How-to-undervolt-AMD-RYZEN-5800X3D-Guide-with-PBO2-Tuner VS corecycler

i underclocked mine, i used this guide.

You still got the option to undervolt the CPU (guide) and the GPU (video tutorial). Maybe the power spikes are too extreme in some cases.

I have two PC's that each have a 5800X3D in them. In my Ghost S1 Build I am running a Noctua L12 (Ghost Edition), With the CPU Undervolted via PBO it runs well. Average gaming temps are around 70 Degrees. Full load the CPU gets closer to 80-85ish degrees. I wouldn't recommend this setup for long sustained work loads as you will leave some performance on the table but otherwise it works fine. I have my fan ramp up to 100% when it gets past 75 degrees. The fan / Cooler are relatively quiet. Very well built cooler.

I ended up using this guide: https://github.com/PrimeO7/How-to-undervolt-AMD-RYZEN-5800X3D-Guide-with-PBO2-Tuner/blob/main/README.md

getting lucky with the binning and undervolting as much as possible (+1 after last stable known value) (tested in multiple days with corecycler using Prime95 and y-cruncher)

Read the developer's statements from GitHub:

What I learned coming over from Intel is that in addition to stressing full loads on the CPU and memory systems for stability testing is that you also need to test bursty light loads as well. Corecycler uses Prime95 (default), Ycruncher, and Aida (you need to download) to test single core stability. I use OCCT & Prime95 to get a rough idea for a stable all core load, then use Corecycler with Ycruncher to test light loads.

Here's a template version of my Excel worksheet I used during my process. It's critical to tracking all your testing. https://we.tl/t-UHg9gnTNaGThe key is to ONLY test 1 core at a time with CoreCycler (https://github.com/sp00n/corecycler) until it has been validated for 12 hours. If it fails the Y-CRUNCHER test (y-cruncher error, hard reboot/freeze, BSOD) under the parameters shown in that sheet, you need to lower the CO setting on that core by 1-2 and test again until it passes for 12 hours.You want to also make sure to keep an eye on Windows Event Viewer to see if the core you are testing is producing any WHEA errors as well. That will alsocount as a FAILED validation run, and you'll need to lower it 1-2 and restart testing in Y-Cruncher again.Once it's passed for 12 hours, mark that core/line green on the sheet, 0 it out again in BIOS, and move onto the next one.Here's my CoreCycler config file. It's already setup to start at Core 0, 720m, y-cruncher, 2 threads, with all the correct testsselected. https://we.tl/t-gK3fc40BR1

You can test it extensively with corecycler ( https://github.com/sp00n/corecycler) with y-cruncher set as the stress tester.

Not all cores are equal. You could try Curve optimiser per core, test with Corecycler. Someone made an automated tool for it. Link