The Gypsum Compute Cluster got an upgrade this summer, more than doubling in size from 400 to now 824 GPUs (graphics processing units).
Gypsum was born in 2016 thanks to a $5 million grant from the Massachusetts Technology Collaborative to establish a large, state-of-the art high-performance computing system, for the purpose of accelerating research, discovery, and innovation, and to foster collaborations with industry partners who have a high need for computational resources. Located at the Massachusetts Green High Performance Computing Center in Holyoke, MA, the Gypsum Compute Cluster is among the top academic clusters in the country, and now offers 2,472 CPU (central processing unit) cores, and 25 terabytes of RAM (random-access memory), in addition to the 824 GPUs.
New computer science faculty member Mohit Iyyer, puts this into perspective: “I had access to one GPU when I was in grad school, and last year in an industry lab I had four GPUs. Now my students have access to hundreds of GPUs, which significantly speeds up the development cycle. They can train tons of models at the same time, and evaluate them all, whereas before you could only train one model at a time. Months of work can be done in a week.”
For more information on the Gypsum Compute Cluster, visit https://ds.cs.umass.edu/resources/compute-clusters