Delta: Balancing GPU Performance with Advanced System Interfaces

Gropp, William; Boerner, Tim; Bode, Brett; Bauer, Greg

Permalink

https://hdl.handle.net/2142/117179 Copy

Description

Title

Delta: Balancing GPU Performance with Advanced System Interfaces

Author(s)

• Gropp, William
• Boerner, Tim
• Bode, Brett
• Bauer, Greg

Issue Date

2023

Keyword(s)

• Delta
• HPC
• High Performance Computing
• GPU
• NSF
• NSCA
• Usability
• Accessibility

Abstract

Advancing the productivity of computational researchers happens in a variety of ways from hardware advances to software developments. Delta–the NSF-funded advanced computing resource awarded to the University of Illinois at Urbana-Champaign’s National Center for Supercomputing Applications–seeks few compromises between its GPU-focused computing power, high-performance file system, and the interfaces that will make it easier to use by a broader set of research communities than many of its predecessors. This paper surveys the system hardware and approach to usability and accessibility that will make Delta a major player in democratizing access to the kind of computing power necessary to meet the growing demands of artificial intelligence and machine learning researchers as well as traditional simulation science.

Type of Resource

text

Genre of Resource

technical report

Language

en

Sponsor(s)/Grant Number(s)

National Science Foundation (award OAC 2005572)

Owning Collections