Hopper (supercomputer)

Hopper (supercomputer). Hopper was a Cray XE6 supercomputer deployed at the National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory. Named for pioneering computer scientist Grace Hopper, it served as a primary resource for United States Department of Energy research from 2010 until its retirement in 2015. The system was instrumental in advancing computational science across fields like climate modeling, materials science, and high-energy physics.

Overview

Hopper was installed at NERSC in 2010 as part of the Department of Energy's ongoing investment in leadership-class computing facilities. The system was operated by staff at Lawrence Berkeley National Laboratory to support thousands of researchers from national laboratories, universities, and industry. It succeeded the earlier Franklin (supercomputer) and was itself succeeded by the Cori (supercomputer) in 2016. Hopper's operational period coincided with a significant expansion of computational science within the DOE Office of Science.

Architecture

The system was built by Cray Inc. as a Cray XE6 model, utilizing the company's proprietary Gemini (interconnect) for high-speed data transfer between nodes. Its compute nodes were based on AMD Opteron processors, specifically the "Magny-Cours" series, with each node containing two sockets of 12-core CPUs. Hopper's memory hierarchy included a total of 212 terabytes of DDR3 memory and utilized the Lustre (file system) for parallel storage. The entire system ran on the Cray Linux Environment and was cooled using innovative liquid cooling techniques to manage its substantial power consumption.

Performance and rankings

At its launch, Hopper achieved a peak performance of 1.28 petaFLOPS as measured by the LINPACK benchmark. This performance earned it the 5th position on the worldwide TOP500 list in November 2011, marking a significant achievement for NERSC and the Department of Energy. Throughout its service life, it consistently ranked among the top systems globally, also featuring prominently on other benchmarks like the HPCG list. Its computational efficiency was critical for large-scale simulations requiring millions of processor hours annually under the INCITE program.

Applications and research

Hopper supported a vast portfolio of scientific research funded primarily by the DOE Office of Science. Major projects included high-resolution climate models developed for the Intergovernmental Panel on Climate Change and complex simulations of combustion for cleaner energy technologies. Researchers used it for quantum chromodynamics calculations in particle physics, density functional theory studies in materials science, and analyzing data from the Large Hadron Collider at CERN. The system also enabled breakthroughs in astrophysics, such as modeling supernova explosions and the formation of galaxies.

History and development

The procurement and deployment of Hopper were managed through a competitive process led by the Department of Energy and Lawrence Berkeley National Laboratory. The system was named in honor of Grace Hopper, a United States Navy rear admiral and pioneer in computer programming. After its installation in 2010, it underwent several upgrades, including expansions to its interconnect network and storage system. Hopper was officially retired from service in 2015, with its computational workload transitioning to the next-generation Cori (supercomputer), which was based on Intel Xeon Phi and later Intel Xeon processors.

Category:Supercomputers Category:Cray supercomputers Category:Lawrence Berkeley National Laboratory Category:2010 establishments in California