Re: CPU Temp Question Concerning Core 2 Quad

GSV Three Minds in a Can wrote:
Bitstring <6ghnp3p7pup116oks9os30jn19s8nnvpph@xxxxxxx>, from the wonderful person soinie@xxxxxxxxxxxx said
I'm running a Core 2 Quad 6600 2.4 ghz and the temp when idle is about
25C according to Asus PC Probe II, although I've seen it up to 34C
with tasking. I want to moderately over clock this (P5E WS Pro) setup
and would like to know what to look for in terms of temp increases and
what might be considered a threshold temp. I'm using a Scythe Mini
Ninja and Arctic Silver 5, and thanks to those who recommended an
after-market heat sink.

Heat is not going to be the problem - there is a thermal protection mechanism on the die, iirc, and probably another on the motherboard, and you are ~30 degrees the right side of a problem anyway. If the CPU gets too hot it will throttle the clock back.

If you (have to, or choose to) ramp the VCore up too far in pursuit of extra GhZ then you may fry something .. but It'll be volts and not degrees C that does the damage.

Before you go overclocking though, try a PROPER stress test at your current clock rate - Prime95 torture test (maybe 4 copies, one pointed at each core - I don't think the current one knows about loading more than one core at once).

34c is 'loafing' IMO, but then it's hard to really work 4 cores in everyday use. This assumes it's being measured right - not always the case with Asus Mobos (what's the case temperature registering at? Heck, how cold is the room it is in??)


I would find a utility that uses the internal CPU temp sensor. I haven't
tested these. They read out the digital temp sensor. Programs like
Asus Probe would likely be reading the analog sensor (CPU diode + hardware
monitor chip).

Intel Thermal Analysis Tool (TAT) - this was not intended for public
distribution, but is used by overclockers. There may be a later version,
but I'm not going to search all over for it. This is just the first
copy I found.

http://www.techpowerup.com/downloads/392/.html

This one does a similar measurement, and was not written by Intel.

CoreTemp
http://www.alcpu.com/CoreTemp/

Both of those programs rely on knowledge of Tj_max. That is the maximum
junction temperature before the processor starts to throttle. The reason
the measurement programs need to know that constant, is the measurement
mechanism has a "relative" and not an "absolute" temperature readout.
In other words, if Tj_max is 85C and the measurement returns a value
of zero, then the calculated temperature is 85C. If the value returned
is two, then the calculated temperature is 83C. Something like that.
So it is important to know the Tj_max of the processor types. A 65nm
processor could have a different Tj_max than a 45nm processor. The
various families, because they have different die sizes and quantities
of cache, and different power dissipation numbers, will also have a
different Tj_max. So the measurement tool needs a table of values, to
compute an absolute value from the relative readout. When those
two measurement programs give a wrong result, the theory is that
the Tj_max value is wrong.

Maybe Speedfan (almico.com) can read the digital temp sensor.
The release notes here, says the capability to read it was
added in version 4.32. Current release is 4.33.

http://www.almico.com/speedfan.php

So the issue with programs like that, is that they've been updated
with the latest values of Tj_max.

To load the CPU to 100%, this version of Prime95 will run on all
four cores at once. To use this program, when it asks to "Join
GIMPS ?" say No. Next, a custom dialog will appear. The maximum
memory to be tested will be shown in the lower right hand
corner. On my 1GB machine, it shows 767MB as the amount of memory
to test. I edit this value, and select something smaller, like
200MB. By doing that, enough memory is left on the computer, so
I can do other things while Prime95 runs. You can leave the other
settings as is. When you are finished with that dialog, the program
will launch a number of threads. For a Q6600, you should see four
threads launched. And your processor should start to heat up. The
program has a "stop" and "exit" option in the left-most menu,
so when you're finished, you can use that menu to get rid of the
program.

http://www.mersenne.org/gimps/p95v255a.zip
http://www.majorgeeks.com/Prime95_d4363.html (if mersenne is down, use this one)

When I want to do an actual integrity test, and not just heat up
the CPU, then I might set the memory to test back to the 767MB
value, as then as much of the available memory is tested as
is possible.

Those are the tools I know about. Have fun :-)

Paul
.

Relevant Pages

  • Re: SpaceTime & memory.
    ... IOW's if one has no memory, ... cannot exist in measurement or preception. ... I further reason the PoN are independant of our ... description of nature. ...
    (sci.physics.relativity)
  • Re: SpaceTime & memory.
    ... cannot exist in measurement or preception. ... saying "an object with no memory has no time". ... I further reason the PoN are independant of our ... This is a step to far: our imagination is a description of nature. ...
    (sci.physics.relativity)
  • Re: system resource measurement
    ... accurate measurement of the amount of pages used incrementally any ... filesystem buffer cache. ... Don't be misled and look at free memory as an indicator. ...
    (Debian-User)
  • Re: Memory Leak
    ... Don't believe task manager. ... It's not really good measurement of memory your ... "Greg" wrote in message ...
    (microsoft.public.dotnet.languages.vb)