LocalLLaMA

2841 readers

1 users here now

Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.

Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.

As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.

founded 2 years ago

MODERATORS

[email protected]

Advice Wanted - Self Hosting Industrial GPUs (kbin.social)

submitted 2 years ago by [email protected] to c/[email protected]

7 comments fedilink hide all child comments

I just bought a "new" homelab server and am considering adding in some used/refurbished NVIDIA Tesla K80s. They have 24 GB of VRAM and tons of compute power for very cheap if you get them used.

The issue is that these cards run super hot and require extra cooling set ups. I was able to find this fan adapter kit on eBay. But I still worry that if I pop one or two of these bad boys in my server that the fan won't be enough to overcome the raw heat put off by the K80.

Have any of you run this kind of card in a home lab setting? What kind of temps do you get when running models? Would a fan like this actually be enough to cool the thing? I appreciate any insight you guys might have!

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 5 points 2 years ago* (last edited 2 years ago) (1 children)

I've got a k80 and it's... underwhelming.

It's CUDA API is very old (11.4). Nothing works with it - you have to compile all the things from scratch.
The last driver that supported it is nvidia-driver-470 which is not even included anymore in 22.04 ubuntu...
Under debian, you can't (I couldn't...) install both cuda-drivers-470 and nvidia-driver version 470.
It doesn't mix well with other modern cards like 3090.
It idles at around 70W and when in use makes my R730 sound like an industrial vacuum cleaner.
It's not even a really-24Gb card. It's two 12Gb cards wearing a trench-coat.

I does run 30B models tho. And it is cheap.

[–] [email protected] 1 points 2 years ago

so i am looking to get a k80, p40 or 3060 regarding the support for cuda in future i see that it is possible to use a old gpu without the current cuda version even if a program requires it? or is it not usable in some programs today? compiling from scratch isnt a problem and drivers is something i can probably handle to but are there more real problems for future proofing?