NVIDIA K80 for PyTorch

I was wondering if anyone of the forum has any experience using a K80 in 2021 with PyTorch.

I am running a GTX 1070 with 8GB of VRAM, and my models are too big. I have purchased a GT710 to free up some of the VRAM occupied by my operating system, but it appears by current project is going to require 16+GB of VRAM.

Originally I was hoping to be able to purchase a pair of RTX 3060’s or get NVIDIA’s researcher discount on a Titan, but it appears that that is not going to happen in the near future.

Because of this I have been wondering if a K80 could be a viable, cost effective alternative for me.

Has anyone used a K80 recently with PyTorch? If so what is performance like? is it worth it considering I can purchase one for <300 USD?

As always thank you very much for your time and assistance.

In case anyone ever stumbles upon this. I have abandoned this idea due to the fact that I would have to mod rather loud fans on to the K80 heat sink. I will simply wait until graphics card prices drop down to normal.

I am interested in this question. I do not mind the loud noise or cooling.