My lab plans to purchase machines with a deep learning framework to support our work (I have two other colleagues). The 2 options that we have are:
- Buy 3 PCs with a single GPU
- Buy 1 Server Rack with 3 GPUs
My colleagues and I work on different problems, thus, we will run different DL models. My question is, which of the above option to be the best option? If we bought the 2nd option, is it possible to tell our DL model to run on a specific GPU (so that three of us can work in parallel)?
Thanks in advance.