thermistor in the
socket area, which lacks precision and has a higher latency (it is
read by the BIOS and so takes a fraction of a second to respond).
Certain kinds of failures, such as a CPU heatsink falling off, may result
in the processor crashing, before the software that checks the
temperature can turn off the computer. The hardware-based
solutions on more modern boards, are more reliable in that regard,
and turn off instantly without software intervention.
*******
So my first question is, you refer to a LED flashing. Are you talking
about "PLED" near the AGP slot ? The user manual shows that LED is
green in color.
Many of the Asus motherboards, have one green colored LED which is
connected to +5VSB. The purpose of the LED, is to monitor whether
standby power is still present. You're not supposed to add or remove
hardware components, when the green LED is operating.
Now, that LED should really only have two states. It's off, if the
switch is off on your ATX power supply. It comes on, solid green,
when the supply is switched on at the back. Even if the computer sleeps,
and the fan is not turning, that LED can still be lit solid green.
If that LED is seen to flash, that is indicative of a bad ATX power
supply. It means the ATX power supply has a weak +5VSB rail,
or it could mean some other rail is shorting out, and the
supply is protecting itself against overcurrent. In any case,
you'd swap out the power supply and try another one.
If you're referring to a "Power" LED on the front of the
computer case, there are more possibilities there, and
I really can't tell you for certain, that a flashing LED
tells you anything. It would normally be used as a "sleep"
indicator, but there might be other reasons for it to blink
that I'm not aware of.
*******
With regard to temperature issues:
1) Make sure any heatsinks present in the computer, are secured
in place. Sometimes, a "push-pin" or a clamp, breaks loose,
and then either the CPU or the Northbridge, aren't being cooled.
2) The thermal interface material, between the CPU and heatsink,
should be checked, any time there is an abnormally high CPU
temperature. If you kept track of what a "normal" temperature
was, when the computer was new, then you'd be better able to
judge whether the value read now, is abnormal.
3) Computers have a couple low speed buses, used for control
purposes. The first of those, might be SMBUS. It is a
serial bus, operating at perhaps 10KHz. (The ones I used in
the lab, ran at 400KHz, so there are several choices for
operating speed.) In any case, the SMBUS can be practically
as slow as an old dialup modem.
That is a problem, when you use two different temperature
measurement programs at the same time. If the first program
starts to access the SMBUS, and the second program
attempts to access SMBUS a fraction of a second later, the
value read by the first program gets corrupted. This causes
the value displayed on the screen, to be "strange". You'll
see normal readings for a while, and then one "strange" value.
The solution in that case, is to disable one of the temperature
reading programs, such that you're only using one at a time.
More modern computers, have the SuperI/O hardware monitor interface,
sitting on the LPC bus. That is a nibble wide, higher speed
bus, which is not prone to corruption like the other bus is.
A modern motherboard has fewer excuses for corrupted temperature
readings, because it uses the LPC bus. Older boards use SMBUS.
The SMBUS could have been "bulletproof", if all software
authors could agree on a semaphore scheme to prevent
corruption. That didn't happen, so we're left with SMBUS
corruption on readout, as a result. As long as only one
program is used at a time, you can get correct readings
from SMBUS.
*******
So I would start, by doing various mechanical checks. Turn off
the power. Then, wiggle all the connectors and reseat them,
as on an older computer, there could be corrosion. Sometimes,
the problem is "tin whiskers" can grow between contacts, and
a visual inspection is necessary to catch those. We had that
problem at work, with some pure tin setups we were using. The
previous gold plated components we made, didn't do that.
Next, if the machine isn't any better, and the green LED on
the motherboard is still blinking, try another ATX power
supply. That might be all it needs.
Power supplies back then, may have had -5V as one of the
output rails. The motherboard may or may not check for
that voltage. In some cases, the motherboard won't run,
unless you can find an old supply to use with it.
This is an example of a supply you can buy, which still has
a -5V signal on pin 18. This kind of supply, is what you
use, if the motherboard refuses to run with one of the
newer standard of ATX supplies. This unit also has a
pretty decent +3.3V/+5V rating, suited to older systems.
Your board doesn't use an ATX12V 2x2 connector, so could be
running the VRM module from the 5V rail (perhaps 12 amps or
more).
http://images10.newegg.com/NeweggImage/productimage/17-103-915-06.JPG
The only problem with a supply like that, might be the
cable length isn't long enough. But if that is the
only kind of supply you can get, with -5V on it, then
fitting a $7 extension cable is a fix for it. Only use
an extension like this, if you have no other choice.
These should not be used on a whim, as they degrade the
accuracy of the 3.3V rail (the remote sense function).
http://www.newegg.com/Product/Product.aspx?Item=N82E16812198023
HTH,
Paul
