skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Teraflop Computing for Nanoscience

Abstract

Over the last three decades there has been significant progress in the first principles methods for calculating the properties of materials at the quantum level. They have largely been based on the local density approximation (LDA) to density functional theory (DFT). However, nanoscience places new demands on these first principles methods because of the thousands to millions of atoms present in even the simplest of nano-structured materials. Recent advances in the locally self-consistent multiple scattering (LSMS) method are making the direct quantum mechanical simulation of nano-structured materials possible. The LSMS method is an order-N approach to first principles electronic structure calculation. It is highly scalable on massively parallel processing supercomputers, and is suited for performing large unit cell simulations to study the electronic and magnetic properties of materials with complex structure. In this presentation, we show that the LSMS accomplishes the first step towards understanding the electronic and magnetic structure of nano-structured materials with dimension size close to 10 nanometers (nm). As an example, we describe a 16,000 atom calculation of the electronic and magnetic structure calculated for an iron nanoparticle embedded in iron aluminide crystal matrix.

Authors:
 [1];  [2];  [2];  [2];  [2];  [3]
  1. Pittsburgh Supercomputing Center
  2. ORNL
  3. Florida Atlantic University
Publication Date:
Research Org.:
Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
Sponsoring Org.:
USDOE Office of Science (SC)
OSTI Identifier:
978181
DOE Contract Number:
DE-AC05-00OR22725
Resource Type:
Conference
Resource Relation:
Conference: 2006 Conference on Computing in Nanotechnology (CNAN '06), Las Vegas, NV, USA, 20060626, 20060629
Country of Publication:
United States
Language:
English
Subject:
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE; APPROXIMATIONS; ATOMS; DIMENSIONS; ELECTRONIC STRUCTURE; FUNCTIONALS; IRON; MAGNETIC PROPERTIES; MULTIPLE SCATTERING; PARALLEL PROCESSING; SIMULATION; SUPERCOMPUTERS

