GPU

Work in progress.

Talking to fermilab and couple other places that are using GPUs.

Waiting for quotes from fermilab. The quote from CMU is below.

We will need a new rack.

My understanding is that it would be better to keep this system independent of the main T3 CPU cluster.

Quote from Carnegie Mellon:

One (210120-52G) is for two generic GPU/CPU nodes 50K each. That hardware is the current sweet spot in terms of GPU/CPU power per dollar.