configure / compile kernel for latency-performance (tuned adm style)
I need the cpu frequency at 100% max and constantly available for latency/cpu sensitive workloads. Typically, i'd use tuned-adm and call it a day but im on openrc and dont plan on using systemd, which tuned-adm is based on. I tried setting _makemenuconfig=y and hoping to find a full premptive RT model but nope. Any suggestions? My only other option is to compile a https://wiki.linuxfoundation.org/realtime/preempt_rt_versions with the full premptive model, but im on the artix distro and the only available kernels are zen, linux stable, and most recently lqx.
Back to top |
|||||
Well first of all, Liquorix defaults to the performance governor, which leaves your CPU at 100%. This doesn't allow maximum boosting at all times though, that's handled in the CPU hardware with information from motherboard on VRM.
As for preempt RT, I don't recommend going down that route. It requires a lot of invasive kernel patches and it's incompatible with most out-of-tree kernel modules. A good example is the proprietary nvidia driver, you can look at this recent thread for more information: forums.developer.nvidia.com/t/nvidia-driver-bug-scheduling-while-atomic-for-preempt-rt-kernel-on-rtx3060-gpu/233400 Back to top |
|||||
I couldn't tell if you were already aware of this, but "lqx" is the Liquorix kernel: aur.archlinux.org/packages/linux-lqx
I think it likely that an RT kernel PKGBUILD for Arch already exists, too. Back to top |
|||||
All times are GMT - 8 Hours
|