Not really. I’ve had to do quite a bit of experimentation.
My setup that I’ve settled on:
Rocm system libraries from Arch Linux
PyTorch nightly for Rocm pip installed into a venv (see instructions on pytorch homepage)
Set HSA_OVERRIDE_GFX_VERSION to 11.0.0. This is just for the RX7600 and it tells it to use the RX7900 code as the pytorch version hasn’t been compiled with 7600 support.
Hmm, that’s weird. I was able to run Stable Diffusion locally with Linux + RX6600.
Probably because I used Easy Diffusion. At first, I couldn’t get the GPU acceleration to work, and I was constantly running out of RAM (Not using VRAM), so my system always froze and crashed.
Turns out it was a ROCM bug, that I don’t know if it’s fixed by now, but I remember “fixing it” by setting an environment variable to a previous version.
Then, it all worked really good. Took between 30 seconds to 2 minutes to make an image.
I find that even if you get a touch primary device, make sure to get one with a keyboard, Ubuntu, Fedora, doesn’t matter, KDE, Gnome doesn’t matter, the touch only experience on linux is simply not great. Make extra sure to get the keyboard with it if its optional.
+1 have been trying to make a Linux tablet work. Gnome is alright but it’s got a crap CPU and 2gb of ram and nothing lightweight has good touch support annoyingly
I am, very hesitantly, optimistic for the new smithay based compositors. Cosmic doesn’t have touch support yet, but it’s super light weight, I get better perf then I do even with KDE. I plan on swapping to it full time on my tablet when it gets touch support. (and when some touch friendly gui stuff is available). you also have catacomb which is an actual mobile compositor. Very promising stuff, but still very far out
I was trying things along the lines of hyprland, sway and i3. I have this idea in my head that a touch screen tiling WM would work really well (from what I’ve seen that’s what people love so much about the iPad nowadays anyway)
Hyprland has something called hyprgrass I think which enables touchscreen gestures, still in the process of figuring out how to install that in NixOS though. (it’s got a nix.flake but it’s not in nixpkgs and I’m still unsure of how to install flakes to a traditional configuration.nix setup)
You could probably look into something like paperwm or Niri, I think scrollable window managers have a lot of potential to be a novel but good touch experience
EDIT: Im not sure if niri support touch, I havent tested it, but I think i might actually try it myself when I get the chance now
Open.qa it is an OpenSUSE tool but it can be used to auto test installs of any OS/software. Their open build service also automates and tests package building
Other should stop just using it because it doesn’t work for you? Wouldn’t it make more sense that those who do work on it keep improving it so it doesn’t break?
The main differences between distros boil down to:
init system
default configurations and applications
release cycle
package manager
Most end users don’t mess around too much with their init system and software configuration. With the rise of mainstream distros and application developers opting to ship desktop applications as snaps/flatpak/appimages, the last two points have less importance than ever.
IMHO, considering this, most of the discussions surrounding distros is relatively silly. After using Linux for almost 20 years at this point, I think I can safely say I could be productive on most popular distributions, with minor adjustments to my workflow.
For a new user? Just pick one of the main distros, that supports the software you need, and roll with it for a while. It won’t make much of a difference. Distro hopping doesn’t make one learn much outside using a different package manager.
Despite all the progress in terms of Wayland, I still find my laptop to be unstable with plasma + Wayland on fedora 38. Many visual bugs, when the screensaver is entered and I move my mouse again, the screen just stays black until I close and open the lid.
Some booting and spontaneous shutdown issues too, but I assume that’s something else. (Framework 12 DIY)
Wayland limits me more than I’d like, with no global hotkeys and general low hackability. The only thing keeping me on it is the fact that I can’t figure out how to get fractional scaling on gnome xorg (also on fedora on a framework)
Scaling is one of the major things that suck. Probably on xorg too through. And especially with multiple screens in different ratios and uncommon ratios (like the frameworks 2:3 one)
Yeah, same experience on Wayland + GNOME for me. I want it to work, but stuff just breaks too often for me to accept at this point. How much of that is Wayland and how much of it is other things failing to work properly with it is kind of immaterial. Regardless, I’ll happily jump ship when it’s more baked, but now isn’t that time.
I would count myself among the people who dont have a huge attachment to x11 and am excited by the modern approach provided by wayland.
