U.S. Risks Losing Edge in HPC, Supercomputing, Report Says
There are other troubling trends, according to the ITIF report: Analysts expect China later this year to bring online two supercomputers capable of 100 Pflops of performance. By contrast, the DOE in 2014 signed up IBM and GPU vendor Nvidia to build two 150-Pflop supercomputers—Summit, for the Oak Ridge facility, and Sierra, for the Lawrence Livermore National Labs. However, those systems aren't due to go online until 2017 or 2018, and another supercomputer—Aurora—isn't due until 2019. This is important, because the future of HPC is in developing exascale supercomputers—those that operate at 1,000 Pflops or more, according to the ITIF authors. The United States, China, Japan and the European Union are among the countries and regions in the race to develop the first of these exascale systems, which is scheduled to happen by 2020—though there are reports the timeline may be pushed back to 2023. Here the United States has a continued advantage, the ITIF report said. "Real national leadership in HPC comes from the combination of superfast systems, designed in a functionally operational, system-efficient, and cost-efficient manner, something at which the United States has long excelled compared with peer nations," the author said.Given this, the country needs to protect and extend its advantage, they wrote. The authors laid out steps Congress must take, including holding hearings on Obama's NSCI and the increasingly competitive global HPC space; authorize the funding necessary for NSCI over the coming years; and reform export control regulations to address the what's happening in the HPC space. In addition, the current and future presidential administrations and their agencies and departments must make technology transfer and commercialization activities a priority for the country's network of national labs and emphasize HPC in federal work training programs. In addition, they should stress HPC in manufacturing extension partnership engagements and help small and midsize enterprises (SMEs) access high-performance computing.
It will be important for the country to keep that advantage, they wrote. It would be risky for the United States to depend on foreign tech vendors to access leading HPC environments that no longer are produced in the country, and "U.S. industries and enterprises benefit through first-mover advantage by having quicker access and availability to leading-edge HPCs," the authors wrote. In addition, supercomputers are key advantages in exports, employment and economics.