Atemu

Oracle ZFS is so obscure by now that it’s irrelevant.

As usual with Oracle products, it’s just a means to squeeze poorly led companies.

If you’re not going to post what I asked for, nobody can help you.

Or just generally df -h | grep tmpfs and look for any significant usage.

I don’t know what this tool is or how it gets its “memory” metric. If you want to continue to use it, please ascertain that these values correspond to RSS by cross checking with i.e. ps aux. RSS is the memory exclusively held by a given process which is typically what mean by the “memory usage” of any given process. Note however that this does not count anonymous pages of a process that are swapped or shared with other processes.

Going into my task manager (Resources), I can see my using is using roughly 18/32GB of RAM despite closing all apps.

This does not tell you (or us for that matter) anything without defining what “using” means here. My system is “using” 77% of RAM right now but 45% of memory is available for use because it’s cached.

Please post the output of free -h aswell as swapon.

Next, please post the contents of /proc/meminfo.

Do you use ZFS?

There cannot be differences between CoreCTRL and LACT; they use the exact same kernel interfaces.

Your issue is therefore also not connected to any of these GUIs but how the kernel applies your policies.

I’d recommend you try to reproduce the issue using just the raw kernel files and report the issue upstream: https://gitlab.freedesktop.org/drm/amd/-/issues/

Well then it sounds like it works just fine but your chosen value isn’t stable.

The originals remain untouched.

It is possible to override existing commands with aliases though. This is useful for setting flags by default. I have alias ls='ls --color' for instance such that whenever I run ls, it actually runs ls --color, providing colourful output.

Note that aliases are only a concept within your command line shell though. Any other program running ls internally won’t have the flag added and wouldn’t be able to use any of the other aliases either (not that it would know about them).

It’s very easy to program your own “proper” commands though on Linux. If you had some procedure where you execute multiple commands in some order with some arguments that may depend on the outputs of previous commands, you could write all that as a shell script, give it some custom name, put it in your $PATH and run it like any other command.

You could make aliases that are easier to remember for you.

If you e.g. had trouble remembering that mv does a rename, you could alias rename=mv. Ideally just put whatever you would have googled in “linux command to x” as the alias.

That’s the power of Linux; you can tweak everything to your preferences and needs.

Chances are that it doesn’t work there either. What actually does the OC is the kernel; the GUIs merely write the desired values into the correct files in /sys.

This is a configuration declaration abstraction issue. Systemd timers and services are more like primitives.

In NixOS, we have an abstraction that allows simple declaration of a service and timer that runs some script.

As an example, I use this to export my paperless for backup daily in a way that is safe (paperless itself cannot run during that time, guaranteed by systemd) and simple:

https://github.com/Atemu/nixos-config/blob/ca0d39eb98c62424208487f973573478268048b4/modules/paperless/module.nix#L59-L95

(Even without NixOS domain knowledge you should be able to follow what’s going on here.)

All that’s needed in order to cause a systemd timer to be created for this service is to declare the startAt = "daily"; at the bottom.

LACT. Though I don’t know if it can OC Nvidia, Nvidia support is quite new.