Source: The Platform | Nicole Hemsoth | June 26, 2015
Korean auto manufacturer, Hyundai, has tapped into the top-tier U.S. Titan supercomputer, as well as other smaller systems at Oak Ridge National Lab, to cultivate key materials science breakthroughs that will decrease the weight (and up the efficient) of next-generation cars.
The first question might be how non-U.S. companies are using top supercomputing resources and what competitive issues might be at stake. We will get to that in just a moment, but the second critical question is why, when computational prowess is at the heart of the product development process for auto manufacturers, are they not using their own internal clusters? If it was a matter of sheer capacity, there is always the ability to burst out into a cloud-based resource or to add more nodes, especially for longer-term research and development projects that are not “burst” workloads where an extra thrust is needed to finish one-time or infrequent simulations.
As Dr. Claus Daniel, Deputy Director for the Sustainable Transportation program at Oak Ridge National Lab tells The Platform, it is not that companies, no matter where they are located, could not have the high performance computing resources if they needed them, it’s a matter of focus. “It comes down to both attention span and risk,” he explains. “Even though computational power is a driver for their competitiveness, the cost and investments in the kind of scale found at national laboratories is almost impossible for any single company to afford. And further, having full time staff dedicated to making these systems run with full utilization and get the full power of the applications is not what these companies make easy investments in. They have shareholders and the higher the risk, the harder it is to make investments like this.”
While Daniel agrees that there are other factors that require the help of national labs, including finding ways to get widely-used engineering codes to scale and perform well on massive machines, the root of the issue is that this takes dedicated people. Even if a company is based off U.S. soil, the “trickle down” of computational investments in R&D are worthy of investment, at least for the Department of Energy and other agencies. The publicly-funded Department of Energy fleet of supercomputers at the lab will also be home to other projects for Hyundai, which while based abroad, has over 30,000 employees in the United States, with a growing number of those position in the R&D segment. Hyundai currently manufactures around 700,000 cars in the U.S., a number that is expected to increase, according to Daniel. This was key to the Department of Energy’s decision to allow a foreign company to use U.S. supercomputing resources, although it is not uncommon for non-U.S. companies to win award to run key workloads on machines housed at national labs, Daniel says.
The problems being solved on national lab systems will have an impact on how U.S. auto makers develop their own future vehicles as well. From dense functional theory calculations for promising materials, to looking at what elements go into those materials, Daniel says Hyundai is using a number of systems to understand how new materials for lighter weight vehicles might be processed and cast. He notes that there are other projects he cannot detail, but the goal is to create more efficient, lighter weight cars, which would have an impact on energy reduction goals put forth by the DoE. Oak Ridge has already worked with carmaker Ford to create lighter weight aluminum frames for its F150 series of trucks, and the national lab has also worked with large semi-truck makers to refine undercarriage aerodynamics.
With so many high-end supercomputers at Department of Energy and other sites in the U.S. , as well as supercomputing resources designed specifically around the needs of manufacturing and other segments (the National Center for Supercomputing Applications, for example) one might wonder why we do not tend to hear about this more often.
For one thing, in many cases, the partnerships extend to specific projects, including engineering projects that push scalability limits compute-wise for companies’ internal infrastructure, and in some cases, for work that requires software tuning and scalability expertise that only national lab or research centers can provide. For instance, on the manufacturing front, multiple universities and supercomputing centers have worked to scale computational fluid dynamics and other simulation codes—a process that would internal developers years, not to mention extend past what current licenses allow.
What is unique about the current partnership between Oak Ridge National Lab and Hyundai is that beyond R&D elements, Hyundai will pay Oak Ridge for use of the supercomputing resources. In other words, it’s not just a research exchange. Instead, for another “immediate need that comes with competitive advantage” Hyundai will tap ORNL resources for a price, in much the same way other companies use NCSA, the Rocky Mountain Supercomputer Center, and others. While the projects around creating lightweight materials for next generation automobiles is an important component of the collaboration, both are also extend the R&D reach to include other elements, including research into fuel cells and alternative energy storage. These are Department of Energy objectives overall—and from a competitiveness standpoint, advances in this area that are shared bolster competitiveness overall by bringing new potential technologies into the mainstream market.