A set of patches for the Linux kernel's Intel P-State driver aim to ensure Intel Xeon servers can hit their highest clock frequencies following changes in the server's performance profile. It turns out some special handling is needed to update the highest frequencies of a CPU after boot to ensure the performance profile is properly reflected.

The patches aim to deal with the performance profile being changed by Intel Xeon servers such as via the BMC or via Speed Select Technology to ensure that the highest frequency is properly propagated into the Linux kernel's Intel P-State CPU frequency scaling driver. The [1]patch series cover letter explains:

"Intel Xeon servers, which are capable of dynamic performance changes, are unable to achieve the highest frequency when the performance profile is changed.

The highest frequency at which a CPU can operate is not fixed and can vary after the system boots. These changes can be initiated by switching to different performance profiles using the Intel Speed Select Technology interface. Additionally, adjustments can be made remotely through a BMC (Baseboard Management Controller) interface. Administrators can select various performance profiles to align with specific performance requirements, as these choices will directly influence the total power consumption and cooling requirements.

Whenever an administrator switches to a different performance profile that alters the highest frequency, the hardware sends an interrupt and update the new highest frequency at which the system can operate. This interrupt can be enabled via the MSR_HWP_INTERRUPT register, and only if support is indicated by the CPUID[6].EAX[15] = 1.

To enable changes to the highest frequency, add a CPU features flag and enable the HWP (Hardware P-states) highest performance change interrupt when it is supported by the CPU."

The intel_pstate patch goes on to confirm that this could lead to the CPU reported maximum frequency not being properly reported via sysfs and that -- more importantly -- if the CPU's highest performance level is increased post-boot, the CPU may not reach the full expected performance.

From the patches and simply checking the presence of a particular bit, it's not immediately clear if this new HWP Highest perf change feature is just for upcoming Intel Xeon 6 servers or also affects existing Intel Xeon servers. In any event these patches are now out for review and could be mainlined as soon as next month's Linux v6.11 merge window.



[1] https://lore.kernel.org/linux-pm/20240618035122.438822-1-srinivas.pandruvada@linux.intel.com/