作者: L. Oliker , T.L. Windus , J.J. Dongarra , P. Luszczek , T.C. Schulthess
DOI:
关键词:
摘要: On August 15, 2002 the Department of Energy (DOE) selected Center for Computational Sciences (CCS) at Oak Ridge National Laboratory (ORNL) to deploy a new scalable vector supercomputer architecture solving important scientific problems in climate, fusion, biology, nanoscale materials and astrophysics. ''This program is one first steps an initiative designed provide U.S. scientists with computational power that essential 21st century leadership,'' said Dr. Raymond L. Orbach, director department's Office Science. In FY03, CCS procured 256-processor Cray X1 evaluate processors, memory subsystem, scalability architecture, software environment predict expected sustained performance on key DOE applications codes. The results micro-benchmarks kernel bench marks show be exceptionally fast most operations. best are shown large problems, where it not possible fit entire problem into cache processors. These exactly types ultra-scale simulation. Application found markedly improved by this architecture: - Large-scale simulations high-temperature superconductors run 25 timesmore » faster than IBM Power4 cluster using same number Best parallel ocean (POP v1.4.3) 50 percent higher Japan s Earth Simulator 5 times cluster. A fusion application, global GYRO transport, was 16 Power3. increased allowed fully resolve questions raised prior study. transport AGILE-BOLTZTRAN astrophysics code runs 15 Molecular dynamics related phenomenon photon echo 8 previously achieved. Even 256 system already outperforming other supercomputers thousands processors certain class such as climate modeling some applications. This evaluation outcome meetings both high-performance computing (HPC) vendors application experts over past 9 months has received broad-based support from community agencies.« less