If you had a $300 budget for a GPU or any PCI-E coprocessor SPECIFICALLY FOR #AI / machine learning, no gaming, new or used, what would you buy? My RTX 3060 ti with it's paltry 8gb vram just doesn't cut the mustard for Stable Diffusion / llama.cpp
#machinelearning #deeplearning #stablediffusion #llama #chatgpt
I was looking at a second hand Tesla P40. They go for around $250, 24GB VRAM, and still reasonably relevant in the AI/Machine Learning space. You might be able to pick up a couple for cheap. I'd avoid the Tesla K series though.
Thanks! Set up a lightning wallet address in your profile so I can zap you!
I will do, just working on the self-custody thing. 😁
I self custody my cold storage, but trust GetAlby with my lightning wallet. Use CashApp to top up, I just never keep much money on Alby or CashApp
I was looking at running Start9's StartOS. It looks good, but I have to wait untilaI can move my server. It's currently in storage due to moving provinces. It'll take me a couple of mohts to get moved properly. In the meantime, everything is in cold storage. 🙂
https://github.com/Start9Labs/start-os
I've been running one for both inference and gaming for about a year. $200 for the card and $50 for fans and a holder. It works well for the price, but lacks some newer architectural features required by some pipelines. if you can justify spending more, a 3090 is faster per dollar and easier all around.
I got a 3060 too but with 12gb of ram, since 40 series is out maybe you could get one cheap.
Oddbean new post about login | logout