E8500 C0 Temp which one is the real one?

[kinnsy]

RealTemp idle - 32C

CPUID Harware Monitor idle - 37C

SpeedFan idle - 42C

I did notice that the difference is 5C or a multiple of that, any significance?

[Kelainefes]

Truth is none of the two, the sensors in 45nm chips are crap. I take a peek at the CPU sensor and not at the cores when needed it seems to me that it is much more accurate, just a little offset (reads -5 to -10C).
Anyway i think that the difference is because Realtemp uses the wrong TJMax, 95C instead of 100C; you can correct that in the settings if needed.

[kinnsy]

Thanks, I was tending to the middle reading anyways.

[kitfit1]

I always used Speedfan to watch temps (because it works at - temps). I always had it set to a -10 offset, because that then tallied with the idle temp of my Mach2. So Kelainefes is about right, speedfan with a -10 offset should see you about right.

[Kelainefes]

What i meant was that with my E6750, wich being 65nm has decent DTS sensors, i had to set the CPU temp to +8 to match the core temps under load.
Maybe below zero it's different, i wouldn't know.

[Billabong]

Originally Posted by Kelainefes

Anyway i think that the difference is because Realtemp uses the wrong TJMax, 95C instead of 100C; you can correct that in the settings if needed.

You're slightly incorrect there, only the old versions use 95°C, since the Tj document released at the August IDF Conference unclewebb has used 100°C as the TjMax for 45nm (version 2.70 onwards I believe) and in Oct at the Taipei IDF Conference they released the 65nm "Target Tj" information which has been incorporated in subsequent beta releases.

The latest version 2.84 Beta has an enhanced sensor test that utilises prime95.

@OP, if your version of Real Temp is using 100°C for TjMax then it is the correct reading using this information.

My E8600 at 4.0GHz (1.225V bios) under a Thermalright Ultra 120 eXtreme idles at 29 - 32°C depending on the time of day though I have one faulty sensor that will not read below 35°C they both function accurately above that.

With regard to idle temps, Intel has repeatedly stated that the sensors were never intended to report idle temps and therefore probably deliver erroneous readings, unclewebb has a calibration process with regard to Real Temp that you can use, it involves lowest vcore/FSB/multi settings in the bios to give a reading that "should" approximately be ~5°C above ambient air temp.

For more information I suggest you refer to his thread at the XS Forums, it is a good read especially towards the end with the results of rge's test drilling of the IHS on multiple cpu's and the resulting "inside the die" temp readings achieved using a thermal probe.

Cheers.

[kinnsy]

Gordon Bennett Billabong, you read all that thread?

So, its TJMax of 100 in realtemp and If I'm reading things right no readings are reliable/accurate until 35C and above?


Oh yea, thanks all.

[Kelainefes]

No readings are reliable until you are very (dangerously) close to the TjMax, according to Intel.

BTW RealTemp 2.70 uses 95C i just downloaded the 2.84 Beta and it uses 100C.

[Billabong]

Originally Posted by kinnsy

Gordon Bennett Billabong, you read all that thread?

So, its TJMax of 100 in realtemp and If I'm reading things right no readings are reliable/accurate until 35C and above?


Oh yea, thanks all.

The 35°C is only for my one faulty sensor on my particular chip, it is by no means a yard stick

Kelainefes is nearly correct below with what he states though the line of divergence is substantially lower than he intimates, there are graphs showing the scaling divergence in that thread. I also have the .pdf from the IDF Conference that shows the same but I do not have it here at work, and yes, I actually have read the whole thread, though it was a lot smaller when I first started using Real Temp and reading it (Q6600 release), now days it's just a matter of keeping up with the last two - three pages.

Originally Posted by Kelainefes

No readings are reliable until you are very (dangerously) close to the TjMax, according to Intel.

BTW RealTemp 2.70 uses 95C i just downloaded the 2.84 Beta and it uses 100C.

My bad, iirc it was around 2.70.4 that 100°C was adopted

[Kelainefes]

Intersting, i'll have a read of that thread and see if i can find out where my cores really at.
If the PDF you got includes graphs of the divergence betweenr the actual temp and the rported temps i would be most grateful if you could post a link!