The CPU core clock drops during gameplay despite a stable overclock setting.
The CPU core clock drops during gameplay despite a stable overclock setting.
I've noticed recurring problems where upon starting games (AC Odyssey and League of Legends), my core clock decreases from 5000 MHz to 4700 MHz. This reduction stays until I restart the system. HWinfo indicates the core power limit has been surpassed, yet this alert didn't appear during stability tests in Cinebench r23 and Prime95. It seems these benchmarking programs might be more demanding than playing League of Legends.
My overclock configuration is as follows:
CPU ratio: 50
Avx offset: -1
Ring ratio: 47
CPU core voltage mode: override mode
CPU core voltage: 1.35
LLC mode 3
PL1: 4096
PL2: 4096
My Cinebench r23 results are around 9800, and the overclock remained stable for three hours of Prime95 small FFT avx off. Core temperatures sit between 36-40°C idle and 75-80°C under load.
I would appreciate any guidance on resolving this issue. Thanks.
Well, the only thing restricted is Vcore, so I’d start there, though I’m sure 1.35 should still be sufficient. Also, is it just a coincidence that the CPU downclocks to the same clock speeds as your ring bus? You might want to try adjusting the ring bus to check if the CPU will downclock accordingly.
The issue isn't about the drop itself but rather the concern with the drop. The specifications indicate different core counts for various frequencies: 4.9GHz uses 1,2,3 cores, 4.7GHz uses 4,5 and 4.6GHz uses 6,7,8 cores. However, it appears you've adjusted the CPU clock speed to 5.0GHz without proper support, causing the CPU to resist operation. Ring ratio should match the CPU's actual ratio—set it to 50 if maintaining the 5.0GHz configuration.
You are using MSI Dragon Center or a comparable MSI control application? Certain iterations of this software contained an issue where it erroneously adjusted the turbo power thresholds. If you have this installed, remove it. Another potential concern is that the maximum capacity stored in the PL1 and PL2 power limit registers is 4095.875 W. Writing 4096W to this register would cause an overflow, resulting in a reading of 0 Watts. To prevent this, ensure the PL1 or PL2 power limit registers are never set above 4095W in the BIOS.
Are you running the most recent BIOS version? Some of these issues have already been resolved.
Consider trying ThrottleStop 9.4.6.
https://www.techpowerup.com/download/tec...ottlestop/
Share screenshots of the main interface, TPL, and FIVR windows, especially when your system is constrained to the 47 multiplier. You may also open the Limit Reasons window to monitor any throttling indicators under the CORE column. After testing, fully close HWiNFO.
Display some screenshots and activate the ThrottleStop Log File feature if you continue experiencing difficulties. Once testing is complete, exit ThrottleStop so it can finalize the log file. The file will be located in your ThrottleStop / Logs directory. You can paste the information here.
Keep the ring ratio at 47. Very few people ever adjust the core and ring speed simultaneously. This approach demands excessive voltage for reliable results, especially with many Intel processors. It mainly generates excessive heat without significant performance improvement.
In the BIOS, it is advisable to turn on Ring Down Bin. This setting maintains the CPU ring/cache at 300 MHz below the CPU core frequency, helping avoid instability. If the core requires throttling, the ring will also reduce accordingly.
This power management issue appears to occur in certain games when the Nvidia GPU is active. That’s why applications like Cinebench R23 run smoothly without problems, while some games may throttle and slow down despite not using the GPU.