What are the normal CPU temperatures for a P43.0c?

External Since: Aug 23, 2004 Posts: 3

I replaced the stock Pentium 4 3.0GHz Northwood heatsink from Intel with a
Zalman 7000-ALCU with some Arctic Silver 5 thermal paste. Everything is
fine, I just want to know what temperature I should expect to confirm
whether or not I installed everything correctly. I don't overclock, I was
looking for a more silent heatsink, and the Zalman has been great for that.

I have a P4C800-E Dlx, and it currently reports temperatures as follow:

Room temperature: 25 °C.

CPU idle: 33 °C
MB idle: 32 °C

CPU load: 48 °C (or so)
MD idle: 36 °C (or so)

These temperatures seems accurate, because when I start my computer in the
morning, they are very close to the room's ambient temperature.

I've heard people having their CPU runs at 29 °C on idle, so I'm wondering.
I applied the thermal paste as a rice grain drop in the middle of the CPU
before installing the heat sink on top - so the heatsink spread the paste
itself. I very carefully cleaned the CPU cover before, so I can certify it
was perfectly cleaned of the old thermal paste.

External Since: Jul 27, 2004 Posts: 2158

In article <iWqWc.75456$F85.1412835@wagner.videotron.net>, "povmec"
<raymond_hill DeleteThis @hotmail.com> wrote:

> I replaced the stock Pentium 4 3.0GHz Northwood heatsink from Intel with a
> Zalman 7000-ALCU with some Arctic Silver 5 thermal paste. Everything is
> fine, I just want to know what temperature I should expect to confirm
> whether or not I installed everything correctly. I don't overclock, I was
> looking for a more silent heatsink, and the Zalman has been great for that.
>
> I have a P4C800-E Dlx, and it currently reports temperatures as follow:
>
> Room temperature: 25 °C.
>
> CPU idle: 33 °C
> MB idle: 32 °C
>
> CPU load: 48 °C (or so)
> MD idle: 36 °C (or so)
>
> These temperatures seems accurate, because when I start my computer in the
> morning, they are very close to the room's ambient temperature.
>
> I've heard people having their CPU runs at 29 °C on idle, so I'm wondering.
> I applied the thermal paste as a rice grain drop in the middle of the CPU
> before installing the heat sink on top - so the heatsink spread the paste
> itself. I very carefully cleaned the CPU cover before, so I can certify it
> was perfectly cleaned of the old thermal paste.

You can estimate the temperature to expect, by taking

CPU_temp = Case_air_temp + HSF_thermal_resistance * CPU_max_power
= 36C + 0.25C/W * 81.9W = 56.5C

In this case, I'm using the thermal design power (TDP) for the
3.0C as listed in the datasheet. The datasheet will have a couple
of thermal resistance number, and I remember the values as roughly
about 0.25C/Watt or so. The estimate of case air, I'm basing on
the measured motherboard temperature, and some people find that
the placement of the MB sensor on the motherboard, doesn't give
a very good estimate of the real case air temperature.

So, in your case, either the case air temp is actually cooler than
the measurement suggests, or you haven't got the processor running
at 100% load. I use Prime95 for that, and there can be other more
pathological applications than that, for achieving 100% load. With
Hyperthreading, you might be able to use a second compute bound
program, to squeeze a little more heat out of the processor. I've
never tried that, as I have HT disabled currently.

I'd say you are doing pretty good for that heatsink.

In terms of idle CPU power, I've measured the power consumed by the
Vcore circuit, and on my 2.8C/FSB800 processor, it is 13.2W via the
12V feed. As power conversion could be 90% efficient, at the CPU it
is probably getting about 12W. Your idle temp cannot go below the
case air temp, and again we are stuck with the less than accurate
MB sensor -

= 32C + 0.25C/W * 12 = 35C

In other words, your idle CPU temp should be about 3C above
the actual case air temperature, if you can find a way to
measure the actual case air temperature. This of course assumes
that both out computers have the same kind of background task
computing load in Windows (not likely).

A good thermistor is a 1% tolerance device, but after it has
been soldered to a motherboard, I don't know if it maintains
that tolerance or not. Thermistors can shift if esposed to
mechanical shock or bending, and I haven't read anything on
how much shift in tolerance to expect from a soldering operation.

HTH,
Paul