Framework Desktop Power Mode Tuning For Better Performance Or Power Efficiency
Continuing on with our Framework Desktop benchmarking powered by AMD Ryzen AI Max "Strix Halo", today we are looking at the performance and power impact of power mode tuning for this review sample powered by the AMD Ryzen AI Max+ 395 SoC. A wide variety of benchmarks were done across the power saver / balanced / performance power modes for looking at the impact on performance as well as thermals and power consumption.
👁 Framework Desktop with GNOME Power Modes
Earlier this month I provided a deep dive into the power and thermals for the Framework Desktop while this article is looking now at the impact of the different power modes for the Framework Desktop under Ubuntu 25.04 Linux. Via the GNOME desktop interface the power modes can be easily adjusted depending upon whether you are striving for greater power efficiency or wanting the best possible performance regardless of the power/thermal impact. Today's benchmarks will help quantify the difference.
👁 Framework Desktop with AMD Ryzen AI Max+ Linux power mode benchmarking
The Framework Desktop was running with the AMD Ryzen AI Max+ 395 with Radeon 8060S Graphics, 128GB of LPDDR5-8000 memory, 2TB WD_BLACK SN7100 NVMe SSD, and running Ubuntu 25.04 with an upgrade to the Linux 6.16 stable kernel and the latest Mesa 25.3 graphics drivers.
When switching to the power saver mode the amd-pstate-epp powersave governor was in use like the balanced mode while the Energy Performance Preference (EPP) shifted from "balance_performance" to "power". The ACPI Platform Profile also shifted to "low-power" in the power saver mode. Meanwhile going from the default balanced mode to the performance mode changed amd-pstate-epp to using the "performance" governor while setting the EPP value to "performance" and the ACPI platform profile to "performance".
When running a wide variety of benchmarks on this Strix Halo powered Framework Desktop the CPU power consumption, CPU temperature, memory temperature, system power consumption (wall power), and system temperature was monitored too with our benchmarking.