Core Max Vs. CPU Package Temps

[Captain Bad]

Is there supposed to be a significant disparity between these two temperatures?

At the time of me writing this thread, my Core Max temp is 43°C, however my CPU Package temp is idling at 52°C.

I've noticed in other threads that the CPU Package temp is the preferred one to be using, but couldn't help but wonder about the significant difference in temps or if that's abnormal.

 

[Martin]

Well, that's indeed a larger difference than expected.
What CPU do you have and what kind of workload are you running ?

 

[Captain Bad]

5820k and nothing outside of the extraordinary... right now I'm just browsing using Chrome and I've got a varying CPU usage of 1%-4%. I have seen the CPU Package temp spike up 69°C when gaming sometimes however. At the time of writing this, Core Max is sitting at a comfy 41°C and CPU Package is 51°C.

 

[Martin]

Have you maybe changed the Tj,max value in HWiNFO ? That would explain the difference, since Core Max values are relative to Tj,max but the CPU Package value is an absolute one.

 

[Captain Bad]

I have not... unless I did it accidentally. Is it possible to do something like that on accident?

 

[Martin]

No, that should not happen. But just to be sure:
- Please right-click on one of the Core sensors, select "Adjust Tj,max" and tell me what value you see there.
- Also please attach the HWiNFO Debug File

 

[Captain Bad]

No, that should not happen. But just to be sure:
- Please right-click on one of the Core sensors, select "Adjust Tj,max" and tell me what value you see there.
- Also please attach the HWiNFO Debug File

1) All "Adjust Tj,max" temperatures read 95°C
2) HWiNFO Debug File attached

 

[Martin]

That's correct, so the value reported is as well.
I don't know what might cause such a large discrepancy. There's also one difference between those values - the Core values are immediate, but the CPU Package is an average of the last 256 ms samples. But according to your description, the values are quite steady (not just single peaks of difference), so either all values are correct and there's indeed a larger difference across the CPU die (which is a bit unlikely), or there's a flaw in the temperature diodes inside the CPU. I think only Intel can provide a better answer.

 

[Captain Bad]

Looks like Intel's stumped too. The rep asked me to run the diagnostic tool, which I did, and came up with a PASS and no issues. If you can find the humor in it, he had no idea what was going on either with the CPU Package temp and mentioned the Core Max temperature is the one I guess they prefer? Maybe prefer's not the right word, but the one they're more familiar with. Guess the mystery of the CPU Package continues.

Thank you for the help.

 

[Martin]

I guess this was just the 1-st level support, so maybe if you push them a bit they will forward this issue to a higher level and might come with a better answer.

 

[parsec]

I'm seeing a similar thing with my i7-5820K, not that I think HWiNFO is a problem. As we say in the USA, don't shoot the messenger.  

For example, at idle with C States enabled, my Core temperatures are in the mid to upper 20s C. HWiNFO provides four other CPU temperatures on my board (ASRock X99 Extreme6, Nuvoton NCT6791D). CPU Package, 39C - 43C, PPO, 34C - 41C, CPU, 34c - 37C, and CPU (PECI) 34C - 37C. There may be more for all I know, hiding under names like AUXTIN1.

The Intel Extreme Tuning Utility shows one "CPU Temperature" that seems to be the same as the CPU (PECI) in HWiNFO, and the Core temperatures (which match HWiNFO Core temperatures.) During stress testing that CPU Temperature seems to simply be the maximum Core temperature at the time the data was read.

My Z97 and Z87 boards with various Haswell processors (but not Haswell-E of course) when used with HWiNFO also show Core, CPU and CPU (PECI) temperatures, but they are much closer if not identical to the highest Core temperature.

It looks like Intel has additional temperature sensor readings on Haswell-E processors, whose meanings were not part of the Intel support training.

The BIOS/UEFI of my board allows you to set the Tj Max temperature, which I set to 100C. The default value is 105C. I'll reset that setting to 105C and see what happens.

 

[TairikuOkami]

Sorry to bump into your thread, but since I have the same question.
My CPU shows 35C and CPU Package shows 85C. Which one is correct?
I mean, should I be concerned? I bought a cooler, because at idle it was 105C.
According to the FAQ, I should ignore core temp and focus on CPU package right?

