HPC Cluster Expansion

GRC’s dedicated High Performance Computing Cluster has just been doubled in computational power (now a theoretical peak of over 30 TFlops without accelerators) and memory (now over 15 TB), and will soon have 5 times more storage (more than 100 TB).

30 new nodes were supplied by Novatte, are based on Quanta’s HPC Multi-Node Servers, and each have two 10-core Xeon E5-2680 v2 CPUs (20 cores/node), 256 GB of RAM per node, two SAS drives in a RAID 0 configuration and Infiniband connectivity between the nodes and to the shared storage.

Besides the usual performance increase that comes with newer generation CPUs, the larger number of available cores and the fact that these nodes have 3 times more memory per core than our previous “thin” nodes has already allowed, one week after deployment, a ~36 fold speedup in a particularly demanding band structure calculation our researchers are currently performing.