Welcome

The University of Maryland has a number of high performance computing resources available for use by campus researchers requiring compute cycles for parallel codes and applications. These are:

Zaratan: Our flagship cluster, intended for large, parallel jobs, housed off campus and maintained by the Division of Information Technology. It consists of over 380 nodes with dual socket (128 cores per node) AMD Milan processors. Twenty nodes also each contain four Nvidia A100 GPUs. All nodes have at least 512 GB of RAM, with six large memory nodes having 2 TB of RAM. All nodes have HDR-100 infiniband (100 Gb/s) interconnects, and there is 2 PB of fast BeeGFS scratch storage.

The following clusters have been retired, but information about them are still provided on the website for historical reasons:

Deepthought2: Our previous cluster, retired in September, 2022. It was intended for large, parallel jobs, housed just off campus and maintained by the Division of Information Technology. It consisted of over 480 nodes with dual socket (20 cores per node) Ivy Bridge 2.8 GHz processors, forty of which had du al Nvidia K20m GPUs. Most nodes had 128 GB of RAM, with a few having 1 TB of RAM. All nodes had FDR infiniband (56 Gb/s) interconnects, and there was 1 PB of fast lustre storage.
Juggernaut: An older experimental cluster, providing compute resources to some users who could not be added to the Deepthought2 cluster because of constraints of its data center, and a testing ground for the next cluster at UMD.

Comparison of Clusters

The following table tries to compare the HPC resources:

Cluster	Number of compute nodes	Cores/node	Processor speeds	Memory/node GB	Nodes with GPUs	Interconnect	Disk space	Licensed Software
Zaratan	386	128	2.45 - 3.5 GHz	512 (6 nodes have 2 TB)	20 (quad Nvidia A100 40GB)	HDR-100 Infiniband GPU nodes have HDR Infiniband	2 PB BeeGFS 10 PB AuristorFS	Intel compiler suite Matlab Distributed Compute Server
Deepthought2 (RETIRED)	488	20 (1TB nodes have 40)	2.8 GHz (1TB nodes are 2.2 GHz)	128 (4 nodes have 1 TB)	40 (dual K20m)	FDR Infiniband	1 PB lustre	Intel compiler suite Matlab Distributed Compute Server

Click on the cluster name above for more detailed information about a particular cluster.