K8NSpro PWR_Fan connector

jt3 wrote:

Yes, I've been using the GigaByte monitoring utility, about which I have
some other concerns, viz., what accuracy can one expect from the voltage
readings? The numbers suggest a precision of 0.001 V, which I suspect. The
reason I'm curious is that it reported the core voltage high--~1.58 V while
the 12 V is about 12.23-12.3 V. The 3.3 comes in at about 3.24-3.26. I
adjusted the core down to 1.51. One of the reasons I suspect the precision
of the reporting is that the 12 stays mostly at 12.23 and hops briefly to
12.30 every so often, probably no more than 5% of the time. The 3.3 jumps
approximately evenly between the two values. The core is steady at 1.51
under load (Prime95). These are all taken during the Prime95 run which has
now been running about 28 hours. I chose the option which gave me a mix of
tests, without paying much attention to the nature of the mix.

If you have a look at the motherboard, you may be able to read the part
number off the chip that does the Super I/O and hardware monitor function.

A typical chip has an 8 bit ADC, with a full scale reading of 4.096V.
Doing the math, 4.096 divided by 256 values (8 bit ADC) gives a base
resolution of 0.016V . But, when a hardware program reads the values,
sometimes averaging is used, which will change the apparent step size
of the number stream. In any case, the resolution is not 0.001.
And in the case of voltages higher than 4.096V, the input to the
monitor is scaled with a resistor divider, and that scaling again
changes the step size the user sees (i.e. bigger steps). Basically
256 values spread over 12V, for the 12V rail.

Any ADC needs a reference voltage of some sort. Some measurements are
ratiometric, which means the same reference voltage feeds the sensor,
as feeds the ADC. A ratiometric measurement, removes the 1 or 2%
tolerance of whatever they are using for a reference voltage. Power
supply readings are not ratiometric, which means they could be off
by 1-2%, as well as the resolution limit of 16 millivolts, as well
as the tolerance of any scaling resistors. Temperature readings are
even more complicated, as some chips now use an internal lookup
table, to map from voltage to temperature, and that mapping is not a
linear function. It would have been better if they stuck with
reading the temp as a voltage, as then you could look at the
raw values and do your own analysis. Not that these details
are important or anything...

Treat the measurements as pretty crappy, unless the values make
you happy :-) And remember, that for BIOS readouts at least, some
BIOS were known to apply an offset to the measured value, meaning
even more voodoo at work. (That was typically done, when temp was
measured in the CPU socket area, with a thermistor.)


Your suggestion about the thermal paste is one I take seriously. In the
initial installation, I used the 'processor in a box' package as it was sold
to me.

There wasn't much chance for error in that, but even there I had trouble,
for the first thing I noticed as I started running the machine (prior to
installing XP--this was over 2 years ago, of course) was dangerously high
reported temperatures (BIOS), which induced me to purchase the Zalman
cooler, as the dealer could offer no alternative.

This CPU appeared to be running at 70 deg C just idling in the BIOS screen.
Try as I might, nothing seemed to lower the temperature. I purchased Arctic
Silver, applied scrupulosly according to their directions, using a razor
blade to spread it (rather than putting a small quantity in the center of
the heat spreader and forcing it uniformly out upon application of the heat
sink). I never liked that method of application, even though they (Arctic
Silver) asserted it as the proper and only way.

However, I shortly afterwards found that the temperature reading problem
was in the microcode of the CPU, or at least the way the BIOS (Award) was
reading the CPU sensor, and a later BIOS update changed the whole picture,
and totally removed any reading excesses. The initial installation of XP
was fraught with hang-ups, blue screen type, and every sort of thing that
makes more sense in light of the motherboard's subsequent history. But
since that time, the CPU has been reinstalled twice, using the same method
as they recommended, so I don't have great confidence in it. The only thing
arguing against it, that I can see, is that, so far it has stayed
rock-steady at 34C all through Prime95, and the 4 times it quit with the 380
W psu, only one was any farther than, say 10 minutes from boot, and this is
all at idle. It has never quit while trying to do anything with it. The
stoppage with the 500W psu, is another matter altogether, since it wouldn't
even turn on afterwards. I only tried turning it on, say, over a 1/2 hour
period, before pulling it apart and then putting the 380 back in. But it
wouldn't even light the LED, which is why I was so sure it was the psu. But
now, that same psu has been running longer than all its time before, and
under Prime95, even if it isn't the heaviest possible load, without so much
as a burp. Shortly, it will have run longer than it did with the 380 since
the XP reinstallation. Since I am well into doing nearly everything over
that I have done on the machine for about the fourth time, pulling the cpu
cooler out and reinstalling it shouldn't be so bad, but the Zalman is a mess
in its installation. Since it has been a while (September) since the board
came back the irritation should be dulled, but I find myself ever more
inclined to throw the machine into the middle of rush-hour traffic. It
would be nice if I could just get the thing to stop again as it seemed to
want to do a couple of days ago--half of me wants to believe it has 'gotten
well' while the less credulous half of me knows that doesn't happen.

