NCCS GPU CLUSTER AVAILABLE NOW!
The NASA Center for Climate Simulation (NCCS) announces a new resource for NASA-funded scientists: a graphics processing unit (GPU) cluster running on the ADAPT Science Cloud and specifically built for accelerating artificial intelligence (AI), machine learning (ML) and deep learning (DL) workloads using GPUs. This leading-edge platform is easy to access and the preinstalled software/libraries provide foundational tools that enable scientists to maximize their workflows.
The new cluster allows science teams to exploit GPU architecture to run very large simulations, conduct predictive analytics, use portable programming models, standard AI/ML frameworks, and foster virtual collaboration.
GPU Cluster from Aspen Systems
22 nodes, each with:
- Dual Intel Xeon Gold 6248 “Cascade Lake” CPUs, 20 cores each @ 2.5 GHz
- Four NVIDIA V100 GPUs with 32 gigabytes (GB) of VRAM each
- 768 gigabytes (GB) of RAM
- 3.8 terabytes (TB) of local NVMe storage (mounted at /lscratch)
- Dual 100-gigabit (Gb) HDR100 InfiniBand
- Dual 25-gigabit (Gb) Ethernet, bonded for high availability
Accessing the Cluster
To gain access to the new GPU cluster, users should contact the NCCS User Services Group and request access to the GPU system on the ADAPT Science Cloud. For more information about using the system, see GPU Usage on ADAPT.
Impact to Science
“The new GPU cluster addresses a pressing scientific need,” explained NCCS scientist Ellen Salmon. “It is intended as an initial testbed for NASA research scientists who have reached the limits of computing using traditional CPU processing capabilities. In order to explore and process the increasingly large datasets coming from satellites, sensors, and simulations, we need to provide our scientists with an efficient means of pushing beyond traditional processing boundaries.”
Sean Keefe, NASA Goddard Space Flight Center