The simple answer, is that the temperature reading, and the ACTUAL temperature are not the same. The important thing, is to stay within the "number" limits. For FX-6300, don't let the "number" go over 70. The actual core temp is likely higher by ~10-20C+, but this is irrelevant as that is taken into account for the "limits" that have been set.
Think about this for a moment...
On the FX-8XXX series, it's common to have CORE temps reading ~10C cooler than socket temps. The socket temp is measured under the CPU by an actual temp probe, the CORE temp is supposed to be INSIDE the CPU (there are parts of the chip dedicated to reading temp, but the output from them is all thrown together into an algorithm to come up with a number to "report"). There is no possible way, within the laws of thermodynamics within this context, for a place OUTSIDE of the heat-source to be hotter than the source itself, unless the source were emitting radiation outside of thermal bands that was being absorbed/converted to thermal by a near-by outside object (not likely).
The complex answer to this conundrum, is that the core temp reported is indeed just a "relative" number. In fact, I would venture to guess that the actual core temp reporting is off by approximately -10-20C+ across the entire range (this explains the sub-ambient reporting at idle, and how we are able to achieve dissipation characteristics we do to typical HSFs with such "low" deltas compared to intel chips. Think about that for a moment... how would we dissipate double the thermal energy across a lower delta, it doesn't make sense). The amount that it is "off" by, depends on whether it is an FX-8 or FX-4/6, as the number of inputs "for" the temp-reporting construct winds up effecting the output, based on the algorithm. I have no doubt that a ~60C reporting on an FX-8 chip is probably ~80C+ actual, and ~70C reporting on an FX-6/4 chip is probably ~80C+ actual as well. Actual core temps could be far higher than that still. 100C is boiling point for WATER. It's not a melting point for sand or anything. Look at "max temps" for lots of other semiconductors, they tend to be ~100C+/-10. This points to an obvious "offset" in place here with the reported AMD temps.
Edited by mdocod - 3/8/14 at 9:01am