Derecho-GPU

Derecho GPU is a GPU-accelerated computing resource within the Derecho system designed for machine learning, AI, and GPU-enabled scientific applications. It consists of dedicated GPU nodes equipped with NVIDIA A100 Tensor Core GPUs and high-bandwidth memory.

Each GPU node includes 4 NVIDIA A100 GPUs, 512 GB of system memory, and high-speed NVLink interconnects between GPUs, enabling efficient multi-GPU computation.

This resource is optimized for workloads requiring GPU acceleration, including deep learning, simulation, and GPU-enabled HPC applications.

Ask about Derecho-GPU

File Transfer

Derecho uses NCAR’s GLADE storage system with multiple file spaces.

Supported Methods Data Transfer Node URL
SCP derecho.hpc.ucar.edu
RSYNC derecho.hpc.ucar.edu
SFTP derecho.hpc.ucar.edu
RCLONE | RECOMMENDED derecho.hpc.ucar.edu https://rclone.org/downloads/
GLOBUS | RECOMMENDED https://www.globus.org/

Storage

File System

Directory Path Quota Purge Backup Notes
Home /glade/u/home/<username> 50 GB Not purged Yes User home directory. Ideal for small scripts, source code, and configuration files that benefit from backup.
Scratch /glade/derecho/scratch/<username> 30 TB / 10M 180 Days No Temporary space. Derecho's scratch file system also includes a limit of 10 Million on a users' total number of files
Work /glade/work/<username> 2 TB Not purged No User work space. Ideal for compiled code, conda environments, and similar large holdings that do not require backup.
Campaign Storage /glade/campaign N/A Not purged No Project space allocations (via allocation request)

Jobs

Jobs on Derecho GPU resources are submitted through PBS Professional and run on GPU-enabled nodes. Each GPU node includes 4 A100 GPUs and associated CPU resources.

Users request GPU resources in their job scripts by specifying the number of GPUs per node. Example:

#PBS -l select=N:ncpus=64:mpiprocs=4:ngpus=4

GPU workloads include machine learning training, AI inference, and GPU-accelerated simulations. Jobs are scheduled using the same queues as CPU jobs but require explicit GPU resource requests.

Queue specifications

Name Purpose CPUs GPUs RAM Jobs
30 days
Wait Time
30-day trend
Wall Time
30-day trend
Derecho--GPU AMD EPYC 7742 (Milan) 2,25 Ghz NVIDIA A100 (40 GB) 8 GB per CPU Core