Why are there such discrepencies in CPU temperature readings?

[PhotoLab]

<<The P4 does that via its internal hardware, you don't even need ACPI for it.>>

Yes, P4 does....but the question is does your motherboard, bios, and OS correctly implement ACPI standard today?

Do you hear your fans [CPU/Pwr/Case] changing speeds, or do they run same speed always?

 

[Peter]

With BIOSes from 2000 or later, ACPI implementations usually are quite OK. Windows 2000 and particularly XP make good use, but on desktop machines don't do fan control (yet?). The problem here is that starting/stopping fans can be done safely, but fan _speed_ control is hardly doable if you don't know what fan the user puts on the board. In a notebook or other fully controlled assembly, this can be done and is being done.

 

[PhotoLab]

Yes, the BIOSes today should have ACPI down pat, but the OSs I wonder. I was looking through ACPI specs (cited earlier) for OS, and it can be done via OS, but there would have to be a thermal spec database (I think) for each motherboard model (like Video Cards). I just don't have the time to research this tidbit.

On an Intel Board, like the D850GB (423 pin), their literature explains how two fans's speeds are controlled by thermal changes. The other two fans run at full power voltage. Four total hookups.

One purpose of controlling these two fans was to decrease noise. Somewhere around 36 degrees C, is where active cooling is engaged.

Fans aren't suppose to "start/stop" but run constantly; you change the voltage going to fan, and this changes RPM. Lower RPM, and noise drops off.

See fan specs at bottom of this URL

http://www.dynatron-corp.com/products/c...ngfan/coolingfan_model.asp?id=9&cid=32

Notice CFM, RPM, and dBA numbers; as you increase RPM, the noise (dBA) and airflow (CFM) increases.

Hence, if you don't hear any fan speed changes in your computer, when it reaches around 36 C, and more, there is a good chance ACPI is not fully implemented.

BTW, ACPI in Bios may only affect CPU, not fans (at least in my 2000 Award Bios). There, you control CPU throtling by 12.5% duty cycle increments of either on mostly or off mostly states.

 

[PhotoLab]

Beatle:

I think I found the answer to your question via Intel pub.


Intel® Pentium® 4 Processor
In the 423-pin Package
Thermal Design Guidelines

8.5.1 Thermal Diode


The processor thermal diode should not be relied upon to turn on fans, warn of processor cooling system failure or predict the onset of thermal control circuit. As mentioned earlier, the processor?s high thermal ramp rates make this unfeasible. An illustration of this is as follows. Many thermal diode sensors report temperatures a maximum of 8
times per second. Within the 1/8th (0.125 sec) second time period, the thermal diode temperature is averaged over 1/16th of a second. In a worst case scenario where the silicon temperature ramps at 30°C/sec, or ~3.75°C/0.125 sec, the processor will be ~3°C above the temperature reported by the thermal diode sensor. (Change in diode
temperature averaged over 1/16th seconds = ~1°C, temperature reported 1/16th second later at 1/8th second when the actual processor temperature would be 3.75°C higher, see Figure 14)

 

[PhotoLab]

This URL may be of interest.

 

[Peter]

Originally posted by: PhotoLab

Yes, the BIOSes today should have ACPI down pat, but the OSs I wonder. I was looking through ACPI specs (cited earlier) for OS, and it can be done via OS, but there would have to be a thermal spec database (I think) for each motherboard model (like Video Cards). I just don't have the time to research this tidbit.

BTW, ACPI in Bios may only affect CPU, not fans (at least in my 2000 Award Bios). There, you control CPU throtling by 12.5% duty cycle increments of either on mostly or off mostly states.

You might want to read up on ACPI "Thermal Zone" objects. These include temperature measurement, as well as active (fan) and passive (throttling) cooling devices. Both active and passive cooling is triggered to BIOS(!)-provided temperature thresholds.

Thus, the OS need not know the particular system's thermal properties, it gets them provided by the system's BIOS. (That's one of the main points of doing ACPI after all - provide abstraction of design-specific stuff to a generic OS.)

 

[PhotoLab]

>Both active and passive cooling is triggered to BIOS<

See: Windows 98 Does Not Support ACPI Passive Cooling Mode

"Windows 98 does not provide support for Passive Cooling Mode as described in the Advanced Configuration and Power Interface (ACPI) specification. Windows 98 only supports Active Cooling Mode."

OS is involved in process, and yes, BIOS is too...