High-performance Computing Cluster Facility
The High-performance Computing Cluster Facility, located in 51 Hammond Building, takes advantage of modern advances in processors. The cluster was built with an emphasis on core density for performance and a small physical footprint.
Assembled in the spring and summer of 2012, the cluster is made up of a single master node and 47 computational nodes built on the 7th generation of Proliant Servers from HP. The master node is a DL585 running on an AMD 6276 “Interlagos” 16-core processor at 2.3 GHz and 24GB of DDR3 RAM. With increased computing capacity comes increased storage demands, and the cluster is configured with 8TB of user storage across two RAID 5 configurations.
Each computational node is built on the DL165 platform and uses two AMD 6276 “Interlagos” 16-core processors, or 32 cores per node, at 2.3 GHz. This results in a total computational processor count of 1,504 cores.
Additionally, each node has 64GB of DDR3 RAM, with the exception of four dedicated nodes having 96 GB RAM. The cluster also offers a node dedicated to GPU processing, offering two NVIDIA C2075 Tesla cards for a total of 896 GPU cores and 12 GB of dedicated memory. The entire platform across two racks is networked together using a low latency 20Gb/s Infiniband fabric.