motherboard overreporting temperatures?

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
qviri
Posts: 2465
Joined: Tue May 24, 2005 8:22 pm
Location: Berlin
Contact:

motherboard overreporting temperatures?

Post by qviri » Wed Jun 22, 2005 3:57 pm

Hey,

when I turn on my computer after it was on since the previous day and go into BIOS right away, the motherboard reports 28 deg C for both SYSTEM and CPU, and the temperatures slowly (~1 C/minute for CPU) rise from there.

There's no way any of the components on this board were warmer than ambient before powering up, and I was in BIOS about 15 seconds later. Ambient was 23 C tops.

Did the components really heat up this quickly? I think it sounds highly, highly unlikely. Can I assume my board overreports the temperatures by ~5 C then?

Would it keep the same 5 C difference at higher temperatures (eg, 53 C real/58 C reported), or would it change?

swivelguy2
Posts: 404
Joined: Sun Jul 04, 2004 9:18 pm
Location: Illinois, USA

Post by swivelguy2 » Wed Jun 22, 2005 4:54 pm

Did the components really heat up this quickly?
I thought about this some, and I ended up doing some thermodynamics. This will be rather mathematical and very likely quite unhelpful, but here it is.

Let's say your CPU has a thermal mass C (Joules/Kelvin) and outputs a constant power P (watts). It is connected to the outside air by the heatsink, which has a thermal conductance S (watts per kelvin). The total amount of heat in the CPU is H = TC where T is the temperature of the CPU relative to ambient.

Then the total heat change of the CPU is: dH/dt = P - ST

We can re-express this: C(dT/dt) = P - ST

then we can separate variables and integrate to get the result

T = P/S + (Ti - P/S) * exp(-S/C * t)

where Ti is the temperature at some arbitrary time 0. From this, we see that after infinite time, we expect the CPU temperature to be P/S, as we would expect from simple thermal equilibrium.

If we knew (or guessed) P, S, and C, we could easily determine the answer to the question "Did the components really heat up this quickly?" Or, somebody else with a better sense for these things might have an intuitive answer.

Anyway, what I do know is that during startup and while in BIOS, the CPU runs at full-tilt, 100% power. So it is quite possible that the CPU did heat up 5 degrees in 15 seconds - especially because at such close temperatures to ambient, the heatsink will do almost nothing.

Tibors
Patron of SPCR
Posts: 2674
Joined: Sun Jul 04, 2004 6:07 am
Location: Houten, The Netherlands, Europe

Post by Tibors » Wed Jun 22, 2005 5:33 pm

If your CPU gets some load, then it heats up very quickly. 15 seconds may sound fast, but it isn't for temp changes of a CPU. So you can't tell from this that the mobo overreports by 5°.

Use the method described in Calibrate Your CPU Temp Reporting to find the answer.

qviri
Posts: 2465
Joined: Tue May 24, 2005 8:22 pm
Location: Berlin
Contact:

Post by qviri » Sun Jun 26, 2005 6:12 am

Tibors wrote:Use the method described in Calibrate Your CPU Temp Reporting to find the answer.
I see that my motherboard isn't reporting the CPU diode temperatures, seeing as no change is reported for about 15 seconds after engaging CPUBurn. Damn cheapo motherboard.

I guess it's all relative then. Any recommendations as for critical socket temps for an Barton mobile with max die temp listed as 95^C? I'm thinking up to 60s should still be safe.

alglove
Posts: 363
Joined: Fri Feb 06, 2004 11:21 am
Location: Houston, TX, USA

Post by alglove » Mon Jun 27, 2005 1:22 pm

For whatever it's worth, my motherboard shows a 10 degC difference between socket temperature and die temperature when the CPU is under load. For example, when socket temp is 56 degC, die temp is 66 degC. MSI KT880 Delta-FSR motherboard, Athlon XP 3000+ CPU.

qviri
Posts: 2465
Joined: Tue May 24, 2005 8:22 pm
Location: Berlin
Contact:

Post by qviri » Mon Jun 27, 2005 5:19 pm

alglove wrote:For whatever it's worth, my motherboard shows a 10 degC difference between socket temperature and die temperature when the CPU is under load. For example, when socket temp is 56 degC, die temp is 66 degC. MSI KT880 Delta-FSR motherboard, Athlon XP 3000+ CPU.
All right, thanks, I have a rough reference value now.

GDukeman
Posts: 27
Joined: Thu Jun 16, 2005 7:28 am
Location: Colorado, USA

Post by GDukeman » Wed Jun 29, 2005 1:54 pm

Most posts I've seen about socket vs die show about 10c difference so it is a good rule of thumb. Also, even through your motherboard does not show the temp, SpeedFan may be able to access it. It finds way more sensor capable components on my system than the BIOS or even MBM shows.

qviri
Posts: 2465
Joined: Tue May 24, 2005 8:22 pm
Location: Berlin
Contact:

Post by qviri » Wed Jun 29, 2005 2:13 pm

Well, Speedfan finds a third sensor, however, readings off that one vary wildly from a maximum of 74^C to a minimum of -7^C, fluctuate with no obvious pattern and actually fall after engaging CPUBurn...

I heavily doubt that one will be of any use, but thanks for the suggestion, made me double-check.

Post Reply