Stephen R. Besch |
06-01-2011 11:27 AM |
From a scientist's perspective, what you are asking is whether anyone has yet built an "empirical" model of CPU temperature vs Load. For the isolated CPU case, I'm quite sure that the chip manufacturers have done something like this and you should refer to their data sheets or talk to their engineers. The problem is, as has been adequately pointed out above, that when a CPU is mounted on a mother board and put into a case, everything changes. However, there is nothing stopping you from doing this yourself if you want to go to the trouble of collecting and analyzing enough data. Most of the issues have already been pointed out. You would use the CPU core temperature that is reported using standard OS tools after applying various loads to the CPU. However, you need to repeat this experiment under a range of environmental conditions: Specifically temperature and humidity (which affects heat capacity of air). Each individual Machine would need to be characterized individually for reasons stated already by others. Finally, there is the issue of "kinetics" - that is, how long does the CPU, given its thermal environment and heat capacity, take to reach a steady state temperature at each CPU load.
Once you have spent several weeks collecting the data for one or two machines, you could indeed build a mathematical model that would reasonably well represent the Load vs. Temperature data with ambient environmental conditions as input parameters. Perhaps the bigger question is why? Simply watching core temperatures (either by eye, or with a software monitor, and then making suitable adjustments would probably be all that is needed in almost all cases - unless you are writing a scholarly paper on the subject. In any case, Have fun!
|