"But the same motherboard should be consistent."
Classic argument, but not really accurate. The in socket thermistor is trying to read the CPU core temp through quite a few layers of ceramic. Different CPU's may vary slightly in how they fit in the socket, making them closer or farther away from the actual sensor. Another problem is that the socket is affected by airflow in the case. For example, I use a Compunurse to read the temp at the core's surface on my 1.2ghz @ 1.4ghz (Tbird). My motherboards sensor (Iwill KK266-R) consistently was around 5C lower than the reading on the core's surface. At the time, I had a side mounted 120mm (68CFM) fan blowing on the mobo, right below the socket. I recently switched to the 130CFM YS Tech 120mm, and now my temps are almost 10c lower from the motherboard sensor. Also, HSF's that blow down on the CPU can have the same affect. The Athlon ORBS are a classic example of this. They would always perform very well in tests, but this was mostly due to the way they flowed air directly into the CPU socket and cooled the thermistor. I have had endless conversations with Mikewarrior about this. There are just too many variables to take into account. The only real way to test accurately is with an internal thermal diode, which the Athlon's now have in the Palomino cores...