That shows that there is definitely an issue with inadequate heat removal under load.
My laptop, an i7-9750H, 6 core 12 thread, 16 GB RAM, when running
“sudo stress --cpu 12 --timeout 1200” (20 minutes)
never exceeded 85C at the beginning and after 3 seconds was stable at 65-70C at 3200 MHz clock. gkrellm showed 99% cpu load.
Within 30 seconds after the end temps were 44-46C, 1-2% load, and 800 Mhz clock.
While I realize there may be a significant quality difference in the hardware design between your system and mine, it seems the only real issue you have is inadequate cooling under load. Your method of controlling the temp while using dnf seems to work .
Interestingly enough, it seems “stress” does not adequately do a full stress on the cpu. I run boinc using the rosetta project and at ~17% load processing 2 work units (2 cores) the CPU freq goes to 4200 MHz and temps go to 70-72C, which seems a heavier processing and thermal load than that provided by stress with 100% cpu load on 12 cores. Thus it seems reasonable that dnf may stress the system more than the “stress” app actually does.
Thank you @computersavvy for testing and sharing the result.
I run an i7-10750H, 6 core 12 threads and 16 GB RAM. According to intel the ‘average temperature rate’ is 50-65° celsius. While on power supply my notebook’s temperature is mostly 50-55 ° celsius, while on battery it is about 45° celsius.