• 8 Posts
  • 185 Comments
Joined 1 year ago
cake
Cake day: June 12th, 2023

help-circle
  • Proxmox is based on kvm/qemu, and is very resource conservative. There is virtually no impact on performance due to the hypervisor, even on older processors. Scheduling on the cpu and hypervisor makes running multiple VMs at the same time trivial as well. RAM and I/O bandwidth are the two things that can affect performance. Running out of RAM due to too many VMs will grind you to a halt, but so would running too many applications or containers on bare metal. Running everything off of one spinning sata disk will make it impossible, but again, same downfall on bare metal.

    Those minimal impacts to performance are a minor nuisance compared to the ability to run experiments and learn on sandboxed VMs. Now that TrueNAS has better virtualization support, it has caught my eye as a better homelab solution, but I will always have a proxmox server running somewhere in my stack just due to the versatility it gives me.






  • Just my opinions here:

    • I have essential services running on a separate computer, 8gb pi4 right now. Stuff like NetBoot.xyz, homepage, etc, lightweight and resource low but need to be always up. That way if your main server needs to go down, you still have those services running.

    • I have bought second hand enterprise equipment for most of the hardware I have. Basically anything with ddr4 and pcie 3 or above will crush most things you would like to do. Grabbing an intel with quick sync will help with Jellyfin, but you can add a graphics card for transcoding if you want, a quadro p2000 or higher will be fine. Building is a viable option as well, but you may spend more for less powerful but more efficient hardware.

    • software is probably the most controversial. I went with proxmox on my main server, giving me the ability to run whatever I want whenever I want. It’s not perfect, but gets the job done and has helped me learn A LOT. But flip a coin or roll a dice on what software to run as a newbie, it will all be a learning curve, and everyone will tell you why what they use is superior.

    Whatever you do, you’re not wrong. Run things that tickle your fancy and move at your pace. You’ll mess up, step back and punt a lot. Remember to backup essential data before you wipe. Have fun, and good luck on your travels.





  • He touches on my major issue with all these companies, data mining without compensating the people that created that data. I have to pay for the operating system, get served ads, AND you get to make extra money off my information too? This kind of shenanigans would be tolerable with a free OS, or maybe one that compensated you like brave browser. The blatant fleecing of the consumer here is sickening. I’m glad data mining your screenshots is the last straw for people.





  • Picked up an AMD instinct mi25 to try and do just that. Can get easy-diffusion working after some cussing and voodoo. Cannot get rocm to do ANY llm of any kind, feels like a waste of video ram

    Also have a tesla p4 that runs most text-to-image models rather well, but have been unsuccessful at any llm either, even oobabooga can’t seem to run on it.

    Have given up because the software stack keeps advancing and leaving my hardware behind. I don’t have $3000 for an a100 or $1300 for an mi100 sooo… until the models can run on older/less powerful hardware, I’m probably sitting out of this game. Even though I’d love to be elbow deep in this one.