this post was submitted on 07 Mar 2026
68 points (95.9% liked)

Selfhosted

60281 readers
421 users here now

A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don't control.

Rules:

Detailed Rules Post

  1. Be civil.

  2. No spam.

  3. Posts are to be related to self-hosting.

  4. Don't duplicate the full text of your blog or readme if you're providing a link.

  5. Submission headline should match the article title.

  6. No trolling.

  7. Promotion posts require active participation, with an account that is at least 30 days old. F/LOSS without a paywall has exceptions, with requirements. See the rules link for details.

Resources:

Any issues on the community? Report it using the report flag.

Questions? DM the mods!

founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[–] Shimitar@downonthestreet.eu 6 points 3 months ago (2 children)

I plugged in an NVIDIA gpu in my server and enabled ollama to use it, diligently updated my public wiki about it and now enjoying real time gpt: OSS model responses!

I was amazed, time cut from 3-8 minutes down to seconds. I have a Intel Core7 with 48gb ram, but even an oldish gpu beats the crap out of it.

[–] mierdabird@lemmy.dbzer0.com 2 points 3 months ago

In that same vein I got an AMD Pro V620 32GB off ebay and have been struggling to get it to POST on my x570 motherboard, but I finally tried it on my old ASUS b450-i with a Ryzen 5 2400GE and with a few BIOS setting changes it fired right up.

Now I need to figure out what I'm doing wrong on the x570 board so I can run the V620 combined with my 9060XT for bigger models

[–] sharkaccident@lemmy.world 0 points 3 months ago (1 children)
[–] Shimitar@downonthestreet.eu 2 points 3 months ago

NVIDIA Corporation GA104GL [RTX A4000] (rev a1)

From lspci

It has 16gb of VRAM, not too much but enough to run gpt:OSS 20b and a few other models pretty nice.

I noticed that it's better to stick to a single model, I imagine that unload and reload the model in VRAM takes time.