by hedora 8 hours ago
I do all my gaming and LLM inference on devuan. (That laptop should chew through 30-120B parameter LLMs, depending on how much RAM it has.)
I don't have any of the problems mentioned by you or sibling comments, full stop.
Up until about a year ago, audio was janky as hell, but then as part of the great de-Poettering, they switched from PulseAudio to PipeWire. I've had zero issues since then.
Copy paste works. Login works. X11 runs at native panel speed (144Hz) with bug-for-bug parity with windows 3d acceleration, but open source (AMD drivers). XScreenSaver works (and can lock the screen). I can't comment on any of the stuff you mention in the second paragraph. I assume it's a bunch of broken Wayland workarounds?
Anyway, instead of switching to Mac, just switch to a stable distribution. Devuan is Debian minus systemd, so essentially everything works out of the box. Even crap that requires systemd usually works, since they install stubs. LLMs like Claude will happily admin it once you tell it that which init system to target.
> the great de-Poettering
Well, systemd, including logind, which was incredibly unstable for me under manjaro (basic session management was flaky).
Devuan uses elogind, which is a fork of logind that doesn't require systemd. I haven't noticed any problems with it.
Systemd is the big one obviously, that's why I mentioned sysv. I'd never heard of logind, so I thought it was sysv.
> ... what other Poettering software is in normal Debian?
A rube-goldberg version of SSH that somehow depends on many things that should be totally unrelated to SSH. Because ofc Poettering needing to mess with everything does need to have systemd notifications available from SSH.
This excuse allows to link many libs and pretend the backdoor attempts are unrelated to systemd, like the liblzma one.
Not of course it's an excuse that doesn't run very far seen the following undisputable fact: the XZ backdoor only affect systemd-enabled SSH. Ouch. Facts do hurt.
I cannot wait for the day a good hypervisor comes out for Linux that runs perfectly on a systemd-less Linux distro, like Devuan (I already used Devuan, but not as an hypervisor).
Basically "Proxmox but systemd-free". I know I've got the FreeBSD+bhyve option too. And at long last I'll be systemd-free again.
Also, a DNS client and NTP reimplementation that have both had security bugs, the broken replacement for syslogd, etc.
After a while, I just lost track. Happily, I don't have to deal with any of that stuff these days.
I'd expect devuan to be a decent hypervisor host, but haven't tried. There's also SmartOS (a few forks away from being Solaris), which looks like it had a release this year. It includes native ZFS.
Honestly, at this point, I'm looking at devuan as the last stop before I jump on the FreeBSD train. It looks like they let you choose between X11 or Wayland, at least for now. I got Steam to work in a FreeBSD VM, but it face-planted because the VM host didn't support any sort of 3d acceleration.
Hopefully enough users will revolt to keep X11, systemd-free Linux viable, but I wonder if that particular niche (which still works great out of the box) is going to end up less popular than the BSDs.
It's the 128gb variant, and I chose it for that reason exactly. I can dedicate up to 96gb to the iGPU
You can probably just dedicate 1GB for the framebuffer, and then let Linux dynamically allocate memory to it at runtime. As far as I can tell this doesn't impact performance, so there's no downside. (Older AMD stacks required a static partitioning under Linux, I think).
I haven’t had much time with it, but I’ve had to set the split in the BIOS. There’s probably a way to do it from within Linux though. Also hoping some progress is made on using the AMD NPU in Linux. I know it only recently got kernel level support.
I thought I had to split it in the BIOS, but then I just didn't (this is on a 2025 machine), and llama ended up with the same available "GPU" ram either way (confirmed by running inference on it).
Oh that's fantastic, I'll give it a try. thank you!
Finally there's good news for Linux users: https://github.com/FastFlowLM/FastFlowLM
Amazing, thank you!