• 0 Posts
  • 39 Comments
Joined 10 months ago
cake
Cake day: May 28th, 2024

help-circle




  • Highly suggest putting Caddy on a machine, forwarding port 443 and 80 to caddy, and then letting it do your reverse-proxy stuff. Register a domain name, give it your IP address, and then tell caddy that ‘immich.yourdomain.bleh’ goes to port 78789 and plex goes to ‘media.yourdomain.bleh’ port 89898 – Caddy handles all of the TLS stuff, handshaking, you name it - so you can have secure sites with proper certs.

    Then make sure those things are isolated from your home network through vlans if your router supports it.

    You can get fancier with it using a tailscale and getting some datacenter IP to forward into your network



  • kitnaht@lemmy.worldtoLinux@lemmy.mlAMD vs Nvidia
    link
    fedilink
    arrow-up
    1
    arrow-down
    2
    ·
    edit-2
    28 days ago

    Nobody is bitching. Rage less. My constructive point is that NVidia is a better option. NVidia’s CUDA stack is software - and unfortunately for us, that means it’s also paired with their hardware.

    Many people care if choosing something is going to hobble their workflow. In this point, if you’re using Blender, choosing AMD is going to hobble your productivity. I’m just stating facts.



  • kitnaht@lemmy.worldtoLinux@lemmy.mlAMD vs Nvidia
    link
    fedilink
    arrow-up
    1
    arrow-down
    2
    ·
    edit-2
    28 days ago

    Blender works with AMD hardware just great

    No it doesn’t. That’s our point. It works 25% as fast as its competition. That’s not “working just great”…it’s working slowly and like shit. The whole damn point of a GPU is to accelerate that work. The work that your AMD-HIP is doing in blender, could take an hour, and the NVidia would pump it out in 20 minutes.


  • kitnaht@lemmy.worldtoLinux@lemmy.mlAMD vs Nvidia
    link
    fedilink
    arrow-up
    7
    arrow-down
    5
    ·
    28 days ago

    Everyone’s gonna suggest AMD here because of your requirement of no-proprietary drivers; but unless you’re some sort of high-value target to a foreign government, I honestly choose the more pragmatic route of just using the proprietary NVidia driver and going NVidia. Especially if I’m not budget constrained on card.

    The fact of the matter is, AMD has just simply fallen behind. NVidia cards are (and have been for like 3 generations now) more performant. There is good reason why they dominate the market right now; they’re just simply better.

    It really depends on how far you want to take your zealotry on open source; there are parts of the CPU microcode that can see everything you do. Those are proprietary. Your bios is proprietary. You’re probably running 100 different proprietary blobs even IF you choose not to use the drivers that NVidia supplies; so why hobble yourself with a slower card that doesn’t have CUDA instructions? (often also very good for AI work if you are interested in that at all)

    I certainly understand wanting to push that direction for the sake of pushing that direction but - is performance and stability less important than using a proprietary driver?











  • So, I migrated to 5.x and I don’t know if it was just me, or a change in the WebUI or something, but Sonarr stopped wanting to pull files in. I’ve been holding out on the Sonarr upgrade because last I looked at it, it wouldn’t auto-migrate you over, etc.

    But when I went to upgrade it - it said that now auto-migrates, and it does. However, the old migrated rules looked kinda dirty, so I was panicking a little. The imported/converted stuff all worked, mind you, I just didn’t like how they looked. In the end, I ended up really really liking the new Sonarr system, though I did have to ask an LLM how to format some new regex.