Missing sensors, wrong reading

Chike

Active Member
Hi,

First time poster here.
Not sure if HWiNFO or BIOS bug
I have noticed that when I change BCLK frequency there are 2 sensors missing and one show wrong reading.
On the right BCLK is 100 and everything seems fine, on the left BCLK set to 101.23.
IA Cores Power and GT Cores Power are missing and CPU Package power is far off, by the temps it is easy to tell it should be higher than those on the right side.
Only BCLK and vcore were changed in BIOS.
Motherboard: MSI Z87-G43
2cwpqp3.png
 
Well, this is interesting.. CPU Power Measurement is reported by the CPU and it's possible that when BCLK changes its internal estimation algorithms don't work properly. I have seen issues with this in the past.
Please attach the HWiNFO Debug File so I can check this in more details.
 
Another thing I've noticed is that it takes some time for the iGPU VID to show, some times it is only few seconds other times it can take a minute. In the attached debug file it did not show at all, seems like enabling debug make it worse.

EDIT: I must add that I have the same issue with CPUID HWMonitor, IA Cores and GT disappear from power, and Package show the same values as HWiNFO.
 

Attachments

  • HWiNFO64.DBG.zip
    43.8 KB · Views: 1
Thanks for the debug file. It confirms my assumptions - when you change the BCLK it seems the internal power controller logic in CPU has problems calculating the power consumption.
I'm afraid this is a bug in CPU and the only way how to fix this would be a BIOS update that would contain a new CPU Microcode update that would fix it. But I'm not sure if Intel ever attempted to fix this bug. Somewhere I've seen stated that if you overclock the CPU, the power consumption data might not be accurate.
Regarding the iGPU VID: I have seen this too and it seems to be an issue in the CPU as well - sometimes it doesn't provide VID and just returns 0.
 
Thank you.
I see this is an issue across all monitoring software I've tested. I guess I can live without the power sensor(s) temps should be sufficient.
 
I think this is rather influenced by SVID support. Many overclockers disable SVID (is this maybe your case too?) and it seems that in that case the internal CPU power measurements can't work properly.
 
Martin said:
I think this is rather influenced by SVID support. Many overclockers disable SVID (is this maybe your case too?) and it seems that in that case the internal CPU power measurements can't work properly.
Yes, that is what it is, I have just confirmed forcing SVID communication enabled and the readings are fine.
Thank you
 
Back
Top