Hi Carren, > > Note: while this is true of AMD CPUs (k8temp is a PCI driver) this > > isn't how Intel CPUs work. The coretemp driver reads the temperature > > values from MSRs (model-specific registers) i.e. from the CPU directly. > > So does that make the coretemp values more accurate/reliable than the > thermal_zone temps? I stand corrected. Yes, I misunderstood how Intel CPUs report the digital temp sensor. But Carren, whether the CPU reports the data through PCI or through MSRs, it is just a digital temp sensor. (So, what does that mean for temp control? It just means another input.) In response to your other question -- why would your digital temp sensors be lower? I can't claim any knowledge about how the chip is actually engineered. But my SWAG (Scientifically Wildly in-Accurate Guess) would be that the digital sensors are measuring the temperature much closer to the execution core, while the analog thermal_zone sensors are measuring the whole-chip average temperature. Intel Core Duo chips had the very first releases of the Merom design, and probably are more prone to overheat in the "misc" circuitry. That is, the execution unit just got a huge overhaul and is incredibly efficient. So if the digital temp sensor is close to it, it gives Intel some nice benchmarks, but isn't the main suspect of overheating on this particular design. Or I could be completely wrong. David
