- Joined
- Dec 27, 2008
Here is an inquiry about core temps vs. cpu temps and the answer from an AMD tech. Clarified some issues for me but not all. Here were my questions:
PROBLEM DESCRIPTION: I wish to know what is the relationship between temperature readings reported for the core vs. the cpu socket/die/socket area.
1.How close should these agree? When stress testing at full load my reported temps for cpu/die/socket area are 37 deg. C higher than reported core readings. At idle, they are about 17 deg. C higher than core. If anything, you would think core temps would be higher than cpu/die/socket temps. What's more, reported core temps at idle are well below ambient with air cooling. Something seems screwy to me here. I have used various software utilities that monitor hardware and they agree with each other and with reported temp in bios, which of course, is an idle temp.
2. Is core temp a real temp or simply an artificial construct used as an algorithm base for calculating cpu/die/socket temp?
3. Which should I go by, core temp reading or cpu/die/socket temp reading when evaluating safe temperature levels of the processor under load?
4. Finally, what is the thermal shutdown temp of this processor and is that a core temp or a cpu/die/socket temp?
Thanks
Here is AMD tech's reply:
1. There is a correlation between core temps and case (CPU thermal lid) temps; however, the core temperature will generally be less than the case temperature. The case temperature includes heat from capacitors and other elements whereas the core temps are simply the temperature of the cores. Since temperatures are reported through the BIOS, the temperatures you see are highly dependent upon whether or not the motherboard manufacturer implemented our guidelines properly or not. There have been incidents in the past where a motherboard manufacturer has chosen to implement temperature reading differently than what we recommend. In these cases, the reported temperatures were not as accurate as possible and may have been confusing.
2. The core temp is an actual temperature.
3. AMD recommends that the case temperature (cpu temp) be used when assessing shut-off temps, etc.; however, bear in mind that AMD cpu’s have automatic shut-offs when certain internal temperatures are reached.
4. The thermal shutdown temp is based upon the case temperature and is approximately 95C.
Two things were helpful to me from the tech's reply. One was to learn that you should expect to see the CPU temp to be higher than the core temp. I assumed the other way around and at idle it may well be. However, under load the capacitors, mosfets, etc add to the CPU environment temp. Makes sense.
The second thing that was helpful was to learn what is the thermal shutdown temp is and that it is based on the CPU temp, rather than the core temp. Of course, this is interdependent on how the motherboard mfg. implemented AMD recommendations and how well they calibrated the sensors. Having said that, His reply also suggests there is another thermal shutdown temp which is indigenous to the processor and I assume that is monitored by the thermal sensor embedded in the processor itself and correlates with core temp. He is rather coy about that temp.
The major unknown still remains how accurately the motherboard manufacturer calibrated the cpu socket environment temp sensor.
PROBLEM DESCRIPTION: I wish to know what is the relationship between temperature readings reported for the core vs. the cpu socket/die/socket area.
1.How close should these agree? When stress testing at full load my reported temps for cpu/die/socket area are 37 deg. C higher than reported core readings. At idle, they are about 17 deg. C higher than core. If anything, you would think core temps would be higher than cpu/die/socket temps. What's more, reported core temps at idle are well below ambient with air cooling. Something seems screwy to me here. I have used various software utilities that monitor hardware and they agree with each other and with reported temp in bios, which of course, is an idle temp.
2. Is core temp a real temp or simply an artificial construct used as an algorithm base for calculating cpu/die/socket temp?
3. Which should I go by, core temp reading or cpu/die/socket temp reading when evaluating safe temperature levels of the processor under load?
4. Finally, what is the thermal shutdown temp of this processor and is that a core temp or a cpu/die/socket temp?
Thanks
Here is AMD tech's reply:
1. There is a correlation between core temps and case (CPU thermal lid) temps; however, the core temperature will generally be less than the case temperature. The case temperature includes heat from capacitors and other elements whereas the core temps are simply the temperature of the cores. Since temperatures are reported through the BIOS, the temperatures you see are highly dependent upon whether or not the motherboard manufacturer implemented our guidelines properly or not. There have been incidents in the past where a motherboard manufacturer has chosen to implement temperature reading differently than what we recommend. In these cases, the reported temperatures were not as accurate as possible and may have been confusing.
2. The core temp is an actual temperature.
3. AMD recommends that the case temperature (cpu temp) be used when assessing shut-off temps, etc.; however, bear in mind that AMD cpu’s have automatic shut-offs when certain internal temperatures are reached.
4. The thermal shutdown temp is based upon the case temperature and is approximately 95C.
Two things were helpful to me from the tech's reply. One was to learn that you should expect to see the CPU temp to be higher than the core temp. I assumed the other way around and at idle it may well be. However, under load the capacitors, mosfets, etc add to the CPU environment temp. Makes sense.
The second thing that was helpful was to learn what is the thermal shutdown temp is and that it is based on the CPU temp, rather than the core temp. Of course, this is interdependent on how the motherboard mfg. implemented AMD recommendations and how well they calibrated the sensors. Having said that, His reply also suggests there is another thermal shutdown temp which is indigenous to the processor and I assume that is monitored by the thermal sensor embedded in the processor itself and correlates with core temp. He is rather coy about that temp.
The major unknown still remains how accurately the motherboard manufacturer calibrated the cpu socket environment temp sensor.
