LLMpediaThe first transparent, open encyclopedia generated by LLMs

Delta (supercomputer)

Generated by DeepSeek V3.2
Note: This article was automatically generated by a large language model (LLM) from purely parametric knowledge (no retrieval). It may contain inaccuracies or hallucinations. This encyclopedia is part of a research project currently under review.
Article Genealogy
Expansion Funnel Raw 52 → Dedup 0 → NER 0 → Enqueued 0
1. Extracted52
2. After dedup0 (None)
3. After NER0 ()
4. Enqueued0 ()
Delta (supercomputer)
NameDelta
Active2023–present
LocationNational Center for Supercomputing Applications, University of Illinois Urbana-Champaign
PurposeScientific research
ArchitectureHewlett Packard Enterprise Cray EX, AMD EPYC CPUs & AMD Instinct GPUs
Power~1.5 MW
Speed~20 PFLOPS (CPU), ~39 PFLOPS (GPU)
Cost$10 million (NSF award)

Delta (supercomputer). Delta is a heterogeneous, GPU-accelerated supercomputer deployed at the National Center for Supercomputing Applications (NCSA) at the University of Illinois Urbana-Champaign. Funded by a $10 million award from the National Science Foundation (NSF), it entered production in late 2023 as a key resource for the NSF's Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program. The system is designed to expand the capacity for computational science and artificial intelligence research across a wide array of scientific domains.

Overview

Delta serves as a national resource for the United States research community, providing advanced computing capabilities through the NSF ACCESS program. It is operated by the National Center for Supercomputing Applications, a leader in high-performance computing since the era of the Cray-2 and the Blue Waters system. The supercomputer is specifically engineered to support workloads that blend traditional computational fluid dynamics with emerging machine learning and data-intensive applications, filling a strategic niche in the national cyberinfrastructure ecosystem. Its deployment underscores ongoing investments by federal agencies like the National Science Foundation to maintain American competitiveness in scientific computing.

Hardware and architecture

Delta is built on a Hewlett Packard Enterprise (HPE) Cray EX supercomputing platform, integrating both CPU and GPU nodes to create a heterogeneous architecture. The system features 448 CPU nodes, each equipped with two 64-core AMD EPYC "Trento" processors, and 112 GPU nodes, each containing four AMD Instinct MI250X accelerators. This design provides a balanced platform for diverse workloads. Interconnectivity is handled by the high-performance Slingshot network, a technology developed by Cray (now part of Hewlett Packard Enterprise). The entire system is supported by a Lustre (file system)-based parallel filesystem for high-speed data operations and is housed within NCSA's existing data center facilities, which were previously home to the Blue Waters supercomputer.

Software and applications

The software environment on Delta is managed via the HPE Cray Programming Environment, which includes optimized compilers, libraries, and tools for both CPU and AMD Instinct GPU programming. Key supported programming models include OpenMP, MPI, and ROCm for GPU acceleration, enabling researchers to port applications from other systems like Frontier (supercomputer) or Summit (supercomputer). Primary application domains span climate modeling, bioinformatics, materials science, and astrophysics, with specific codes such as WRF (Weather Research and Forecasting Model) and GROMACS being commonly used. The system also supports frameworks like TensorFlow and PyTorch for artificial intelligence research, facilitating hybrid simulation and data analytics workflows.

History and development

The project was initiated following a competitive award from the National Science Foundation in 2021, as part of the agency's mid-range computing infrastructure program. The National Center for Supercomputing Applications was selected to lead the deployment, leveraging its decades of experience operating systems like Blue Waters and Comet (supercomputer). The system was named "Delta" to signify incremental but powerful advancement in capability. Major hardware components from Hewlett Packard Enterprise and AMD were integrated and tested throughout 2022 and 2023, with formal acceptance and transition to production status occurring in November 2023. The deployment timeline was coordinated with the sunsetting of the Comet (supercomputer) at the San Diego Supercomputer Center.

Performance and rankings

In peak theoretical performance, Delta's CPU partition delivers approximately 20 petaFLOPS, while its GPU partition provides around 39 petaFLOPS, for a combined peak of nearly 60 petaFLOPS. This performance places it as a capable mid-tier system within the global supercomputing landscape. While it did not appear on the TOP500 list immediately after deployment, its architectural similarity to the Frontier (supercomputer) suggests high real-world efficiency for supported applications. Its primary metric within the NSF ACCESS program is not raw LINPACK performance but rather sustained scientific throughput across diverse research communities, including those from the LIGO collaboration and the IceCube Neutrino Observatory.

Category:Supercomputers Category:University of Illinois Urbana-Champaign Category:National Center for Supercomputing Applications Category:2023 establishments in Illinois