The fact that you've applied paste multiple times, tells me you've had
enough practice. I asked the question, because there are a few people out
there, who feel a crooked heatsink, or a heatsink installed dry, makes
sense. So I have to ask questions, to see if they are aware of what
they're doing.

The "dot in the center" approach to paste, should be fine if the
metal has been "primed" with paste at least once. The idea is to
force all the air out, so if the metal was primed, there should be
less room for air. I spread mine with an edge, to reduce the need to
"squash" it. And I use a bit more than the minimum quantity, as I
like to see a little bit wet the edge of the joint. If the motherboard
is outside the case, I can look between the heatsink and CPU, to see
if any paste is visible in the joint between them.

I would hope the temps rise a bit when you Prime. If the temp doesn't
budge, maybe the sensors are mislabelled.

When you see a power supply sensitivity, it can be that the power supply
is too small for the system. You can calculate/estimate the power consumption
and work that out. The Antec supplies, due to their flexible output arrangement
(load any single rail as heavy as you want), should reduce the probability
of flaky behavior. At least for their Truepower series. If I saw some
flakiness, I'd reach from my clamp-on DC ammeter, and measure how much
power is actually being drawn. But gadgets like that are too expensive
for home DIY use. My clamp-on meter, is the most expensive meter I own,
and the rest are cheap $30 to $100 multimeters.

When a power supply won't turn on afterwards, that is called "latch-off".
A supply which thinks it has suffered an overload, can insist that the
user flip the switch on the back, to reset the power supply. The Vcore
on some motherboards also can feature latch-off, and they won't clear
their latched state, even if you press reset. Flipping the switch on
the PSU may be enough to clear a fault like that too. If a supply
is not designed properly, it may be overly sensitive to loading,
cutting out and latching off at random.

Some previous generation motherboards, use "put-put" mode for Vcore.
What that means, is if Vcore detects an overload, the Vcore circuit
may continue to try to start afterwards. The circuit gets a bit warm,
and the nice thing is, it doesn't latch off. Generally, latch-off
is useful for circuits where the designers want to protect things to
the max extent possible, meaning fewer fires/smoke etc. But the
nuisance of having to flip the switch or unplug, to clear the latch.


I'll post back to the thread when I've got any results worth reporting.

Thank you very much for your time and efforts,

Joe


What most people don't think about, is it is a miracle that this stuff
works in the first place. That is why I'm not surprised when a computer
works flaky. The flakiness in computers only really started to disappear,
once good design methods, spread from engineer to engineer. There was
a time, when some companies simply didn't understand how to build stuff.
Now, companies can make "Xerox copies" of reference designs, and know
very little of what they're building. Kinda useful in the video
card industry. And to a much lesser extent with motherboards.

Paul

"Paul" wrote in message ...
jt3 wrote:

Yes, I've been using the GigaByte monitoring utility, about which I have

snip

nuisance of having to flip the switch or unplug, to clear the latch.


I'll post back to the thread when I've got any results worth reporting.

Thank you very much for your time and efforts,

Joe


What most people don't think about, is it is a miracle that this stuff
works in the first place. That is why I'm not surprised when a computer
works flaky. The flakiness in computers only really started to disappear,
once good design methods, spread from engineer to engineer. There was
a time, when some companies simply didn't understand how to build stuff.
Now, companies can make "Xerox copies" of reference designs, and know
very little of what they're building. Kinda useful in the video
card industry. And to a much lesser extent with motherboards.

Paul

Should you still be monitoring this thread, I'm posting back, not because I
have any noteworthy info, but more, the absence of it :-)
The machine has been working just fine ever since, with virtually no hiccups
of any sort. Works on either power supply, seemingly just fine. The only
questionable occurrence has been today when the GigaByte monitor wouldn't
give a reasonable reading--giving trash like CPU temp=96 deg C, fan speed
46,000 rpm, etc. When I would try to check any further, XP would tell me
that it had to close it down, and would I like to send a message to MS? I
love that one.
Anyhow, rebooting the machine seemed to fix that, though it is a puzzle,
since I would presume it reads it through their version of the Award BIOS,
and that seems passing strange, especially since the app was trying to
report battery voltage as well (that was about 46,000 volts, as well!) which
was not normal, as well as not reporting a number of the normal parameters.
Still, nothing to be especially worried about, at least, not yet!

