Zen 2 CPU Core Voltage vs VR VOUT Voltage

TADP0LE

New Member
During idle, my 3900X (running at stock) CPU core voltage drops to around 1V, with occasional spikes to around 1.4-1.5V, which I believe is standard behaviour. However, the VR VOUT value never drops below 1.4V at idle. Under load both values sit around 1.25V. From what I understand, CPU core voltage and VOUT are supposed to show the same values. Is there an explanation for why VOUT never reports values as low as CPU core at idle, and is this normal? This is on an MSI X570 MEG ACE with the ABBA Agesa BIOS.
 
The "CPU Core Voltage (SVI2 TFN)" value should match the VR VOUT value.
While sampling the SVI2 value is relatively quick, the VR VOUT value requires more complex operations to read out. So it is my assumption that either your polling rate is not short enough to capture the occasional drops or the VRM doesn't provide data with sufficient accuracy.
 
The "CPU Core Voltage (SVI2 TFN)" value should match the VR VOUT value.
While sampling the SVI2 value is relatively quick, the VR VOUT value requires more complex operations to read out. So it is my assumption that either your polling rate is not short enough to capture the occasional drops or the VRM doesn't provide data with sufficient accuracy.
Thank you for the reply. I was using a 500ms sampling rate, I tried again using a 100ms sampling rate and found that there was still a discrepancy between the values as in my earlier testing. I've tested now using a locked 1.2V on the CPU core, and now the values are the same all the time. I find it strange that the values previously matched when under high load, but not at idle when the CPU Core Voltage (SVI2 TFN) value would drop well below the VOUT. In the event that the values are not equal, which value should I consider more accurate?
 
I believe the SVI2 TFN should be more accurate. This value is directly provided to CPU from VRM via a different interface and is crucial to be accurate because the CPU relies on it for its own telemetry and power management.
The VR VOUT is provided by the same VRM, but another interface which is not required to be as accurate as SVI2.
 
Alright, thanks. Am I correct in saying that the VR VOUT value is provided by the Nuvoton LPCIO chip? Is the discrepancy in reporting something I should worry about? Or is it fine to ignore?
 
No, VR VOUT is provided by the same VRM, but using a different interface, which is not deemed critical for accuracy or reliability.
 
Back
Top