Improved settings on the 5900X outperform manual adjustments!
Improved settings on the 5900X outperform manual adjustments!
Seems like the clock stretching on your Asus board was causing issues. Turning it off and adjusting the voltage offset helped lower temperatures and improved the score to 22046. It looks like checking core vs effective clock during Cinebench made a difference, though. With the Asus x570 running only Agesa 1.1.0.0, I might hold off until the BIOS is more stable.
It's good to see you're making progress. AGESA 1.1.8.0 should address some issues with new features, so patience is key. MSI released a BIOS update yesterday that includes AGESA 1190, surpassing version 1180. After updating, the curve optimizer seems to be functioning properly. An odd observation is that AGESA1100 can now set a negative value of -35, but it won't allow values below -30. I followed this sequence: -10, -10, -10, -10, -10, -30, -30, -30, -30, -30, -30, -30, -30, -30, -30, -30, -30, -30, and -30, -30. This resulted in a lower score in CNr23 from 22400-22500 to 22200. Perhaps experimenting further with these settings would help. The RYZEN Clock calculator 2.0 is expected to support 5xxxx Ryzen profiles, which could simplify the stable overclocking process. They mentioned a 2.1 update next month that will enable per-core optimization automatically—hopefully this will let us fully utilize our processors as promised.
It seems there were some differences in how older and newer BIOS versions handled core settings. The previous version limited you to 30 counts, while the latest allows up to 30 but with different core allocation strategies. The reason for prioritizing certain cores over others depends on your workload—some setups benefit from balancing or optimizing specific cores. I see others using different approaches, so it might come down to testing what works best for your tasks.
I checked a beta version with Agesa 1190 on the x570 Asus board. It looks like the curve optimizer performs much better then, with no weird errors or unexpected reboots. This time I set voltage and LLC to auto, which helped notice the difference. For my CPU, I limited the four fastest cores to -10 and then further to -25, except for a few combinations that wouldn’t boot or caused crashes. The results were solid. I’m hoping they have an updated calculator for the 5000 series so I can try overclocking there too. Right now I’ll stick with XMP settings.