Ryzen 7800X3D - - - GHz decreases in PBO settings during the game?
Ryzen 7800X3D - - - GHz decreases in PBO settings during the game?
Hello, I changed the PBO configurations. During the OCCT test at -30, the system would restart. I adjusted it to -25 and it operated smoothly for two hours without issues. Yet, the PBO auto settings indicate a GHz decrease in the negative curve I manually set. It seems they maintain a minimum of 4.7 GHz, whereas some drop to 3.9 in manual mode. Interestingly, Cinebench doesn’t show this issue—it stays at 4.8 GHz and drops to 4.5 in auto settings. This is helpful for testing but I’m puzzled about the GHz variation in the game. Same game, same environment, same quality.
The change in performance metrics you observed is likely due to different system conditions. The game maintains a stable score of 4.8 on 8 cores at certain points, but drops to 3.9 at -25, despite no crashes in OCCT or other software. This inconsistency may relate to load variations or environmental factors affecting the auto-lock setting.
The same performance level can maintain 4.8 on 8 cores at the same stage until -20, but it decreases to 3.9 at -25. At -25, insufficient energy (low voltage) prevents operation at higher frequencies. Negative CO is best suited for temperature regulation; if you manage adequate cooling under the maximum junction temperature for each CPU, you won’t have to compromise voltage too much. When core voltage is too low, power delivery increases and could hit built-in limits that limit frequency.
This might be a vdroop situation. Consider applying an LLC configuration at the third level. It varies from board to board, possibly affecting levels 1,2,3,4, or the overall setup. It should help counter vdroop and provide stability.
I understand your logic, but I have a question. If it can stay at 4.9 in synthetic tests like Cinebench and provide this voltage, why can't the game achieve this? If there is a voltage limit, shouldn't it be valid in Cinebench as well? I've examined the structure of people using PBO, and most of them can detect the 5050 even at -30°C. I tried load line calibration 4, but the same thing happened.
The game might not require full CPU usage or load similar to a benchmark test.