jg said:
Thanks for you detailed reply, Paul.
I must admit I'm not litterate in the sensors and thermistor domain, nor
in the inner workings of these devices (while it is always interesting
to learn something new).
My post was based on observations I made that seems odd to me : a
"case" temp higher than the CPU temp reported by both MBM and the mobo
Bios.
Am I correct if I understand from your reply that because the Mozart2
datasheet was never publicly released, there's a chance that utilities
software such as MBM or Everest report wrong values ?
And while P4G533-LA is not listed here, you can see
some similar motherboards that use Mozart2. For example
P4B533-VM has Mozart2-2 for Case and Mozart2-1 for CPU.
http://web.archive.org/web/20060107075602/mbm.livewiredev.com/comp/asus.html
You say that P4G533-LA is not listed on that page, but there's an P4G533
(without -LA suffix) listed (-LA probably meaning "with integrated Lan
and Audio ?) listing different chip and sensor set than Mozart:
SMSC LPC47M192 : sensor chip
LPC47M192 1 : case sensor
LPC47M192 2 : cpu sensor
In the MBM Configuration wizard, there's only one entry for this board,
and it includes "all revisions".
To make sure I didn't initially select the wrong board, I ran the MBM
wizard again, and here's what the "System Info" (with default settings)
reports:
------Sensor Chip-------
Main Sensor Chip : Asus Mozart2
SMB Address : $E800
Used Bus : SMB CHIP Intel ICH4
Chip Address : $77
-Sensor Chip Selections-
Possible Temp Sensors[1] : Mozart2-1
Possible Temp Sensors[2] : Mozart2-2
Selected Temp Sensors[1] : LPC47M192-1 (406)
Selected Temp Sensors[2] : LPC47M192-2 (407)
Possible Voltage Sensors[2] : MBM Fixed
Selected Voltage Sensors[1] : MBM Fixed (2)
Selected Voltage Sensors[2] : MBM Fixed (2)
Selected Voltage Sensors[3] : MBM Fixed (2)
Selected Voltage Sensors[4] : MBM Fixed (2)
Selected Voltage Sensors[5] : MBM Fixed (2)
Selected Voltage Sensors[6] : MBM Fixed (2)
Selected Voltage Sensors[7] : MBM Fixed (2)
Possible Fan Sensors[2] : Mozart2-1
Possible Fan Sensors[3] : Mozart2-2
Selected Fan Sensors[1] : LPC47xxxx-1 T1 D4 (94)
Selected Fan Sensors[2] : LPC47xxxx-2 T1 D4 (95)
Possible Voltage Sub Types[0] : Mozart2 Standard
Selected Voltage Sub Type : Error (1)
--Sensor Chip Readings--
Temperature Readout[1] : 31
Temperature Readout[2] : 0
=============================
I also run Lavalys Everest and here's what the "Sensors" section reports:
Sensor properties
- Sensor type : Asus Mozart-2 (SMBus 77h)
Temperatures
- Motherboard 55 C (131 F)
- CPU 32 C (90 F)
- Maxtor 4R120L0 34 C (93 F)
Cooling Fans
- CPU 1687 RPM
I wonder why Everest does not report the case fan (while it is in the
BIOS) but that's another story.
What puzzles me is the motherboard temp being so high compared to the
CPU temp. BTW, it's an Intel Celeron 2.7Ghz.
What do you think ?
What I was looking for first, was some background on the thing.
P4G533-LA is not the same as P4G533. There are chip differences, like
a difference hardware monitor.
Some people, based on the available info, would try searching for the
thermistors. I don't know how they find them, but some people have
managed to track down which components measure the temperatures. I've
never had much luck myself.
A second issue, is how do the monitor program writers figure out the
arithmetic necessary for each sensor. For voltage measurement, there
can be scaling resistors, so there is an arithmetic function to fix that.
For temperature, the thermistor has a beta (like 3435) and there is
a nonlinear equation to compute the temperature. Some monitor chips
have a look up table, but it still has to represent the non-linear
behavior of NTC thermistors or diodes.
The temperature equation usually includes a "fudge factor" or temperature
offset. Take the example of the thermistor in the CPU socket. The
thermistor may be touching, by means of a mechanical connection. (Some
motherboards have a gadget to help the thermistor sample the temp of the
bottom of the CPU. When the thermistor doesn't touch, it measures a temp
much less than the real CPU temp. Also, the dynamics would be wrong.
On my computer, I'm currently running MBM. If I start a game, the
temp shoots up in a matter of seconds. If I alt-tab out of the game,
it comes down over a matter of 30 seconds perhaps. And this is because
I have a diode sensor on the CPU die. A thermistor will respond much more
slowly.
The person writing the monitor program, has to "guess" as to how much
diff there is between the real CPU temp and the actual temp. The
temp offset used by Asus Probe, by the BIOS, or by MBM or Speedfan,
could be quite different. Sometimes people get scared needlessly,
because of the offset that is applied.
So, I hope you can see that by using a thermistor, if the info I dug up
is accurate, there is much less chance of getting an accurate reading.
In the recent past, I wasn't running MBM, because everything on my
machine seemed fine. Now that summer is here, I noticed my PSU was
getting hotter than normal (actually by the smell of the thing). My
machine runs Win2K, and is supposed to be using the HALT instruction.
Now, some software I added to the machine, seems to have fouled that up.
I got a copy of RMClock, and there is an option in there, to launch an
idle loop routine, for each core on a processor. The Task Manager no longer
accurately reflects processor loading (it will always read 100% CPU), but
RMClock has its own display of the real CPU load. The RMClock idle loop
(RMCLockHLT.exe) is the thing you see in the Task Manager. And when it
is running, it uses the HLT instruction. This cools the processor, as the
processor enters a lower power state, until it is interrupted by a clock
tick interrupt or the like. This does not help with real 100% CPU loading,
as the temp will still shoot up. But it does reduce the temps when not
gaming, for example.
(Install the program, then look under "Management", click "Run HLT"
if and only if you think your OS is not using HLT. You can always
reboot and stop using RMClock, if there is no improvement. My temps
dropped by more than 10C by doing this. My processor draws about 13W
when using HLT.)
http://cpu.rightmark.org/products/rmclock.shtml
http://cpu.rightmark.org/download/rmclock_225_bin.exe
A typical RMclock display is here (second picture down):
http://www.coolaler.com/forum/showthread.php?t=117069
I'm not saying you are going to learn more than you already know. The
sensors could be mislabeled. That is always possible. The dynamics of
temperature rise and fall, would have helped identify the sensor type,
if the processor diode method was being used. (CPU would rise fast when
you start a game.) But if both sensors are thermistors, their dynamics
will be similar. The offset is purely an estimate, used by the monitor
program writer, and helps compensate for the physical reality of a
thermistor based CPU sensor, not making contact with the bottom of
the CPU. The diode method, at least gets the sensor in the right
place, as it is part of the CPU silicon die.
So start a 100% CPU program running, and watch the sensors. Does the
CPU sensor shoot up immediately ? Do the sensors seem to ignore how
loaded the CPU is ? Those will help provide clues as to what's up
with your machine.
Paul
