I've recently built a new workstation that's utilizing the Intel 12Gen/Alder Lake processor and have installed Fedora 36 Beta. While testing a number of different workloads I've noticed that it's scheduling processes that are maxing out the efficiency cores while the performance cores aren't being utilized or are at a lower percentage of utilization.
Current Kernel: 5.17.3-300.fc36.x86_64 #1 SMP PREEMPT Wed Apr 13 23:08:09 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux
CPU: 12900K
Hi, this sounds like a kernel bug, not a desktop issue. Have you tried reporting this to Linux kernel folks?
I'm going to go ahead and close this because this is well outside our expertise.
Metadata Update from @catanzaro: - Issue close_status updated to: Deferred to upstream - Issue status updated to: Closed (was: Open)
I believe this will be fixed with Linux 5.19, which introduces a fair bit of new code to support 12th Gen Intel CPUs. @jforbes keeps a closer eye on this stuff than I do, but we'll probably see this in the summer rebase of the kernel in Fedora.
Hi, this sounds like a kernel bug, not a desktop issue. Have you tried reporting this to Linux kernel folks? I'm going to go ahead and close this because this is well outside our expertise.
I assumed that there was a Fedora kernel team that patched the mainline kernel with specific patches/backports. Does Fedora just use the vanilla mainline Kernel? I've been on Ubuntu for quite some time and they massage the heck out of their Kernels :)
We keep pretty close to mainline and update to new versions as they are released. If you want massaged kernels, you should probably check out CentOS Stream or RHEL. :)
Cool! This is one of the reasons that I'm giving Fedora another look after abandoning it years ago. I was getting frustrated with out of date software and working too hard to stay current. I'm just getting a better feel of the project. Thanks for clarification!
Ours isn't heavily patched because we move so closely with mainline. We do follow an upstream first policy, but do sometimes backport fixes or new support for staged upstream trees. Unfortunately in this case, the HFI code is a bit much to carry, but it was merged with 5.18, so at least the first iteration will be available soon.
Login to comment on this ticket.