I'm not sure whether this subject is now closed or still active. I just stumbled onto it.
It is a subject I have been trying to track down for several weeks because I bot a QX6700 a few weeks ago & it is overclocked & I don't know how far to go without damaging the cpu. I care about my $s too. I don't have any good answers, but I have a few stray opinions that might be useless & might be wrong.
I think the label "Tjunction" is used in at least three different ways. These are related, but not precisely the same.
****************
(1) The desktop Core 2 cpus have a partial metal heatspreader on top & then some sort of internal thermal interface material & then a silicon layer where the computation is done & the heat is generated & then a supporting board for the silicon. "Tjunction" is then sometimes used to refer to the temperature (that one might be able to measure) at the junction between the silicon layer & the supporting board. Of course, this temperature (that one might measure, at least in principle), will go up & down with computational load, with core voltage, with fan speed, with room temperature, etc.
*****************
(2) "Tjunction" is sometimes used to refer to THE MAXIMUM ALLOWABLE VALUE of the temperature I have defined above in (1). THIS SHOULD BE CALLED TjunctionMAX, but usually it is not. This number will not go up & down with any running conditions at all. It is a fixed tolerance spec set by Intel for each type of chip.
*****************
(3) Intel has implanted some type of thermal sensors somewhere in the silicon layer of their Core2 cpus.
These generate a digital signal that can be picked up & then used to calculate a temperature by using a formula provided by INTEL. This formula of INTEL looks approximately like this:
temperature = constant - digital_signal.
The constant in the above formula is called "Tjunction" by Intel. That is, Intel writes this formula as
temperature = Tjunction - digital_signal.
This third type of "Tjunction" is not something that is measured or can be measured at all. It is just a constant in a formula that has been chosen by Intel for its own use. It is different for different types of cpus, but does not vary with use.
*********************
When the CoreTemp software reports that "Tjunction = 100 C" (or whatever it reports to you) it is not reporting anything measured at all. The author of CoreTemp is simply telling you that CoreTemp has calculated the core temperatures by plugging the number 100 into Intel's formula & then used
temperature = 100 - digital_signal.
and reported the result to you as your core temperature. Of course, we all hope that the CoreTemp has correctly identified the type of your cpu & has gotten the correct constant assigned by Intel for your cpu. Otherwise GIGO.
*********************
Of the three different uses I listed above for the same label "Tjunction", the 1st & 3rd are just fixed numbers (for each type of cpu) that have been set by Intel. Then, has Intel chosen to set these two numbers to be equal? I hope so, but I'm not sure.
*********************
Tcase. I have claimed above that the label "Tjunction" was rather ambiguous. Now what about "Tcase"? I have tried to read the Intel document. It does define "Tcase" as the temperature on the center of the outside surface of the heatspreader. But this section of the Intel document is giving design specs to the thermal design engineers. It does not say that Intel or anyone else has actually placed any thermal sensors on the outside surface of the heatspreader. The thermal design engineers have various goals: to reduce costs & to hold down noise levels., etc. Intel is telling the thermal design engineers, "OK, you can cut back on the thermal design of your computer, so long as you can remove 130 Watts from the cpu without letting the temperature at the center of the outside surface of the cpu heatspreader go above 65 deg C". [I am just using some Intel numbers for a QX6700 cpu as an example. There are other numbers for other power levels & other cpus.] The thermal design engineer is expected to make calculations or tests to ensure his design can meet Intel's specs. If those specs are met, then Intel warrants, based on its own thermal calculations inside the cpu, that the cpu will function for its design life. I have found no claim by Intel that there is a thermal sensor for "Tcase".
Intel's own TAT software does not report such a result, at least when I run it.
*********************
There is at least one more "temperature" we must deal with. That is a temperature that is somehow picked up by the motherboard hardware & then reported in the BIOS & often by proprietary motherboard windows software & by other software such as Fanspeed. I have two ASUS P5B motherboards. They use ASUS software & call this temperature just "CPU temperature". The terms "Tcase" or "Tjunction" are not used by ASUS.
One of my ASUS motherboards has a Core 2 Duo installed. At this time it is reporting "CPU" temperature = 64 deg C. But TAT reports that the two core temperatures are Core 0 = 54 deg C and Core 1 = 55 deg C. Now we can say that this ASUS "CPU" temperature is NOT "Tcase" because the core sensors are in the silicon layer of the cpu where the heat is generated. The Tcase temperature must be cooler than the silicon temperature because the heat is being removed through the heatspreader & heat must flow from a higher temperature in the silicon layer to a lower temperature in the heatspeader or "case".
So then what is the "CPU" temperature? Motherboard makers have been planting thermal sensors in & under & around cpu sockets for many years before cpus contained any internal sensors. These MB sensors have usually been quite unreliable. I claim the MB makers are still at it, giving us unreliable temperature reports. What is the MB "CPU" temperature? It is garbage out. Not related to "Tcase" or "Tjunction". Not to be trusted for anything.
But I would be interested to hear about cpu temperature reports by other motherboards, as compared to TAT for example, and any reports by MB makers as to just what it is they are reporting to us as "CPU" temperature.
BOTTOM LINE
It is "Tcase", the temperature at the outside center of the heatspreader, that must be controlled to ensure that the cpu will work for its design life, but there are no known sensors that provide this information. It is only obtainable by the thermal design engineer based on detailed heat transfer calculations or measurements. But I think there is one thing we do know about "Tcase". It must be lower than the highest temperature in the silicon layer of the cpu because heat must flow "downhill" from the silicon to the heatspreader to get out. I think the thermal sensors that are placed in the silicon layer by Intel & reported as "Core temp" by TAT are near the high temperature regions in the silicon. So it looks safe to me to assume that if all Core temperatures are within the specs given by Intel for Tcase then the cpu is safe.
In other words:
If we observe from TAT reports that Tcore<TcaseSPEC
and we know that Tcase<Tcore because heat flows downhill,
Then we are safe in concluding that Tcase < TcaseSPEC, so Tcase is within specs.
All the above is purely opinion of course. No warranty is expressed or implied. Read at your own risk.
Walt