Live from SC11: Turning Big Data into Big Physics

GPU technology is set to play an important role as the era of data-intensive computing in both science and the enterprise continues to unfold. With that said, it is no accident that the keynote speaker this year was NVIDIA CEO, Jen-Hsun Huang who recognized the intersection of new approaches to crunching big data problems.

While we’ll get to more about the GPU vendor’s presence here later this week, there were plenty of examples on the show floor, especially in the startup camp, that showed just how much of a role GPUs are playing in data-heavy simulations, scientific, industrial and in the enterprise context.

In one particular chat about processor innovations and the role of GPUs in science, Matthew Scarpino, a software developer with Eclipse Engineering, discussed how programming tools that enable GPU benefits, including OpenCL and NVIDIA’s own CUDA is a key to powering massive simulations that involve incredibly large and complex data sets.

Scarpino says that dealing with graphics on the GPU and using the physical equations to manipulate the objects on the GPU instead of having to deal with CPU provides great benefits for the kind of simulations they work with at Eclipse. He claims that for one thing, in addition to getting the computational boost, using a framework like OpenCL or CUDA reduces amount of data transfer between the CPU and GPU, allowing for high-speed simulation.

For his company, OpenCL and GPU technology are keys to achieving the high-end FEA, collision detection and aerodynamics simulations they specialize in. His customers have huge amounts of simulation data, are dealing with billions of data points and billions of vertices in a model, so they need to be able to manipulate these large-scale problems rapidly.

While advances in GPU technology, especially as they relate to HPC are one of the more compelling stories to infiltrate the supercomputing realm in the last few years, one can expect to hear more about what they portend for the future of data-intensive computing, both in the lab and in the enterprise.

On that note, take a look at some of the simulations that were on display as part of this year’s visualization showcase to get a sense of the complexity and scope of the types of problems that are being aided by GPUs in science and engineering.