 

[Martin]

Not easy to say, the correct one might be neither of those two on this AMD CPU.. Though the package temperature seems too high to be true.
Try to check values in idle and under load which of them seems to provide reasonable data.
Or better rely on external (socket) temperature if you system provides such (can't tell from the cut screenshot).

 

[Zone15]

Don't mean to bump this post (again), but I believe I am having a similar issue to the OP. The first thing I noticed was that with my old 2500K my CPU Package temperature and Core Max were almost always within a degree with each other. When I upgraded to a 5820K with the ASUS X99-A motherboard, I noticed that at idle the CPU Package temperature was 10-15 degrees higher than the core temperatures and was around 5C hotter during stress tests.

At first I did not think much of this 5C as I figured Haswell was just different. Then I noticed in Intel's own stress test program, the Intel Extreme Tuning Utility, the CPU Package temperature was reported at exactly 5C lower than what HWiNFO64 was reading. I then checked AIDA64 and noticed it was reading the same 5C higher temperature as HWiNFO64. Confused I checked another temperature monitoring software, this time CPUID's HWMonitor (which is used by OCCT) and it had the 5C lower CPU Package temperature like the Intel test. Is there a reason why HWiNFO and AIDA64 are showing the CPU Package being exactly 5C higher than Intel's own utility and HWMonitor? Which one is the true temperature?

 

[Martin]

It's possible that there are other components in the CPU that are hotter than CPU cores - in case of Haswell CPUs this is possible because of the Fully Integrated Voltage Regulator (FIVR). So since the CPU package temperature reports the average of highest temperatures in the CPU, it might be catching the FIVR.
Not sure which temperatures exactly report the Intel Extreme Tuning Utility and HWMonitor, whether it's really the same value.
Is the 5C difference always (idle and load too) ? Can you please attach a screenshot showing those values ?
Can you also try to do a "Reset Preferences" in HWiNFO to see whether the same difference will still be there ?

 

[Zone15]

Here is a screenshot of the temps from both idle and load.  It is not always exactly 5C but within a degree of 5C, and I believe that is just due to the temperature readings being taken at slightly different intervals.

Idle

Load

Resetting preferences didn't seem to change anything.

 

[Martin]

Looks like tools like HWMonitor and Intel XTU read a different CPU Package temperature value.
Unfortunately Intel doesn't provide enough details about these temperatures, so my assumption is that the value HWiNFO reads includes some additional package source that is hotter.

 

[Zone15]

Looks like tools like HWMonitor and Intel XTU read a different CPU Package temperature value.
Unfortunately Intel doesn't provide enough details about these temperatures, so my assumption is that the value HWiNFO reads includes some additional package source that is hotter.

You don't think it's possible that there is an offset somewhere?  I'm assuming HWiNFO isn't applying a +5C offset to the values but is it possible Intel applies a -5C offset to their temperature reading?  It just seems strange to me that even at idle or load the temperature difference is pretty much exactly 5C.  Even if you look at the difference in the "average" temperature over a long run, it is 5C.  None of the other temperature readings in HWiNFO match up to the Package temperature being reported in Intel XTU or HWMonitor.

 

[Martin]

HWiNFO certainly doesn't apply an offset. I have also noticed from those screenshots that the Min/Max values have a 5C difference between tools. So yes, it's really strange.
Moreover the Package temperature reported by XTU or HWMonitor seems to match the CPU (PECI) temperature in HWiNFO. This is the temperature passed to external hardware monitoring circuits.
The value used by HWiNFO is called Package temperature by Intel too, so from the results I assume the CPU Power Control Unit uses a different Tj,max to calculate it.
Perhaps a BIOS update (with new microcode) might change this...
Another theory would be that the CPU Package temperature reported by HWiNFO is the correct one and all other values including Core temperatures are -5C offset because the Tj,max is adjusted by CPU.

 

[Zone15]

Thanks for the help. I just recently updated to the latest BIOS but noticed this discrepancy in temperatures before that on the old BIOS as well. At this point I'm not sure what temperature to believe so I might just ignore package temperature all together and look at the core temps.