Throttlestop/Undervolting broken by MacOS VoltageShift (possibly)

[gonkulator]

Hi,

I have a triple-booted system with an intel i7 8565u.

I've been undervolting in windows with throttlestop and Linux with intel-undervolt for a long time and it's been really helpful for power consumption and temperature. I tried adding VoltageShift in MacOS and suddenly undervolting seems to be broken everywhere. Throttlestop still shows the offset is being applied but it's not reflected in hwinfo64.

Is there a way I can fix this? Also I have msr-tools on Linux to try and see for myself what's happening but I'm struggling to find out the correct hex codes to get any meaningful info

Thanks

 

[unclewebb]

Post a screenshot of the FIVR window. Does the top middle of the FIVR window show Locked?

Does the ThrottleStop FIVR monitoring table show that the undervolt is being applied?

 

[gonkulator]

Spoiler

No locked symbol, the bios version I am on enables undervolting. It's been working really well up until i tried to add voltageshift to macos, now my temps seem to have suddenly shot up and battery life dropped. I can see the offsets seem to be applied correctly but struggling to confirm outside throttlestop. Not sure if it's an issue unrelated to undervolt, but heat and battery issues seem very similar to pre-undervolt levels

 

[unclewebb]

The FIVR monitoring table confirms that your undervolt is working correctly. After using ThrottleStop to make any voltage changes, you always need to restart HWiNFO before it updates the offset voltages that it reports. This seems to confuse a lot of users.

Always trust the FIVR monitoring table. It updates continuously in real time. HWINFO does not.

 

[gonkulator]

Okay thanks, can I confirm by reading the MSR (0x150 or other) values that undervolt is working?

It does seem to work as you said in windows after restarting throttlestop and hwinfo manually so I think it is working correctly.

I just also want to check in Linux and Windows the same way. The issue might be solely related to that and I would like to compare the MSR values to confirm. Linux returns 0 for all values of 0x150, but I'm not sure if that means anything.

 

[unclewebb]

after restarting throttlestop

No need to do that. Only HWINFO needs to be restarted. As soon as you press the Apply button in the FIVR window, any voltage changes will show up immediately in the FIVR monitoring table.

How MSR 0x150 works is a big secret. Intel does not publicly document this register. Do some Google searching. Maybe someone can tell you some more info.

 

[gonkulator]

Okay, I've tested reading/writing to MSR and confirmed the offsets are being applied both when manually writing and when using intel-undervolt on Linux. I think the issue must be unrelated to undervolt. Wondering if it's just failing thermal paste or something, thanks for the help anyway!