HPC News

In late 2017, IBM will begin delivery of Sierra, the latest in a series of leading-edge Advanced Simulation and Computing (ASC) Program supercomputers. Delivering peak speeds of up to 150 petaflops (10¹⁵ floating-point operations per second), Sierra is projected to provide at least four to six times the performance of Sequoia, Livermore’s current flagship supercomputer. To run efficiently on Sierra, applications must be modified to achieve a level of task division and coordination well beyond what previous systems demanded.

Two Penguin Computing systems installed at Lawrence Livermore National Laboratory, Quartz and Jade, were ranked 41st and 42nd on the TOP100 list of the world's fastest supercomputers. The announcement came during the SC16 supercomputing conference, held November 13–18, 2016, in Salt Lake City, Utah. The systems were procured under NNSA’s Tri-Laboratory Commodity Technology Systems program, or CTS-1, to bolster computing for national security at Los Alamos, Sandia, and Lawrence Livermore national laboratories.

A Lawrence Livermore team's dramatically improved first-principles molecular dynamics code that promises to enable new computer simulation applications was one of the finalists for the 2016 Gordon Bell Prize. The team presented its ground-breaking project at the 2016 supercomputing conference (SC16) held in Salt Lake City, Utah, November 12-18, 2016. "Modeling Dilute Solutions using First-Principles Molecular Dynamics: Computing more than a Million Atoms with over a Million Cores," was the title of the Livermore team's submission for the competition. Using a robust new algorithm, the Livermore team has developed an O(N)complexity solver for electronic structure problems with fully controllable numerical error.

Lawrence Livermore scientists are among those who have been awarded funding to develop software for the Department of Energy's Exascale Computing Project (link is external) (ECP). The ECP selected 35 software development proposals representing 25 research and academic organizations; Livermore computer scientists will lead six of the projects and are collaborators on another seven. The awards, totalling $34 million for the first year of funding, cover many components of the software stack for exascale systems.

Laboratory computer scientists and Norwegian researchers are collaborating to apply high performance computing (HPC) to the analysis of medical data to improve screening for cervical cancer. The convergence of high performance computing, big data, and life science is enabling the development of personalized medicine. "Delivering care tailored to the needs of the individual, rather than population averages, has the potential to transform the delivery of healthcare," says LLNL's Ghaleb Abdulla, Livermore lead on the collaboration with Norway and director of LLNL's Institute for Scientific Computing Research.

A U.S. Department of Energy (DOE) program designed to spur the use of high performance supercomputers to advance U.S. manufacturing has funded 13 new industry projects for a total of $3.8 million. Experts at Lawrence Livermore and other DOE national laboratories will work directly with manufacturing industry members to teach them how to adopt or advance their use of high performance computing (HPC) to address manufacturing challenges with a goal of increasing energy efficiency, reducing environmental impacts, and advancing clean energy technologies.

The Department of Energy will invest $16 million over the next four years to accelerate the design of new materials through the use of supercomputers. Two four-year projects will take advantage of superfast computers at DOE national laboratories by developing software to design fundamentally new functional materials destined to revolutionize applications in alternative and renewable energy, electronics, and a wide range of other fields. Lawrence Livermore will use its Vulcan supercomputer to study transition metal-oxides with the goal of developing new theoretical methods and simulation capabilities for the predictive calculation of the properties of complicated materials for energy applications.

In January 2016, President Obama signed a Presidential Memorandum establishing a first-of-its-kind federal task force to end cancer as we know it. Led by Vice President Joe Biden, the Cancer Moonshot is a national effort to double the rate of progress—to make a decade's worth of advances in cancer prevention, diagnosis, treatment, and care in five years—and to ultimately end cancer. Lawrence Livermore National Laboratory's (LLNL) high-performance computing will play a major role in the initiative.

