HWiNFO causes high Core VID values on some cores

[Dalai]

Hi there,

long time HWiNFO user here, but first time to post in the forum (obviously). Before reporting this as a bug I want to investigate further to check if I'm doing something wrong or missed anything.

Here goes. I wanted to use HWiNFO64 to monitor sensors on a server, but as long as this issue is there, I can't and won't. The issue: As soon as I start HWiNFO64 the Core #0 and Core #1 VID don't come down after HWiNFO is done initializing its stuff which is causing higher power draw and temperatures. If I restart CPU-Z while HWiNFO64 is running the Core #0/#1 VIDs come down again - until I restart HWiNFO64. The VIDs also don't come down when closing HWiNFO64, they stay high until CPU-Z is (re)started. I tried disabling as many sensor types as possible (see INI values below) to see if that changes anything - it doesn't.

Important notes:

• The system is idle in any case
• All ways to read the voltages (CPU-Z, HWiNFO64, IPMI) report (almost) the same values. So, it's not that one program reports something completely different than another program
• If you require more information please do not hesitate to ask

Some information about the hardware and OS:

• SuperMicro AS-2013S-C0R
• AMD EPYC 7261
• Windows Server 2012 R2 (Nov. 2014), freshly installed, without any updates so far (shouldn't matter for this issue)
• All drivers installed

Settings in HWiNFO64.ini, cut to those that are different from the default values

SWSMI=0
SMBus=0
AC97CodecID=0
GPUI2C=0
PCIdirect=0
UseHPET=0
AutoUpdate=0
GPUI2CNVAPI=0
BusClkPolling=0
SummaryOnly=1
WakeGPUs=0
iMEsupport=0
GPUI2Ccaching=0
DecimalSeparator=.
ThousandsSeparator=,
CsvSeparator=,
MinimizeGraphs=1
TextButtons=0
SensorRowShading=1
SensorsTopmost=0
CSMI_SAS_Support=1
DebugDirect=0
GPUI2Cforce=0
MinimalizeSensorsClose=0
KeepTheme=0
WakeGPUsExt=0
PollSleepingGPUs=0
ShowWelcomeAndProgress=1
EnablePchTherm=0
ReorderGPUs=0
NvmlSupport=0
PreferAMDADL=0
SMART=1
ATAStats=1
SupportCorsairAsetek=0
SensorInterval=5000
SmartScanRate=1
EcScanRate=1

Screenshots of CPU-Z and HWiNFO64.

CPU-Z before starting HWiNFO64
[attachment=3044]

CPU-Z after starting HWiNFO64
[attachment=3045]

HWiNFO64 before restarting CPU-Z
[attachment=3046][attachment=3047][attachment=3048]
Here you can see that Core #0/#1 VIDs are higher than all other Core VIDs.

Will post the remaining three screenshots in a following post.

Why is the Core VID not coming down when the system is idle, until I (re)start CPU-Z? Am I missing something?

Regards
Dalai

 

[Dalai]

Remaining screenshots.

HWiNFO64 after restarting CPU-Z
[attachment=3049][attachment=3050][attachment=3051]

 

[Martin]

Well, this is a very strange issue and very difficult to diagnose without having physical access to the machine and ability to debug run-time.
Please attach the HWiNFO Debug File, I will check if I see something suspicious there.

 

[Dalai]

Here you go. I hope the debug file contents are independent of Sensors/Summary only settings. The file "WithCPU-Z" was recorded while CPU-Z was running, the other one while CPU-Z was not active.

Regards
Dalai

 

[Martin]

Thanks. I've got one idea, but really not sure if this will bring anything..
Please try this build: www.hwinfo.com/beta/hwi64_589_3524.zip and let me know if you see any difference.
If not, please try to start HWiNFO without Sensors-only, so that only the main + summary screen will open. Then check when the issue happens - if right after starting HWiNFO without sensors, or after you open the sensors window.

 

[Dalai]

Please try this build: www.hwinfo.com/beta/hwi64_589_3524.zip and let me know if you see any difference.

Thanks. Looks good so far. Voltage stays low, also during HWiNFO's initialization. But I'll do some more tests to see if the persistent driver works just as well.

If not, please try to start HWiNFO without Sensors-only, so that only the main + summary screen will open. Then check when the issue happens - if right after starting HWiNFO without sensors, or after you open the sensors window.

Sorry if you had the impression that the issue only occurs if Sensors-only is enabled. In version 5.88 it doesn't matter how HWiNFO64 is started, Sensors-only, Summary-only or none of them, the issue is present in all cases. The voltage goes up as soon "Completing Mainboard Devices" can be read during initialization. It's just that the initial intention was to use HWiNFO64 with Sensors-only as a monitoring tool, so that's what I started with.

Regards
Dalai

 

[Martin]

Thanks for the feedback. Please let me know after more testing. It would be great if my assumption worked and fixed this issue

 

[Dalai]

I haven't found any problems related to the issue with the new beta version so far. Although at some point I had a weird issue with higher CPU usage caused by some driver (which might have been HWiNFO's driver), it never occurred again, so ignoring this for now.

It seems your assumption was correct and you fixed the issue. Just out of curiosity: Could you explain what the issue was, and how you fixed it?

Regards
Dalai

 

[Martin]

Thanks for the feedback.
Problem was that HWiNFO was checking certain parameters in the CPU using commands that are valid for the AM4 platform, but not for SP3. This caused some unexpected side-effects.

 

[Dalai]

Interesting. Thank you very much for your fast and precise help!

Regards
Dalai