AMD engineer Shyam Sundar S K reportedly submitted a three-patch AMD Overclocking Driver (AOD) series for Linux on October 10, 2026. The reported review sequence continued through October 21. In a report published October 23, Geektopia said the patches were still under review and had not been merged into the Linux kernel.
What the proposed AOD controls would do
AOD would connect Linux to Ryzen platform firmware through WMI, or Windows Management Instrumentation, and expose selected settings through sysfs, a Linux kernel interface for system attributes. The proposed controls cover three power and current limits: Package Power Tracking (PPT), Thermal Design Current (TDC), and Electrical Design Current (EDC). The interface would also provide supported ranges and step sizes for those limits.
The proposal also includes a platform-level switch for simultaneous multithreading (SMT) and a read-only memory-clock value. The SMT switch acts at the firmware level; Linux’s nosmt boot option instead tells the kernel not to use sibling threads. The reported initial controls do not include adjustments for individual-core clocks or voltage.
Why changing the limits would require a reboot
New PPT, TDC, and EDC values would be stored in firmware and take effect after the next reboot, rather than changing active limits immediately. The described attributes would expose supported ranges and step sizes, but not read back the currently active values.
How AOD differs from HSMP and debugfs
AOD and the Host System Management Port (HSMP) operate at different layers. AOD would write settings stored in platform firmware for the next boot. HSMP changes runtime limits in the system management unit (SMU); those changes take effect immediately and are lost at reboot. Technical accounts describe HSMP as an interface for EPYC servers.
The patch series would expose the proposed controls through sysfs, not a graphical AMD tuning app. It also describes raw WMI access through debugfs, a Linux filesystem used for debugging. That interface lacks a stable application binary interface (ABI), so it is not a dependable foundation for user-facing tuning tools.
The reported review timeline began with the October 10 submission. Mario Limonciello was reported to have supplied a Reviewed-by on October 12; Armin Wolf commented on all three patches on October 15; and subsystem maintainer Ilpo Järvinen commented on patches 1 and 2 on October 21.