Across all our server systems based on Intel Icelake-SP, we are observing BMC SEL log entries indicating that the PVCCSA rail (controlled by Loop B of the MP2975) briefly rises to approximately 1.79 V and then immediately returns to the normal operating range of approx. 1.0V.
These events:
-
Occur randomly and are not easily reproducible.
-
Appear approximately once in every 4–5 power cycles (power ON → operational use → power OFF).
-
Do not result in system shutdown although 1.79V is well above the OVP limit programmed.
-
Do not show any fault assertion in the GUI or in the controller status registers.
The BMC reads the voltage over PMBUS from the controller.
We would appreciate your guidance on the following:
-
Is it possible for short-duration voltage spikes to reach such levels without triggering OVP protection, fault logging, or shutdown in the controller?
-
What board-level checks or measurements would you recommend to validate whether this is a genuine transient event? We are not able to capture any waveform as the occurrence is highly random.
-
Are there specific MP2975 registers or telemetry mechanisms we should tweak or change to take care of this?
Appreciate any leads for this.
Regards,
Ashbin