The Los Alamos National Laboratory (LANL) said this week that its High Performance Computing facilities team has performed a redesign for its Trinity supercomputer that saved the Department of Energy facility approximately $2.6 million in material and labor costs.
In 2014 supercomputer company Cray Inc. received a $174 million, multiyear, multiphase contract to provide the National Nuclear Security Administration with the Trinity system, which will deliver over eight times greater performance than the Cielo supercomputer currently installed at LANL. The redesign, which followed a re-engineering of the wiring diagram for the Trinity computer racks’ power feed, was adopted by Sandia National Laboratories for their corresponding supercomputing systems and is being considered by Lawrence Berkeley National Laboratory.
In addition to a new wiring configuration, the work involved using prefabricated copper tray cables and new aluminum power cables, “saving 20 percent in materials cost and a factor of three in weight,” LANL said in a press release. The latest redesign was funded by NNSA and its Advanced Simulation and Computing program, LANL noted.
LANL spokeswoman Nancy Ambrosiano said by email that the LANL saved costs specifically on the system’s installation and integration. “The $2M in savings compares to about $10M in project costs to upgrade the electrical infrastructure for supercomputing in general, and to attach Trinity in particular to that infrastructure,” she said.
LANL’s Trinity high-performance computer performs simulations and provides the NNSA with the capability “to improve geometric and physics accuracy in calculations that can be completed in weeks – not years,” LANL said.
Supercomputers support NNSA’s defense programs through simulations that certify the reliability of the U.S. nuclear deterrent in order to avoid a return to full-scale underground testing. Supercomputing allows scientists to assess the entire weapons life cycle and predict their functionality, model weapon systems to certify their performance, predict problems caused by weapon aging, model phenomena such as the behavior of high explosives, and simulate abnormal environments to test whether weapons meet safety requirements.