Citation Formats

Wang, Yang, Stocks, George Malcolm, Rusanu, Aurelian, Nicholson, Don M, Eisenbach, Markus, and Faulkner, John Sam. Teraflop Computing for Nanoscience. United States: N. p., 2006. Web.
Wang, Yang, Stocks, George Malcolm, Rusanu, Aurelian, Nicholson, Don M, Eisenbach, Markus, & Faulkner, John Sam. Teraflop Computing for Nanoscience. United States.
Wang, Yang, Stocks, George Malcolm, Rusanu, Aurelian, Nicholson, Don M, Eisenbach, Markus, and Faulkner, John Sam. Sun . "Teraflop Computing for Nanoscience". United States. doi:.
@article{osti_978181,
title = {Teraflop Computing for Nanoscience},
author = {Wang, Yang and Stocks, George Malcolm and Rusanu, Aurelian and Nicholson, Don M and Eisenbach, Markus and Faulkner, John Sam},
abstractNote = {Over the last three decades there has been significant progress in the first principles methods for calculating the properties of materials at the quantum level. They have largely been based on the local density approximation (LDA) to density functional theory (DFT). However, nanoscience places new demands on these first principles methods because of the thousands to millions of atoms present in even the simplest of nano-structured materials. Recent advances in the locally self-consistent multiple scattering (LSMS) method are making the direct quantum mechanical simulation of nano-structured materials possible. The LSMS method is an order-N approach to first principles electronic structure calculation. It is highly scalable on massively parallel processing supercomputers, and is suited for performing large unit cell simulations to study the electronic and magnetic properties of materials with complex structure. In this presentation, we show that the LSMS accomplishes the first step towards understanding the electronic and magnetic structure of nano-structured materials with dimension size close to 10 nanometers (nm). As an example, we describe a 16,000 atom calculation of the electronic and magnetic structure calculated for an iron nanoparticle embedded in iron aluminide crystal matrix.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {Sun Jan 01 00:00:00 EST 2006},
month = {Sun Jan 01 00:00:00 EST 2006}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • Wouldn’t it be great to have a teraflop of computing power sitting in your lab, desktop workstation, or remote instrument server? Talk about simplifying workflows, eliminating competition for HPC resources, and allowing more scientists and technicians to get more work done! Well, the computer industry is marketing that capability now in the form of high-end video cards – and for a bargain price – with more and better technology on the market horizon. As the industry evolves to become more oriented toward multi-core and multi-threaded hardware; video card manufacturers are attempting to transition from a niche to multi-purpose market. Onemore » of the products currently getting attention is the Nvidia Tesla family of products based on the Tesla GPGPU (general purpose graphics processing unit). This card contains 128 processor computing core engines advertised as having the ability to deliver an aggregate 518 billion single-precision floating operations per second (518 Gflop), which is being introduced at a $1499 MSRP price-point. Nvidia also offers other commodity graphics cards, such as the GeForce 8800, which appear on paper to have roughly the same performance for roughly half the price – although with half the memory (768M vs the Tesla 1.5 GB). This highlights how the Tesla GPGPUs are essentially redesigned graphics cards (with no video capability, increased memory, and clock changes) that fit into PCI-Express slots in your motherboard. If you believe Nvidia’s claims, two Tesla cards will - for the right applications - turn your lab workstation into a teraflop capable supercomputer. Double-precision versions are projected for a late 2007 introduction with expected 2008 delivery. The Nvidia Tesla GPGPU is one step forward in the many-core revolution that is happening in the computer industry. Instead of making two or four processing cores available to the user, many-core processors offer tens or hundreds of processing cores. Many-core processors promise to provide very high performance-per-dollar and performance-per-watt for many computational workloads. Intel is working on their version of many-core processors but delivery dates appear to be several years in the future. Last year Intel made a large splash with their proof-of-concept teraflop 80-core chip, which they announced might be available sometime in 2011. Intel is also working on something similar to the Nvidia Tesla – codename Larrabee – which will perform in the teraflop range and has a release date of sometime around 2009 or 2010. Larrabee is supposed to have 16 – 24 cores and several nice features. Bottom line: A teraflop lab computer is feasible today as the programmable Nvidia GeForce 8 and Quadro family of graphics cards are available now, Tesla cards will be shipping, and exciting many-core architectures are on the horizon from a number of vendors. Definitely, the potential for parallel processing systems is huge, and GPGPUs certainly provide parallel processing, but are there enough applications out there to take them mainstream and make it more appealing to businesses other than just research firms? Only time will tell as more applications are developed to utilize this computational capability. Right now, programming is required. Recently Google purchased PeakStream, a firm that engaged in abstracting the task of running multiple threads to software with specific GPGPU applicability. However, Google is a visionary software company. Instrument vendors and much of the software industry are still in the early stages of the transition to multi-threaded many-core data processing. Applications that exploit the full potential of parallel processing systems, and GPGPUs in particular, really don’t exist in today’s market. The development of Matlab plug-ins is a very positive sign for the future of GPGPUs and is indicative of Nvidia’s sense of where the market is headed.« less
  • The key to insight is coupling the power of the computer with unique skills of the human. At Sandia National Laboratories' Interaction Laboratory, we call this teraflop visualization. We are concentrating research in three main area: 1) using the computer as a facility for authoring content, 2) adding the physics to model real behaviors, and 3) allowing the human to utilize the improved precision and resolution provided by this new class of compute power.
  • Sandia Laboratories' computational scientists are addressing a very important question: How do we get insight from the human combined with the computer-generated information? The answer inevitably leads to using scientific visualization. Going one technology leap further is teraflop visualization, where the computing model and interactive graphics are an integral whole to provide computing for insight. In order to implement our teraflop visualization architecture, all hardware installed or software coded will be based on open modules and dynamic extensibility principles. We will illustrate these concepts with examples in our three main research areas: (1) authoring content (the computer), (2) enhancing precisionmore » and resolution (the human), and (3) adding behaviors (the physics).« less
  • A workshop was held at the RIKEN-BNL Research Center on the afternoon of October 16, 1998, as part of the first anniversary ceremony for the center. Titled ''Workshop on Physics of the 1 Teraflop RIKEN-BNL-Columbia QCD Project'', this meeting brought together the physicists from RIKEN-BNL, BNL and Columbia who are using the QCDSP (Quantum Chromodynamics on Digital Signal Processors) computer at the RIKEN-BNL Research Center for studies of QCD. In addition, Akira Ukawa, a leader of the CP-PACS project at the University of Tsukuba in Japan, attended and gave a talk on the Aoki phase. There were also others inmore » attendance who were interested in more general properties of the QCDSP computer. The QCDSP computer and lattice QCD had been presented during the morning ceremony by Shigemi Ohta of KEK and the RIKEN-BNL Research Center. This was followed by a tour of the QCDSP machine room and a formal unveiling of the computer to the attendees of the anniversary ceremony and the press. The rapid completion of construction of the QCDSP computer was made possible through many factors: (1) the existence of a complete design and working hardware at Columbia when the RIKEN-BNL center was being set up, (2) strong support for the project from RIKEN and the center and (3) aggressive involvement of members of the Computing and Communications Division at BNL. With this powerful new resource, the members of the RIKEN-BNL-Columbia, QCD project are looking forward to advances in our understanding of QCD.« less
  • A workshop was held at the RIKEN-BNL Research Center on the afternoon of October 16, i 998, as part of the first anniversary ceremony for the center. Titled ''Workshop on Physics of the 1 Teraflop RIKEN-BNL-Columbia QCD Project'', this meeting brought together the physicists from RIKEN-BNL, BNL and Columbia who are using the QCDSP (Quantum Chromodynamics on Digital Signal Processors) computer at the RIKEN-BNL Research Center for studies of QCD. In addition, Akira Ukawa, a leader of the CP-PACS project at the University of Tsukuba in Japan, attended and gave a talk on the Aoki phase. There were also othersmore » in attendance who were interested in more general properties of the QCDSP computer. The QCDSP computer and lattice QCD had been presented during the morning ceremony by Shigemi Ohta of KEK and the RIKEN-BNL Research Center. This was followed by a tour of the QCDSP machine room and a formal unveiling of the computer to the attendees of the anniversary ceremony and the press. The rapid completion of construction of the QCDSP computer was made possible through many factors: (1) the existence of a complete design and working hardware at Columbia when the RIKEN-BNL center was being set up, (2) strong support for the project from RIKEN and the center and (3) aggressive involvement of members of the Computing and Communications Division at BNL. With this powerful new resource, the members of the RIKEN-BNL-Columbia, QCD project are looking forward to advances in our understanding of QCD.« less