CPU Power Reporting Deviation insane at 1000+% max and ~150% idle

brawnyryan

New Member
SPECS:

MSI B550 GAMING PLUS latest bios
RYZEN 5 5600x
XFX SWFT 6700xt
850w CORSAIR RM
32 GB CORSAIR VENGENCE 3200mz
WD BLUE SSD

So, I've been having issues with this system. Packages (wattage, power) on both cpu and gpu dipping to zero. MOSFET transistors overheating ? no. CPU or GPU thermal throttling? no. I originally had a 500 watt power supply and just replaced it. The 850w Gold Plus Corsair seems to have made things more stable but when running a stress test in adrenaline I will occasionally see the gpu as well as the cpu dip to 0 clock and 0 watts, this is all while gpu, cpu mosfet, all other temps are not even close to thermal throttling temp. It got a 11855 score on 3dmark, but 3dmark and all other benchmark software don't seem to agitate the system enough to get a complete dip in power.

Now here is the interesting part. I've been using hwmonitor and msi afterburner to troubleshoot this system. I installed hwinfo and am now seeing the cpu is over reporting its power consumption by a long shot. I wish I would have done this about 5 days ago it would have saved me around 40 hours of my life and literally 50 clean installs of windows but nevertheless. This would make sense as to why I've been having these issues. I am thinking the system is throttling power not due to thermals but over reporting of power consumption.

imgur image here

any suggestions?
 
What you are thinking can’t happen. The zeroing in clocks and power has nothin to do with Power Reporting Deviation.

First of all PRD only matters when the CPU is 100% loaded. In every other situation (<=99% load) this is inaccurate and you shouldn’t pay any attention to it.
PRD is just a report of a deviation in power of what the board is reporting back (feedback) to the CPU about its power consumption and what the CPU is actually consumes. The CPU is unaware of this.

At idle or middle loads (<=99%) this value can be anything from 0% to over 1000% and it doesn’t matter.

If you want to see by how much the board is lying to the CPU about its consumption run a cinebench R20/23 MT and observe PRD and CPU PPT during(only) the test run.
You can post both values here and we’ll see.

As for the 0-ing at clock/power I’ve seen this before in another discussion and I remember that @Martin (HWiNFO author) said that it’s the SMU that it’s responsible for this kind of behavior.

You can try and flash the board BIOS/UEFI at latest version. If you are already there then try older one. The meaning of this is to try different AMD AGESA microcode for the CPU (= different SMU version).
Sometimes same AGESA version but different UEFI version can contain different SMU.

You can quickly find the SMU version in ZenTimings app

Like this one:
1668357421549.png
 
Back
Top