Ultimately, I just want my stuff to work. I am running pop and I tried booting into wayland, since they provide that as an option, but I was getting hardlocks. Something I haven’t had on a PC in over a decade. According to the log files it appeared to be related to wayland, so I switched back to x11 and haven’t had any issues since.
I am happy to switch to wayland, but I’ll be waiting on the pop devs to make it a focus–presumably after cosmic DE is out.
Yes, wayland by design does not let random applications grab events intended for other applications nor does it let random applications take screenshots at any point in time showing other applications screens. This requires applications to do screen sharing differently, and it indeed breaks random applications sending events to random other applications. That is basically all you wail about and an absolutely necessary property of any sensible system and it is very embarrassing that it took so long to get this.
It’s not really worth it, honestly. All netplan does is generate a config for systemd-networkd. It’s better to just configure systemd-networkd directly and have a portable configuration, rather than use Canonical’s proprietary stuff. The documentation is quite good for systemd in general, and with more people using it directly for network config it’s easier to find examples when you need help.
The dock is rendered directly by the compositor in one of the examples; it’s not an external application as it ideally should be. It doesn’t rely on any intricate protocols or systemd services to monitor the states of apps. I added it solely for demonstration purposes.
Lots of great answers here already so I will only address a couple of things that haven’t been mentioned:
Regarding Fedora Silverblue:
Currently, Fedora Atomic Desktops are in a major shift to accept OCI container images for delivery of packages. This means that the built image becomes one compliant to OCI and that we boot into an OCI container as our system. As OCI images are relatively declarative (not to the extent that NixOS does (yet)), it becomes possible to have a set of config files (most importantly, the so-called Containerfile) in which your system is ‘declared’/‘configd’. In case you’re interested into how this looks/works, consider taking a look at uBlue’s startingpoint or if you’re more interested in the scope of configuration into Bazzite and/or Bluefin.
apx is available as a COPR on Fedora Atomic Desktops.
Nix can be installed on Fedora Atomic Desktops using Determinate Systems’ installer.
Regarding Vanilla OS:
They’re also moving to a model that’s very close to where Fedora Atomic Desktops is heading towards. So, expect a similar way to config/‘declare’ your system.
What are your thoughts on the three four distros mentioned above?
It’s a question of polish if you’d ask me. With Fedora Atomic Desktops and NixOS being advantageous due to being more established and better funded. I wouldn’t write off Vanilla OS yet as they seem to know what they’re doing. Though, I wouldn’t keep my hopes up for blendOS as its main developer was unaware of which MAC was configured by default on blendOS (spoiler alert: none, at least at the time).
Furthermore, NixOS is literally its own thing and unfortunately infamous for its steep learning curve. If you can afford to learn and conquer NixOS, then NixOS should be the recommendation; unless (like me) you seek SELinux on your systems.
Between Fedora Atomic Desktops and Vanilla OS; Vanilla OS is still in its major rewrite/revamp. The alpha builds are there, but I wouldn’t recommend using those on production machines. Fedora Atomic Desktops, on the other hand, has been going strong for a while now and the uBlue-team has even succeeded in making the OCI-stuff accessible for the general (Linux) public. So if you want to switch now and NixOS is/seems too hard; then Fedora Atomic Desktops it is. On that note, I recommend to check out the uBlue project.
Which ones are the most interesting, and for what reasons?
Honestly, all of them are really interesting, but NixOS does the most unique stuff; with only Guix doing something similar within the Linux landscape. To give you a taste of some of the wild stuff found on NixOS; there’s the so-called Impermanence module which -to my knowledge- happens to be the closest thing to a usable stateless system we’ve got; period. Consider reading this excellent blog post in case you’re interested to know what this entails.
Can’t even being to agree enough on this. Unless you specifically need something that an SBC - ARM or X86 - offers, a second hand thin client or USFF computer will be a better fit, plus they come with high-quality power supplies and solid cases.
Well, not really. The HP g3 mini is roughly the size of a paperback book and costs around 100€, depending on the specs. Similar devices of slightly older makes are even cheaper.
So, yes, they are physically larger, but still pretty small. Chances are, you don’t actually need a tiny device like a Pi, so you should at least consider SFF PCs.
linux
Top
This magazine is from a federated server and may be incomplete. Browse more on the original instance.