PBO2 Curve Optimizer: Inconsistent AF, more difficult then X299 Mesh OC

So… I’ve spend the past 3 days trying to stabilize an undervolt using PBO2 Curve Optimizer on 5800X and I’m getting nowhere fast.

MPV on Ubuntu 20.04.1 always segfaults with a libavcodec segfault (so technically an FFmpeg segfault) after 6 or 12 hours.

Hypothesis tried to use less negative value than before… Nope.

Some tutorials said to undervolt for a specific load and “keep using the system”. Did that, it crashed faster.

Lowered the CPU Boost Clock override from 200mhz to 50mhz. Still crashed on negative 12.

The pass condition is playing a 40min 4K 60p VP9 video on loop for 24+ hours. It has not succeeded so far.

My control is a 10920X at stock multipliers and that can easily do 20 hours with MPV and libavcodec. (so far)

At this point I think even disabling the Curve Optimizer will cause a crash. This is unbelievably inconsistent, and anyone recommending going straight for 200mhz boost clock override is WRONG. There can still be instability if you run something super sensitive like FFmpeg or something that uses libavcodec like MPV.

I’m literally waiting for my next crash on my 5800X system because it’s inevitable. Memory is 3600Mhz dual rank CL16, and I got rid of “miscompare” errors due to 1800 FCLK memory errors by increasing SOC to 1.14V.

To anyone planning to stability test Curve Optimizer, FFmpeg/libavcodec will find crashes any other standard Windows program won’t find.

This is oddly similar to the mood of the Jayztwocents video where he got nowhere with his Kingpin 3090:

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.