On Tue, Dec 5, 2023, 11:19 Vigneshwar S <[email protected]> wrote: > Hello everyone! > > I've read about the CPU performance scaling in Linux kernel documentation > <https://docs.kernel.org/admin-guide/pm/cpufreq.html> and tried > experimenting with multiple scaling governors. Based on the documentation, > the performance scaling governor > <https://docs.kernel.org/admin-guide/pm/cpufreq.html#performance> should > set the frequency to the maximum possible frequency. >
Correct > I've got two machines with different kernel versions: v3.10.0 (CentOS 7) > and v5.10.106 (Debian 11). When I set the 'scaling_governor' to > 'performance' in the CentOS machine, it is working as expected and the CPU > frequency is set to the maximum possible frequency statically. But when I > set the 'scaling_governor' to 'performance' in the Debian machine, that's > not the case and the CPU frequency is changing dynamically. > I think that is because two probable things: 1. Kernel does not correctly probe max freq of processor Or 2. Kernel detect idle situation, thus decide to lower the cpu freq for quite some time When needs arise, performance will put back cpu to max known freq Performance differ with other lies not on putting max freq, but how fast it.steps into max freq That is afaik > > I've read that 'intel_pstate' scaling driver would bypass the scaling > governor and set the CPU frequency based on its internal pstate > information. But in both CentOS and Debian machines, the 'scaling_driver' > is set to 'acpi-cpufreq' only. > > How can I solve this? Any help on this would be greatly appreciated. > > Thanks! > > Vigneshwar S > _______________________________________________ > Kernelnewbies mailing list > [email protected] > https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies >
_______________________________________________ Kernelnewbies mailing list [email protected] https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
