NSF ACCESS Resources
Local entryways to the nation’s cyberinfrastructure and ACCESS high-performance computing resources.
NSF Cyberinfrastructure Resources
NSF ACCESS Resources/NCAR
Discover local entryways to the nation’s cyberinfrastructure. On-Ramps is an online tool to browse available ACCESS resources to facilitate an increasing educational and research community for utilization of its full range of HPC options that provides educators and researchers with powerful HPC resources when and where they need them. It extends the ACCESS program’s reach, and empowers community members by giving them a larger toolbox to choose from.
Research and Cyber Computing
The ACCESS On-Ramps discovery tool connects the campus community to the nation’s NSF-funded advanced computing systems and services. Explore ACCESS resources and request an allocation below:
The live ACCESS On-Ramps tool is an embedded interactive component; use the links above to browse the full, up-to-date catalog of ACCESS cyberinfrastructure resources.
ACCESS Allocated Resources
The following NSF-funded advanced computing systems and services are discoverable through the ACCESS On-Ramps tool. Browse the full, up-to-date catalog and request an allocation at access-ci.org.
CloudBank
CloudBank Classroom
A cloud-hosted, small-scale JupyterHub for teaching, built on an open-source stack (Python, R, JupyterLab). Each learner gets a modest allocation of persistent storage, memory, and CPUs, signs in with existing campus credentials, and works in a browser preloaded with common data-science tools—no local setup. Instructors distribute materials and manage assignments through integrated tools such as nbgitpuller and grading extensions. GPUs are expected on the platform by fall 2026.
CloudBank Research
Gives researchers access to commercial cloud platforms—Amazon Web Services, Google Cloud, IBM Cloud, and Microsoft Azure—through ACCESS and NAIRR Pilot allocations awarded in dollars. The CloudBank portal provisions cloud accounts, manages authorized users via federated ACCESS login, tracks usage against the allocation, and sends automated spend alerts. Regulated data is supported.
Indiana Jetstream2
Indiana Jetstream2 CPU
A hybrid-cloud platform offering on-demand, programmable cyberinfrastructure from interactive VMs to orchestration services. The standard CPU resource uses AMD Milan 7713 processors with 128 cores and 512 GB RAM per node on a 100 Gbps fabric.
Indiana Jetstream2 GPU
The separately allocated GPU portion of Jetstream2: 360 NVIDIA A100 GPUs (four per node), with 128 AMD Milan cores and 512 GB RAM per node on a 100 Gbps fabric.
Indiana Jetstream2 Large Memory
The large-memory portion of Jetstream2: 32 GPU-ready nodes with 1 TB RAM each and AMD Milan 7713 processors (128 cores/node) on a 100 Gbps fabric.
Indiana Jetstream2 Storage
Storage allocation on the Jetstream2 hybrid-cloud platform.
NCSA Delta
NCSA Delta CPU
124 dual-socket nodes for general-purpose computation across domains such as weather and climate, hydrodynamics, astrophysics, and engineering. Each node has two AMD EPYC 7763 “Milan” CPUs (128 cores), 256 GB RAM, and an 800 GB NVMe scratch disk, on a 100 Gb/s HPE Slingshot fabric.
NCSA Delta GPU
GPU nodes supporting both simulation and AI/ML in five configurations. Capacity is mainly 200 single-socket AMD EPYC 7763 nodes (256 GB RAM)—half with four NVIDIA A100s and half with four A40s—plus dense eight-GPU nodes (A100, MI100) and H200 nodes, on a 200 Gb/s HPE Slingshot fabric.
NCSA DeltaAI
114 NVIDIA quad Grace Hopper nodes on an HPE Slingshot interconnect. Each node holds four superchips, each pairing an ARM CPU (128 GB LPDDR5) with an H100 GPU (96 GB HBM), tightly coupled by NVLink with unified shared memory.
PSC Bridges-2
PSC Bridges-2 Extreme Memory (PSC Bridges-2 EM)
Part of Bridges-2, which combines HPC, high-performance AI, and large-scale data management. Extreme Memory nodes target memory-intensive work such as genome assembly, graph analytics, and in-memory databases: four Intel Xeon Platinum 8260M “Cascade Lake” CPUs, 4 TB RAM, and 7.68 TB NVMe, linked to the Ocean filesystem by 400 Gbps InfiniBand.
PSC Bridges-2 GPU (PSC Bridges-2 GPU)
Bridges-2 GPU nodes are optimized for scalable AI and deep learning and also serve accelerated simulation. Four node types are available (H100-SXM5-80GB, V100 with and without NVLink, and L40S), connected to the Ocean filesystem by dual HDR-200 InfiniBand (400 Gbps).
PSC Bridges-2 Regular Memory (PSC Bridges-2 RM)
Bridges-2 Regular Memory nodes provide general-purpose HPC, machine learning and data analytics, AI inferencing, and pre/post-processing. Each has two AMD EPYC “Rome” 7742 CPUs (64 cores), 256–512 GB RAM, and 3.84 TB NVMe, on HDR-200 InfiniBand.
PSC Bridges-2 Storage (PSC Ocean)
The Bridges-2 Ocean data-management system: a single-namespace filesystem with a 15 PB Lustre disk tier (up to ~129/142 GB/s read/write) and a 7.2 PB tape tier, transparently managed by HPE DMF.
Purdue Anvil
Purdue Anvil AI
21 nodes, each with four NVIDIA 80 GB H100 SXM GPUs, for machine-learning and artificial-intelligence workloads.
Purdue Anvil CPU
A Dell/AMD cluster of 1,000 nodes (two 64-core AMD EPYC “Milan” CPUs each) delivering a peak 5.1 petaflops and over a billion core-hours a year. Nodes have 256 GB RAM (32 large-memory nodes at 1 TB), on 100 Gbps HDR100 InfiniBand.
Purdue Anvil GPU
16 nodes, each with four NVIDIA A100 Tensor Core GPUs, providing 1.5 PF of single-precision performance for machine learning and AI.
SDSC Expanse
SDSC Expanse CPU
A Dell cluster with AMD Rome processors (128 cores/node) on Mellanox HDR InfiniBand in a hybrid fat-tree, with a 3.373 PF peak. Nodes carry 1 TB NVMe and 256 GB DRAM; the system adds 12 PB of Lustre and 7 PB of Ceph storage.
SDSC Expanse GPU
The GPU component of Expanse: 52 nodes, each with four NVIDIA V100s (32 GB, NVLink) and dual 20-core Intel Xeon 6248 CPUs, 1.6 TB NVMe, and 256 GB DRAM, on HDR InfiniBand, sharing the 12 PB Lustre and 7 PB Ceph storage.
SDSC Expanse Projects Storage
5 PB of storage on a Lustre-based filesystem.
TACC Stampede
TACC Dell/Intel Knights Landing, Skylake System (Stampede2)
TACC’s Dell/Intel system built on Intel Knights Landing and Skylake processors for large-scale simulation and data-intensive research—the predecessor to Stampede3.
TACC Dell/Intel Sapphire Rapids, Ice Lake, Skylake (Stampede3)
An NSF-funded Dell/Intel system serving traditional simulation users, data-intensive computing, and emerging users, on an evolutionary path toward many-core and accelerated technologies while keeping a familiar programming model.
Personnel & Ready to Join?
Personnel
Ready to Join?
For more information, please contact us at (601) 979-3913. We look forward to meeting with you and seeing how our program can help you achieve your goals.
Quick Links
Civil & Environmental Engineering and Industrial Systems & Technology
1400 J.R. Lynch Street
P. O. Box 17068
Jackson, MS 39217, USA