Several of my group members and I have started using CUDA for large complex data runs. The problem is our laptops/desktops are basically unusable while these runs are happening. I was asked to investigate setting up a CUDA server or cluster with the following requirements:
Requirements:
- Easy access via remote desktop or some over GUI (some are not comfortable without a GUI)
- CPU code executes on the server as well as CUDA code.
Optional But Preferred:
- Multiple users can perform runs simultaneously.
Can some suggest a framework and a CUDA capable cards. The code we are running doesn’t really require high end numeric computation (Teslas) but if we are using remote desktop we need a card with TCC mode I believe.
TIA