Skip to main content

Project Profile

An image of Delta cabinets in the NPCF machine room.

The Next Big Wave.

Funded by the National Science Foundation’s Innovative High-Performance Computing Program and building on the success of Blue Waters, Delta is the most performant GPU computing resource in NSF’s portfolio, making it a prime destination for advanced scientific research.

More Than CPUs.

NCSA continues its contributions to supercomputing innovation with Delta, a computing and data resource that balances cutting-edge graphics processor and CPU architectures that will shape the future of advanced research computing.

NCSA integrated Delta into the national cyberinfrastructure ecosystem through the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program and partner with the Science Gateways Community Institute to provide platform access serving a broad range of needs. Boasting a non-POSIX file system with a POSIX-like interface, Delta allows applications to reap the benefits of modern file systems without rewriting code. And the Delta team advances accessibility, providing greater usability of the interfaces by the widest possible audience, and in helping emerging research areas, such as computational archaeology and digital agriculture, take advantage of new computing methods.

Delta Might Be Right for You If:

  • you need access to the latest NVIDIA GPUs for your accelerated code
  • you’re interested in transitioning from CPU to GPU while running a CPU workload
  • you have large memory demands for a shared-memory application such as in-memory databases, etc.

Delta provides ample professional development opportunities to adapt research applications to more optimally use its key features. Researchers who currently have GPU projects or are considering migrating to GPU architectures will find ready assistance in migrating the work to Delta.

To apply for a Delta allocation, please visit the Delta allocations page.

Delta Offers:

  • 124 CPU nodes consisting of:
     – Dual AMD 64-core 2.55 GHz Milan processors
     – 256 GB DDR4-3200 RAM
     – 800 GB NVMe solid-state disk
  • 100 quad A100 GPU nodes consisting of:
     – Single AMD 64-core 2.55 GHz Milan processor
     – 256 GB DDR4-3200 RAM
     – 1.6 TB NVMe solid-state disk
     – Four NVIDIA A100 GPUs with 40 GB HBM2 RAM and NVLink
  • 100 quad A40 GPU nodes consisting of:
     – Single AMD 64-core 2.55 GHz Milan processor
     – 256 GB DDR4-3200 RAM
     – 1.6 TB NVMe solid-state disk
     – Four NVIDIA A40 GPUs with 48 GB GDDR6 RAM
  • Five eight-way A100 GPU nodes consisting of:
     – Dual AMD 64-core 2.55 GHz Milan processors
     – 2 TB DDR4-3200 RAM
     – 1.6 TB NVMe solid-state disk
     – Eight NVIDIA A100 GPUs with 40 GB HBM2 RAM and NVLink
  • One MI100 GPU node consisting of:
     – Dual AMD 64-core 2.55 GHz Milan processors
     – 2 TB DDR4-3200 RAM
     – 1.6 TB NVMe solid-state disk
     – Eight AMD MI100 GPUs with 32 GB HBM2 RAM
  • Eight utility nodes will provide login access, data transfer capability and other services
  • 100 Gb/s HPE SlingShot network fabric
  • 7 PB of disk-based Lustre storage
  • 3 PB of flash based storage for data intensive workloads to be deployed spring 2022
Back to top