Joe

jt3 wrote:
"Paul" wrote in message ...
jt3 wrote:
Yes, I've been using the GigaByte monitoring utility, about which I have

snip

nuisance of having to flip the switch or unplug, to clear the latch.

I'll post back to the thread when I've got any results worth reporting.

Thank you very much for your time and efforts,

Joe

What most people don't think about, is it is a miracle that this stuff
works in the first place. That is why I'm not surprised when a computer
works flaky. The flakiness in computers only really started to disappear,
once good design methods, spread from engineer to engineer. There was
a time, when some companies simply didn't understand how to build stuff.
Now, companies can make "Xerox copies" of reference designs, and know
very little of what they're building. Kinda useful in the video
card industry. And to a much lesser extent with motherboards.

Paul

Should you still be monitoring this thread, I'm posting back, not because I
have any noteworthy info, but more, the absence of it :-)
The machine has been working just fine ever since, with virtually no hiccups
of any sort. Works on either power supply, seemingly just fine. The only
questionable occurrence has been today when the GigaByte monitor wouldn't
give a reasonable reading--giving trash like CPU temp=96 deg C, fan speed
46,000 rpm, etc. When I would try to check any further, XP would tell me
that it had to close it down, and would I like to send a message to MS? I
love that one.
Anyhow, rebooting the machine seemed to fix that, though it is a puzzle,
since I would presume it reads it through their version of the Award BIOS,
and that seems passing strange, especially since the app was trying to
report battery voltage as well (that was about 46,000 volts, as well!) which
was not normal, as well as not reporting a number of the normal parameters.
Still, nothing to be especially worried about, at least, not yet!

Joe


Just don't shake the machine or look at it sideways :-)

Must be the good karma that makes it behave.

As for the flaky readout, the problem could be something
to do with RAM. The other day, I had a game that started reporting
inconsistencies in some data structure it was using, and a reboot
fixed it. None of my machines have ECC RAM, so I'll never know
whether the problem could have been RAM or not.

Paul

Yes, I've kicked myself repeatedly that I didn't get ECC RAM when I put this
thing together. It would certainly have made sorting through all this mess
a lot simpler. Though, with the motherboard not handling the memory
properly, as it was doing at the last before the RMA and repair, it might
have been less confidence inspiring than I thought.
But, as you say, don't give the machine any dirty looks, it might take
offense :-)

Thanks for all your input,
Joe

"Paul" wrote in message ...
jt3 wrote:
"Paul" wrote in message
...
jt3 wrote:
Yes, I've been using the GigaByte monitoring utility, about which I
have

snip

nuisance of having to flip the switch or unplug, to clear the latch.

I'll post back to the thread when I've got any results worth
reporting.

Thank you very much for your time and efforts,

Joe

What most people don't think about, is it is a miracle that this stuff
works in the first place. That is why I'm not surprised when a computer
works flaky. The flakiness in computers only really started to
disappear,
once good design methods, spread from engineer to engineer. There was
a time, when some companies simply didn't understand how to build
stuff.
Now, companies can make "Xerox copies" of reference designs, and know
very little of what they're building. Kinda useful in the video
card industry. And to a much lesser extent with motherboards.

Paul

Should you still be monitoring this thread, I'm posting back, not
because I
have any noteworthy info, but more, the absence of it :-)
The machine has been working just fine ever since, with virtually no
hiccups
of any sort. Works on either power supply, seemingly just fine. The
only
questionable occurrence has been today when the GigaByte monitor
wouldn't
give a reasonable reading--giving trash like CPU temp=96 deg C, fan
speed
46,000 rpm, etc. When I would try to check any further, XP would tell
me
that it had to close it down, and would I like to send a message to MS?
I
love that one.
Anyhow, rebooting the machine seemed to fix that, though it is a puzzle,
since I would presume it reads it through their version of the Award
BIOS,
and that seems passing strange, especially since the app was trying to
report battery voltage as well (that was about 46,000 volts, as well!)
which
was not normal, as well as not reporting a number of the normal
parameters.
Still, nothing to be especially worried about, at least, not yet!

Joe


Just don't shake the machine or look at it sideways :-)

Must be the good karma that makes it behave.

As for the flaky readout, the problem could be something
to do with RAM. The other day, I had a game that started reporting
inconsistencies in some data structure it was using, and a reboot
fixed it. None of my machines have ECC RAM, so I'll never know
whether the problem could have been RAM or not.

Paul