Ryzen idle voltages from different sensors & low/typical current BIOS setting

I’ve been wondering about the differences between the reported idle voltages and how they are affected by the Low Current Idle vs Typical Current Idle setting in the bios. Some background: recently I got a new power supply after using my ryzen pc with an old one that did not support modern C6 states on cpus, so I had to set it to Typical Current Idle in the bios as it would often freeze up at idle. I got a new psu to replace it and thought I would be able to set Low Current Idle in bios, but still got idle crashes, so had to switch it back to Typical. I booted into windows (I use linux primarily but have a second drive with windows on it, I can get more sensor readouts there). I’m wondering what the different core voltage sensors mean in Hwinfo64, and knowing what would indicate working C6 states. The psu I have now is an evga 650gq, which should support it. I am running an r5 1400 on an Asus x370 prime pro.

Under typical current idle:
-VID for all cores would stay at 0.9v idle.
-SVI2 TFN would stay at 0.85 at idle.

Under low current idle:
-VID for all cores stays at 0.4v idle, but sometimes fluctuates around or going all to 0.9 per core even if SVI2 stays at 0.85v.
-SVI2 TFN stays at 0.85 at idle, same as before.

There is also the VDDCR CPU sensor which I do not know anything about. If anybody else has investigated this or has any other observations that would be cool. From what I understand, VID is the voltage the cpu requests/thinks it needs for a given load, and SVI2 is the voltage readout directly on the cpu, which is most accurate. I am wondering if the behavior I am seeing is not normal, I would have expected with low current idle for both VID and SVI2 to drop to 0.4 when dropping into a lower power state.

Here is a screenshot of Hwinfo: https://i.imgur.com/UCEkkuM.png

I’ve experienced this bug myself on my AMD based Unraid server.

You may find this to be some interesting reading:

https://bugzilla.kernel.org/show_bug.cgi?id=196683

Reports that I’ve read is that AMD has released bios updates to address this issue to varying degrees of success based on what I’ve heard. Official AMD statements seem to blame the PSU but user reports seem to indicate that the issue is something else.

For me I’ve disabled C-States and its been solid, although interesting to note, that if I load windows on the same hardware I never saw this issue crop up.

VDDCR CPU is what the VRM is outputting, and SVI2 TFN is what the CPU is seeing after vdroop. The SVI2 TFN sensor won’t read when the CPU is in a sleep state and only getting 0.4V, so the lowest you’ll see on that is like 0.8V. You should be able to see the sleep state voltage if you look at VID or VDDCR CPU.

@Klingon00 That’s the same case I experience it. In windows, from what I can tell it seems fine, although I spend much less time there. In linux, eventually I saw an idle freeze although only after a few hours. To fix it I only have to set power supply control to Typical in bios, leaving C6 states enabled doesn’t hurt in my case. One thing to note, however, is when in linux you can use a program called Zenstates to see which C6 states are enabled on your processor. With “Typical” set, Core C6 state is enabled and Package C6 is disabled. With “Low” set both C6 states show as enabled with Zenstates. I am running a recent bios update from last december, though from what I can remember earlier bios releases were not much different. Maybe there’s some very minor in silicon problem or something, will be interesting to see if zen2 has this issue as well.

@bardacuda That is very helpful to know and makes a lot of sense that it wouldn’t report when idle. VID and VDDCR both tend to be in range of each other. Thanks for the input.