Vcore...vs CPU Core Voltage (SVI2 TFN)

[buddywh]

I've read here and other places that the (SVI2 TFN) voltage reading is reported by the processor in the SVI2 telemetry stream and so it's 'the most accurate' indication of the voltage the core(s) are seeing. I also understand that the Vcore reading in the Motherboard section is the output of the VRM as reported by it's monitoring chip, which I have confirmed as I read it's output (on the coil leads) with a DMM as best I can.

So, with that understanding how is it SVI2 holds in very steady range while the Vcore reading is constantly dropping to 1V or less, often for a considerable number of polling cycles if the system is very idle. I'm confused how SVI2 could be varying from 1.3-1.36V, occasional drops lower, even during periods when Vcore is at 1V. I am using a polling period of 500mS, btw, and graphing this to see the voltages juxtaposed.

Can anybody provide clarity to this confusion? Thanks very much.

 

[Martin]

This really depends on how the Vcore is measured and where the voltage is sampled which is specific for each mainboard model.
SVI2 TFN value is most accurate as it's the straight value provided by the voltage regulator.

 

[buddywh]

Is it posssible the VRM is responding by lowering Vcore voltage to 1.0V or less when all cores are in a C6 deep sleep, but the SSVI2 telemetry is still reporting a higher voltage like 1.34V? Or should I also expect to see SVI2 voltage dropping to 1.0 V or less. I'm aware that reporting of CPU core status is limited when cores are in C6 deep sleep, just don't know if that's a part of what I'm seeing or not.

I have an Asus board with a Digi+ VRM, btw, a B550 TUF gaming.

 

[Martin]

How can the CPU voltage be read if all cores are in C6 (sleeping)? At least one of the cores would have to be active to do the read and show it.
Try to activate the "Snapshot CPU Polling" option to get more accurate values in low power states.

 

[buddywh]

How can the CPU voltage be read if all cores are in C6 (sleeping)? ....

That's exactly what I'm thinking! so maybe what I'm really seeing is the 'last value' before all are sleeping and yet the Vcore reading can drop to 1V or less since it's the monitor chip that's reporting it.

I'll try the snapshot polling. Thanks

 

[Martin]

I agree with your understanding That could be true if the Vcore sensor measured by SIO would be reported with some delay.

 

[buddywh]

I agree with your understanding That could be true if the Vcore sensor measured by SIO would be reported with some delay.

I feel like I understand things a bit better now... although, what is the SIO. A new acronym and would it be a problem if it's always a bit delayed with it's reporting.

I attached an image of the the SVI2 and Vcore voltage graphs when the system is relatively quiet, so you can see what I'm talking about.

 

[Martin]

SIO is Super-IO (also called LPC) chip responsible for various functions including sensor monitoring via dedicated probes connected to it. In your case it's the Nuvoton NCT6798D.