Showing results for 
Search instead for 
Did you mean: 


Journeyman III

3900x failing cinebench out of the box

Hi there! 

I've recently switched to AMD and have set up a rig with a 3900x and Asus ROG Strix x570 board. I'm having the hardest time getting this thing to run stable! I'm wondering if anyone here might be able to help me. I've watched a ton of tutorials but I am very much an overclocking noob. 

So out of the box, with only DOCP enabled to allow for 3600mhz on my ram, my CPU is reaching 1.45v (in CPUz) and attempting 4.6MHz clock speeds. This apparently isn't stable because it fails any Cinebench R20 test I run on it (gets an error). I've used Ryzen Master to clock all cores to 4.3MHz and the voltage to 1.35v per recommendations on forums and videos, and I am at a stable build. However CPUz still reports 1.45v average in a Cinebench run.

So I went into the bios and set the clock manually to 1.35, which fixes that, lowers the temps significantly and seems stable, but ANYTHING over 4.3MHz caused my computer to crash (hard crash, not Cinebench error). I feel stuck.

Is this a normal issues? That the out-of-the-box settings are so unstable that it can't even complete one Cinebench test? I've seen that I should set the VDCCR voltage in the bios to 'offset' -0.01v but that also causes system instability as it seems to still want to boost to 4.6MHz+. I feel like I'm missing something here. This is with PBO on 'auto' in the bios.

Anyway, I'm sure this question gets asked a thousand times, but I couldn't make sense of the other threads. If anyone is willing to help me out that would be awesome!


1 Reply
Journeyman III

Re: 3900x failing cinebench out of the box

So to thicken the plot I've found that limiting the voltage of my CPU in bios to 1.35 while enabling PBO (in BIOS) and then NOT using Ryzen Master allows me to get +800 on my multicore Cinebench test at 7562. While with Ryzen Master enabled the build is either unstable or results in a much lower score. Interesting. I'm still wondering if there is any way to get more single core performance...