I am looking for a cheap gpu to run small llm (e.g. qwen 4b q4_k_m) in a home server, and from where im at, I can get the p4 for $ 70 and the p100 for $ 80, are they even worth it as cuda support has ended for both of them. should I get either of these? if so, which one?