weird realtemp reading

[simaomacaco]

Hi guys,

I have been playing with realtemp while doing some nab cpu clocking and am really curious how my core0 is way lower than the other 3 cores. Running i3750k @ 4.5 on stock volts. TJMax different as it seems unlikely I fitted the cooler crap on the other three cores.

 

[Ed_1]

its normal to see 8 -10c delta between core temps .

 

[simaomacaco]

Normal is an artefact of the factory calibration differences for each DTS? Do the reported temperatures in C vary because each DTS is effectively using a different scale for units from TJMax and/or is it that the TJMax (baseline) values vary. Trying to understand how this works hence the questions.

 

[Ed_1]

Normal is an artefact of the factory calibration differences for each DTS? Do the reported temperatures in C vary because each DTS is effectively using a different scale for units from TJMax and/or is it that the TJMax (baseline) values vary. Trying to understand how this works hence the questions.

Hi , I really am not that familiar with the cool down test but I did run it on my system 3570@4.0 and I got similar scaleing as you (I am little cooler (higher values) . the tests bring the CPU under high demand with prime95 then is lessens load (I did not check how it did (by size of fits maybe) .

Form what I understand the farther you get away from Tjmax the less accurate the temp reading is . Maybe there some documentation at there site on the test as to what to look for other than obvious how fast it cools down after load is lessened .

Edit: here documentation , should help .
http://www.techpowerup.com/realtemp/docs.php

 

[simaomacaco]

Yeah, I had a read of that already. I found this online too http://lenry.atw.hu/tjmax.pdf which does a reasonable job explaining some of how it works. My understanding is that the temperatures are broadly similar, maybe core 0 DTS counts in units of 1.05 whereas the other cores are some other unit maybe 0.97. This does not matter to intel as the design of the DTS was just to measure distance fom TJMax, these small variations in the actual scale of the units does not matter much. But also makes an accurate C hard from just using the DTS. Intel seem to recommend using an average from the four cores as the best measure, makes sense given my understanding.

 

[simaomacaco]

Actually the link that's was most useful was: http://www.techpowerup.com/forums/attachment.php?attachmentid=19489&d=1224601917

 

[Ed_1]

Thats a good link .
Only thing I am not fully understanding is the values displayed in sensor movement numbers above the cool down test (sensor movement values).
they get generated before test is run .

 

[simaomacaco]

I think that's variation at a steady load. The main point 'I think' is that the temperature readings are indicative, what matters is the distance for TJMax. It's probably best to consider the average of the cores if one were to want a measure in C, so as to mitigate the variations. Eitherway, as you
pointed out at the very start, 10 degree of delta is normal. . Given the inherent problems in a reported temperature I wonder if it would be better to express distance to TJMax as a percentage?

 

[Ed_1]

I think that's variation at a steady load. The main point 'I think' is that the temperature readings are indicative, what matters is the distance for TJMax. It's probably best to consider the average of the cores if one were to want a measure in C, so as to mitigate the variations. Eitherway, as you
pointed out at the very start, 10 degree of delta is normal. . Given the inherent problems in a reported temperature I wonder if it would be better to express distance to TJMax as a percentage?

yes, it just makes it more complicated once you know there no trouble with sensor (no locking) and results are close it doesn't really matter .
Once you know or see what value are normal , you just keep eye on it. if they change then check what cause could be .

I went through the calibration like on this site with IR gun and it wasn't off much .I would have to lower 1 core a few c and there around 4-5c above room temps with no load ,