I'd defintely go with an M.2 SSD, you can get 1tb for 50€ and 2tb for 100€ now and they're much faster, more reliable and take up way less space.
For ML/AI stuff, you might be just fine using an AMD GPU. AMD GPUs are a lot easier to use on Linux and are also a good bit cheaper. I use Fedora with an AMD GPU and I just installed the packages for OpenCL and HIP and now I can run LLMs on my PC using my GPU. I've also used Stable Diffusion with that GPU on Linux before. If there's something specific you want to do regarding that, I'd look up first if you need an Nvidia GPU for that but from my experience AMD GPUs work just fine.
I'd take a look at AMD CPUs again. Last time I checked they were even cheaper (including mobo price) than Intel even though they're also more efficient (faster and less power draw). Prices might have changed tho. You should probably use a Ryzen 5, a Ryzen 7 will only make sense if you use all cores because game performance is pretty much the same. A Ryzen 3 is more of a budget option tho, I wouldn't use that. If it's in your budget, you should also use the newest generation that uses the AM5 socket because you'll be able to upgrade your CPU without needing a new mobo. I think it also only supports DDR5 RAM, which is more expensive than DDR4. If you use a Ryzen generation that uses the AM4 socket, it's gonna be cheaper but if you want to upgrade you'll need a new mobo with AM5 and new DDR5 RAM in addition to the new CPU.
As for Linux distros, my recommendations are Linux Mint if you want something very easy, EndeavourOS if you want something Arch-based or Fedora if you want something that's not quite as easy as Mint but more up-to-date. I personally use Fedora but I used EndeavourOS before. I detailed why I switched to Fedora in a reply here somewhere.
Used EndeavourOS for a few years too but switched to Fedora Workstation recently. EndeavourOS is still great but I like Fedora more now since it's just easier. A lot of stuff I did manually before like switching ext4 for BTRFS, enabling compression and switching to Pipewire is done by default (also LUKS for full diks encryption which I was too lazy to install before) and I can update my system and install most software through GNOME Software without having to use the CLI. It's also very easy to get OpenCL and HIP working, it's just one package each you need to install. Only downside for me is that it's not as easy to install stuff from COPR than it is from the AUR because you first have to enable the repo for each package you want to install from COPR. I think COPR is more secure tho, especially for someone like me who never looked at the PKGBUILD when installing from AUR.
Police officer emptied his entire mag on the car the suspect was sitting in after an acorn fell on its roof because he thought the suspect was shooting at him. I think his partner did the same. The suspect wasn't hit tho.
To work on it from different devices, use features like issues, basically the same reason people use it for public repos instead of just uploading it as a zip somewhere. Sometimes you have stuff you don't want to release to the public or it's just not ready to release to the public yet.
I don't like corporations using the data I put online to train their AI either. I'd happily give it to people to train a FOSS AI tho. I've also contributed my voice to Mozillas Common Voice project, which is an open dataset that everyone is allowed to use. If it's something that everyone can use and benefit from equally, I'm happy to help. I'm not happy to help some corporation make even more profits tho. At least not without getting payed.
"They" in this case isn't the entire EU tho. There's a lot of different politicians there and Patrick Breyer, for example, was against this from the start. That's probably also who I'll vote for (or the Pirate Party, which he's a part of, I don't know how exactly voting works for the EU).
Fisch @ Fisch @lemmy.ml Posts 11Comments 628Joined 4 yr. ago
You can also just disable transcoding in Jellyfin, that's how I'm running it on Pi 4