Linux 5.16 Merges Fix For One Of The Intel Alder Lake Issues
Merged this Friday afternoon into the Linux 5.16 development kernel is fixing a performance issue affecting some Intel Alder Lake motherboards.
The fix merged a short time ago is the item previously covered within Linux ITMT Patch Fixes Intel "Alder Lake" Hybrid Handling For Some Systems.
As explained in that prior article, TurboBoost Max 3.0 / ITMT (Turbo Boost Max Technology) code within the kernel isn't being enabled for some systems, particularly if overclocking or even any memory XMP / optimal settings. The ASUS Z690 board I've been primarily using for the i9-12900K was affected as are numerous other boards. I've also heard reports of some motherboards running purely stock are even having this issue.
The issue this causes is both the P and E cores being advertised as having the same maximum performance potential with the firmware hard-coding the ACPI CPPC highest performance state for all of those cores.
The fix was merged as part of this week's power management fixes that is included for Sunday's Linux 5.16-rc3 release.
Additionally, still outstanding is the new x86 cluster scheduling code in Linux 5.16 hurting Alder Lake as previously reported on and bisected. I had heard they were looking at reverting or disabling that x86 code but as of writing haven't seen that happen yet for Linux 5.16 to further get the Alder Lake Linux performance into a more optimal state.
The fix merged a short time ago is the item previously covered within Linux ITMT Patch Fixes Intel "Alder Lake" Hybrid Handling For Some Systems.
As explained in that prior article, TurboBoost Max 3.0 / ITMT (Turbo Boost Max Technology) code within the kernel isn't being enabled for some systems, particularly if overclocking or even any memory XMP / optimal settings. The ASUS Z690 board I've been primarily using for the i9-12900K was affected as are numerous other boards. I've also heard reports of some motherboards running purely stock are even having this issue.
The issue this causes is both the P and E cores being advertised as having the same maximum performance potential with the firmware hard-coding the ACPI CPPC highest performance state for all of those cores.
The fix was merged as part of this week's power management fixes that is included for Sunday's Linux 5.16-rc3 release.
Additionally, still outstanding is the new x86 cluster scheduling code in Linux 5.16 hurting Alder Lake as previously reported on and bisected. I had heard they were looking at reverting or disabling that x86 code but as of writing haven't seen that happen yet for Linux 5.16 to further get the Alder Lake Linux performance into a more optimal state.
14 Comments