Improv System Specs

Linpack Performance 2.51 PFlop/s
Processor 2x AMD EPYC 7713 64C 2GHz
Platform PowerEdge R6525
System Interconnect Infiniband HDR200
System Size 825 Nodes

Available Partitions

Improv has several queues defined. Use the -q option with srun or sbatch to select a queue. The default queue is compute.

Improv Queue Name Description Number of Nodes CPU Type Cores Per Node Memory Per Node Local Scratch Disk
compute Standard Compute Nodes 805 2x AMD EPYC 7713 64-Core Processor 128 256 GB DDR4 960 GB (6 TB bigdata Nodes)
bigmem Large Memory Compute Nodes 12 2x AMD EPYC 7713 64-Core Processor 128 1 TB DDR4 6 TB
debug Reduced Walltime Compute Nodes 8 2x AMD EPYC 7713 64-Core Processor 128 256 GB DDR4 960 GB

The compute queue has 68 nodes with a 6TB local NVMe scratch disk instead of the default 960GB NVMe.

File Storage

Users that do take advantage of local scratch space will have access to the shared GPFS file systems.

Please see our detailed description of the file storage used in LCRC here.

Running Jobs on Improv

For detailed information on how to run jobs on Improv, you can follow our documentation by clicking here: Running Jobs on Improv.

With an eye towards future alignment with the ALCF, LCRC has adopted PBS Pro for the Improv cluster.