Selecting “performance” in two desktop menus does not establish identical power-management behavior. That distinction matters in the Fedora 44 results on one Panther Lake laptop.
What the test establishes
Michael Larabel’s August 11 investigation concerns a Framework Laptop 13 Pro with Intel Core Ultra X9 388H. Updating Thermald alone did not close Fedora’s performance gap. Replacing tuned-ppd with power-profiles-daemon made the tested workloads substantially more competitive. These are Phoronix measurements, not PeopleAreGeek tests.
The conclusion also reports better performance per watt and explicitly limits knowledge of other laptops. It does not isolate a particular EPP register as the proven root cause. Nor does it show that every Fedora system needs the same replacement.
A shared interface can hide different profiles
Fedora’s Tuned change documentation explains that tuned-ppd translates the familiar desktop API to Tuned. The three visible modes can map to configurable Tuned profiles through /etc/tuned/ppd.conf. Where Tuned is installed, tuned-adm active reports its active profile. An installed-package list alone does not establish which provider currently handles the interface.
Our diagram follows the selection through the translation layer to effective settings. Keep that chain in a test record alongside the firmware and kernel. Otherwise, “same performance mode” can conceal the variable actually being compared.
Faster does not automatically mean more energy
Consider a fictional fixed job: 100 W for 60 seconds consumes 6,000 joules; 120 W for 50 seconds also consumes 6,000 joules. The second run draws more power but finishes sooner. This arithmetic is not a reconstruction of the review’s measurements.
For a useful local comparison, hold workload, mains/battery state, cooling conditions and background activity constant. Record elapsed time and energy over the same completed job, and repeat after temperatures settle. A profile name is a starting configuration, not a measured result.
September 8: attribute measurements to their author and one laptop; remove an unproven EPP root cause and distinguish installed packages from active profiles.