To finally put this question to bed, are the core temps for AMD CPU's calculated based on the CPU Temp, or are they actually coming from a sensor inside the CPU? Some sources claim there's just one CPU temp and that the cores are calculated, some claim the "CPU Sensor" is just a motherboard reading near the CPU, and that the core temps are actually the sensor readings. Which is correct? Also I found something from an AMD rep that said "The 62C limit is for the CPU Temp, not the cores". Essentially.
That sort of answers the question, but he still says "make sure core temps stay below 62C" Pretty much everything I've found is that the only sensor is the TMPIN1 sensor, or the CPU one. Cores are just estimates based on it.
The sensors are on-die from Athlon 64 onwards, which means they are located on the CPU itself, but I'm not really sure how many there are. Btw, from what I've read the core readings are the actual temps, the CPU temp (Everest/AIDA64) has something to do with distance to TjMax. At least that's my understanding of it.
That's why I though too, but all evidence suggests they are in fact estimates... Even the developer of core temp only shows the CPU temp for these CPU's, because he knows there's only one sensor.
there is no other temp sensor on the phenom 2's other than what located under the heat spreader. the core temps are just a guess as the cores themselves are not equipped with sensors. afaik the tmpin0 or tmpin1 sensor, depending on what mobo you have, normally is the sensor located under the cpu socket. so even then the temps reported in bios are off by few degrees.
That's what Im saying, there's conflicting information everywhere. But it seems as though AMD tends to agree with the 1 thermal sensor (TMPIN0 or 1), and that the cores are merely estimates.
Source one says there's a temperature sensor in each core, which isn't true... Source two says there's one thermal sensor inside the CPU that gives a general reading, which is what me and Firestorm were agreeing on.
That's not true at all, at least one of you were saying there was no sensor at all on the CPU: I quoted that link to back up my earlier post saying the sensors were on-die, I thought that's what you were asking? And there's only one source, not two. I think you need to look at the link so you can read the whole thing.
Oh, I did not in fact read the whole thing. But still, the two statements seemed so contradictory I assumed they were from different sources.... I wasn't saying the sensor wasn't on die, that was Firestorm. I agree that the sensor is on die. It has to be just one... The cores are always the same or within 1 or 2 degrees of each other in some programs (the developer of coretemp talked with the developer of hwmonitor, and they told him the differences were from sensor access delay). Not to mention they go up almost instantaneously when load is applied, all exactly the same amount. That's not physically possible