(Click on pics to enlarge for details)fire!

FixThProbe HeatBackView HeatingFront

Fixing calibrated therm.probe to the heatsink bottom closest to the CPU core ; sensor already fixed + connected on-Die CPU diode from backside of a slotket ; before starting heating should both sensors reveal room temps. Heating the HS/CPU combo with applied AS-II therm.paste with a hair dryer for 5 min till it reaches constantly 86C (onDie diode should reveal 85C after 5 minutes or so also constantly (that onDie diode reaches practically same HS temp & stabilises there! for 0,13u CPUs, for older ones a bit higher temps can be used), if the "measuring Adapter"=gadget is properly calibrated too..) Than I repeated the procedure (cooling/heating) few times during a day to check the gadgets calibration considering also approx.1C drop or so on inner diode, because of not perfect match (th.paste!) between HS & Cpu core measured. ( I checked it also for subzero temps putting this setup into the fridge freezer! too..). This way measuring HS temp I used also for pre-calibrating the gadget electronics range! [Btw, did you know that water starts freezing @ 0C & boiling @ 100C at sea level air pressure? - pop in that water a probe & check temps; this could be for any thermal probe calibration CHECK - a FAST homemade solution!] Burnin-in ...the CPU helps OverClocking IMHO...

But this is almost nothing... The previous & current one Cpu I was pre"burn-in" it inside frying pan on the stove! It helps phase change/melting of the AS-II paste (it does it around 55-60C if I am well informed) onto the Cpu & heatsink surface in-between much faster than normal 200h operation recommended for it (a week) or so. Other milder way is to run a Cpu/HS combo without a fan for awhile, but can be dangerous for AMD CPUs ... .. the results are revealed under Benchmarking section...no s**t !

the Gadget ! - b/w slide of my "measuring" prototype adapter hanged on a curtains .. This Gadget is described under Electronics section, (lazy me, it took me almost a year to make it on a proto-board & mounted it /left lower corner in a box/ inside PC case ...)

. . . Time ago Cpu temps revealed using only thermistors in socket were more accurate (example PPGA Intel CPUs), because the Cpu core was facing socket; now these days because of better cooling it is flipped on outer side towards HSF & so even less accurate than ever before for sure!

How to be certain, if your MoBo is really showing Cpu OnDie diode temps & that are shown properly?

It is not so easy to determine that sometimes!

IMHO, MoBo using the CPU onDie diode is one thing & showing the temperature more or less precisely are 2 different things (properly calibration is involved here & that depends also on each generation of CPUs!) & the MoBo manufacturers mostly do not like us to see the REAL high Cpu temps inside it (why is another question!), but there are exceptions, mostly from the server board manufacturers.

Yes, there are also newer boards using Cpu onDie diode, but mostly for overheating protection & not mainly for displaying the temperature from it but rather using in_socket probe for that like Epox did it with last ones after their support being flooded from their users, complaining about too high Cpu temps (in that case REAL ones!) shown from some their few previous boards.

Internal Cpu Diode does NOT linearly follow ambient temperature changes in the same amount if Cpu is ordinary cooled, but in ... Need2know

. . . . . reduced range !

  How can you find out, if temps reported from Cpu onDie diode are kind a correct ones with these days so power demanding CPUs:

- cold machine booted into Windoze straight, should show approx. between 15 & 20C higher CPU temp than the ambient one immedeately! (booted straight in to the Bios first should reveal there immediately approx. 10C or even more higher temperature than the ambient @ example 20C one)

- than in the next five to ten minutes, the temperature should rise for approx. another 10C without doing anything (idle) & using a SoftCooler program (HLT instruction engaged) & stay there for good. [Without SoftCooler program or it not working properly /no proper HLT instruction/, should rise here few C even more, but than less the same amount under heavy load further]. This I call it BASIC temp which is higher with more power demanding CPUs naturaly, (mine max. 35W only) !

- after for awhile running a Cpu "heavy load" proggy like SiSoftSandra CPU Burn-in Wizard or some other like that, the temp should rise for almost another 10C quite fast (in few minutes or so, for Tuallies only 5C more or so, because looks like SoftCoolers do not work properly with them!).


If you would follow posts on the Net about certain temps reported from users with different newer MoBos & frequencies of their different CPUs, even if are made on same 0,13u process like mine (which consumes approx. 35W full load@1,35G) & mine temps air-cooled, considering also room temps, you could find out, that there is something iffy about their temperature reporting!

. . . . . . There is no way that somebody just with a solid made air cooling @ same room temps like mine & a CPU consuming more power at higher Vcore & a double the speed of mine (& almost doubled thermal dissipation at full load), could have same like mine temps or even lower ones !!! Well, with Water-Cooling setup things are a bit different.

It is just common sense & my curiosity that lead me to made such an experiment from which results can be seen on this sites about MINE temps in different conditions!

Some "consumer" boards actually do show the temps correctly, but they were quite rare IMHO, but lately things are getting better with that... & mine temps are a bit higher than they should be, because I don`t wanna have an airplane (or hovercraft ) infront of my nose..

  If this is not a case, there is something wrong with reporting temps or are NOT really measured from the Cpu core but instead maybe from some other therm.probe laying somewhere around & just maybe in Bios "fixed" (thru it & SMbus programmed HWmon. chip so) to show for example higher than 30C every C like 2 of them or so !!!

But can be also this kind of case; and if you "flash" the bios, than reporting changes, usually to lower values, interesting that ! Also any Win program being used for temperature revealing (like famous MBM) should be adjusted properly (under settings or properties).

Maybe the easiest way is @ idle to check with external calibrated instrument the temperature on the bottom of the HSF running near Cpu core & just add another 10C or so (for Water-Cooled setup, few C more against the bottom of the WaterBlock would be Ok IMHO) & than just adjust the monitoring program! Not 100% accurate, but for everyday use for average user just enough good IMHO! For "advanced" HardWare guy like me, maybe that is not enough & so involves some knowledge in electronics to make an additional circuitry of some kind to do that job of measuring temps properly; it can be like mine analog one or independent digital PIC micro-controlled one (also expencier & always involves soldering on the MoBo - tricky a bit that!). . . .