Originally Posted by CyberDruid
The actual reported temps are nothing to do with the safe spec for operation.
That is what I gathered from my reading on Core Temp and C2D tempertature monitoring.
What have you found that says different?
This is the way I see it.
I agree that the reported core temperatures from coretemp are not comparable to Intel's safe operating spec of 62C, as coretemp reads from the core diodes, and Intel is referring to the hypothetical center of the IHS (Tcase).
What I was trying to point out is that I believe the Tcase temperature reported by PC Probe and SpeedFan is inacurate because they use the wrong Tjunction value of 85C for a Kentsfield when in fact it should be 100C. This is my reasoning.
Since the temperatures are calculated by reading a value of a MSR and subtracting it from the reported value of the Tjunction stored in another register, if the wrong value is used for Tjunction the reported temperatures will be incorrect.
From the how it works for coretemp:
This is how the program works:
Intel defines a certain Tjunction temperature for the processor. In the case of Yonah it is 85CÂ° or 100CÂ°. First of all the program reads from a Model Specific Register (or MSR), and detects the Tjunction temperature. A different MSR contains the temperature data, this data is represented as Delta in CÂ° between current temperature and Tjunction.
So the actual temperature is calculated like this 'Core Temp = Tjunction - Delta'
Coretemp reports the Tjunction for a Kentsfield as 100C, and the Conroe's is 85C (I think). So when core0 reads 45C in coretemp, it is calculated by 'Core0 temp = 100C - 55C'
I believe that PC Probe and SpeedFan "assume" a Tjunction of 85C for the Kentsfield's, not the correct 100C. So using the example above where coretemp reports core0 at 45C, SpeedFan would calculate it as 'Core0 temp = 85C - 55C' outputting an improper value of 30C for core0.
The same also applies for the Tcase (CPU/IHS) temperature, which is the proper value to compare to Intel's "safe" spec of 62C. Since coretemp does not show this temperature programs like SpeedFan are needed.
In the following I'm assuming that CoreTemp displays the proper core
temperatures and Tjunction for my computer. Sorry for all of the screenshots, but I think that they help. They are all taken under load on a hot and humid day, ambient temp around 26C.
This is what CoreTemp looks like:
Those core temperatures seem reasonable to me at 3.2GHz on a hot day.
This is what SpeedFan looks like under default configuration. I used 2 instances so I could show 2 configuration windows. I have uninstalled PC Probe, but I recall it reported the same temperature for Tcase (CPU).
Notice the reported core temperatures are much lower than those in coretemp. 15C lower (except core2, which is 16C lower, but this can be attributed to the difference in polling intervals between the two programs). Exactly the difference in Tjunction values between Kentsfield's (100C) and Conroe's (85C), this can't
be a coincidence.
So I accounted for the 15C difference, and now the core temperatures are the same for both programs.
But now the reported Tcase temperature seems way too low; only one degree hotter than the NB, that doesn't seem right. Once again I added 15C to the reported temperature and things started to make more sense.
Now the core temperatures read by Speedfan match those in CoreTemp and the Tcase temperature seems to make sense being a few degrees cooler than the coldest (least hot) core.
This is what the chart in speedfan looks like after adding 15C to the Tcase temperature. Red is the Tcase temperature, and the others are the cores.
If SpeedFans reported Tcase temperature is off by 15C, then so is PC Probe's.
My guess is the both default to a Tjunction of 85C and therefore calculate temperatures 15C too low.
This is just my situation, yours my be different. There has always been a lot of confusion over the difference between Tcase and Tjunction, and which temperature one is supposed to keep below Intel's safe spec. Now that there are different Tjunction values for different classes of CPU it only adds to the difficulty. I'm not sure I've got it, but I think I have a decent grasp of how this works for kentsfield's, please correct me if I am wrong.
The way I see it Tjunction is a constant of either 85C or 100C and Tcase is the temperature one wants to keep below 62C. So I my case I'm at 49C and happy with that, 15C lower (34C) is just too low to be logical. The core temperatures can go above 62C as long as Tcase remains below 62C for kentsfield's.
Sorry for the long post, what do you think?