The Department of Energy (DOE) has launched three new pilot projects focused on bringing together nearly 100 cancer researchers, care providers, computer scientists and engineers to apply the nation's most advanced supercomputing capabilities to analyze data from preclinical models in cancer, molecular interaction data, and cancer surveillance data across four DOE national laboratories. The Collaboration of Oak Ridge, Argonne and Livermore (CORAL) supercomputing, led by Lawrence Livermore, which is typically used for stockpile stewardship, will be applied to biology to refine the understanding of the mechanisms leading to cancer development and accelerating the development of promising therapies that are more effective and less toxic.

Lawrence Livermore will receive a first-of-a-kind brain-inspired supercomputing platform for deep learning developed by IBM Research. Based on a breakthrough neurosynaptic computer chip called IBM TrueNorth, the scalable platform will process the equivalent of 16 million neurons and 4 billion synapses and consume the energy equivalent of a hearing aid battery—a mere 2.5 watts of power.

Lawrence Livermore National Laboratory (LLNL) and partners announced 10 new industry projects to advance manufacturing using high-performance computing (HPC) under a DOE program. Industry projects ranging from improved turbine blades for aircraft engines and reduced heat loss in electronics to waste reduction in paper manufacturing and improved fiberglass production are among the first to be selected for funding and partnerships with national labs under the U.S. Department of Energy's (DOE) High Performance Computing for Manufacturing (HPC4Mfg) Program. LLNL leads the program and partners with Lawrence Berkeley and Oak Ridge National Laboratories (LBNL and ORNL).

Each of the 10 Phase I projects will be funded at approximately $300,000 for a total of just under $3 million. Selected companies will partner with national labs, which will provide expertise in and access to high performance computing systems aimed at high-impact challenges. The Advanced Manufacturing Office (AMO) within DOE's Office of Energy Efficiency and Renewable Energy (EERE) created this program to advance clean energy technologies, increase the efficiency of manufacturing processes, accelerate innovation, shorten the time it takes to bring new technologies to market, and improve the quality of products.

The collaboration of Oak Ridge, Argonne and Lawrence Livermore (CORAL) that will bring the Sierra supercomputer to the Lab in 2018 has been recognized by HPCWire with an Editor’s Choice Award for Best HPC Collaboration between Government and Industry

Supercomputers at Lawrence Livermore National Laboratory (LLNL) will be retrofitted with liquid cooling systems under a California Energy Commission (CEC) grant to assess potential energy savings. Reducing power consumption and the associated costs at data centers and high performance computing facilities is a leading concern of the HPC community," says Anna Maria Bailey, LLNL high performance computing facility manager, "addressing this issue is critical as we develop ever more powerful next-generation supercomputers."

Lawrence Livermore National Laboratory (LLNL) and the Rensselaer Polytechnic Institute (RFI) will combine decades of expertise to help American industry and businesses expand use of high performance computing (HPC) under a signed memorandum of understanding. Livermore and RPI will look to bridge the gap between the levels of computing conducted at their institutions and the typical levels found in industry. Scientific and engineering software applications capable of running on HPC platforms are a prime area of interest.

EVERYTHING changes. Nowhere is that maxim more apparent than in the world of computing. From smartphones and tablets to mainframes and supercomputers, the system architecture—how a machine’s nodes and network are designed—evolves rapidly as new versions replace old. As home computer users know, systems can change dramatically between generations, especially in a field where five years is a long time. Computational scientists at Lawrence Livermore and other Department of Energy (DOE) national laboratories must continually prepare for the next increase in computational power so that the transition to a new machine does not arrest efforts to meet important national missions.

That next jump in power will be a big one, as new machines begin to approach exascale computing. Exascale systems will process 10¹⁸ floating-point operations per second (flops), making them 1,000 times faster than the petascale systems that arrived in the late 2000s. Computational scientists will need to address a number of high-performance computing (HPC) challenges to ensure that these systems can meet the rising performance demands and operate within strict power constraints.

We present a work-in-progress snapshot of learning with a 15 billion parameter deep learning network on HPC architectures applied to the largest publicly available natural image and video dataset released to-date. Recent advancements in unsupervised deep neural networks suggest that scaling up such networks in both model and training dataset size can yield significant improvements in the learning of concepts at the highest layers. We train our three-layer deep neural network on the Yahoo! Flickr Creative Commons 100M dataset.