Tjmax values are only documented on mobile cpus, not on desktops/server cpus. Mobile cpus do not have an IHS (which is why intel documents tjmax on them) so it would seem unlikely for all mobile cpus then to have the same tjmax as desktops, and intel has responded on their now locked forums "no" to using mobile tjmax for desktops. For example, 45nm mobiles have tjmax of 105 documented, so coretemp, everest, HWM use 105 for all 45nm desktop cpus, because prior to Realtemp, there was no better way of guessing at tjmax.
Then Realtemp introduced measuring the value. All of Realtemp tjmaxes are based on estimating die temps, by undervolting, underclocking at idle, where intel docs show there is no gradient across the die at idle (or less than 0.1C).
There have been multiple experiments showing there is minimum gradient across IHS (less than 1C) under those conditions as well. Die temps have been measured directly with IR, showing no measurable difference (less than 1C) between bare die and IHS temps under those testing conditions (idle, undervolted, underclocked). Hence die temps can be accurately measured within ~1C of error.
For example E6600, which all programs use tjmax of 85C, IR measurement by realtemp show exactly 85C at DTS=0, again suggesting IR measurements are accurate, if indeed tjmax is 85 on E6600.
On many core 2 duos that coretemp, everest, HWM use tjmax of 100C, IR measurements (for example on E6850) show 95C at DTS=0, so many 65nm that have been measured at 95, now have 95 tjmax (realtemp) instead of 100 tjmax based on 65nm mobile cpus (coretemp).
And of course the 45nm have been measured multiple times by multiple people and consistently read 95C, so 95 tjmax is used in realtemp.
The 5c difference on some 65nm, measured IR Tjmax 95 (Realtemp) vs using mobile cpus tjmax 100 (Coretemp, HWM, Everest), was not worth writing another program over. But the 10C difference on 45nm, 95 Tjmax by IR (Realtemp) versus 105 tjmax using 45nm mobile cpus tjmax, which led to ridiculous high temp reporting using 105tjmax , led to Realtemp.Edited by opt33 - 7/3/08 at 11:16am