Top Swiss Tech University Joins Push to Spread GPU-Powered Supercomputing

by Peter Messmer

More performance. Less money. Less power. If computers are to hit new performance milestones in a world of shrinking resources and limited budgets, the industry will have to rethink how it designs and builds these systems.

Hybrid multicore supercomputers – systems that couple high-performance, energy-efficient GPU accelerators with CPUs – are the ideal way to address this issue. Delivering cost-effective, scalable performance while using far less energy than comparable CPU-based systems, these systems, to many industry experts, represent the next 10-20 years of computational science.

Recognizing this, we’re partnering with the Swiss Federal Institute of Technology in Zurich (ETH Zurich) – Albert Einstein’s alma mater – to establish the NVIDIA Co-Design Lab for Hybrid Multicore Computing at ETH Zurich.

Housed in ETH Zurich’s facilities and run by NVIDIA and ETH, the lab will give students and scientists access to GPU-based supercomputing systems and programming environments.

We’ll offer hands-on training, and the latest tools and resources, to help scientists tune their research for current and next-generation GPU-accelerated supercomputers. We’ll also work with scientists to advance their computational models to fully utilize the GPU horsepower.

Driving Advances in Climate, Life Sciences and More

One of the lab’s key aims is to encourage tighter collaboration among computing system architects, integrators, application developers and researchers, providing an open channel for all involved to exchange ideas and experiences.

This, in turn, will be used to speed up the design of new applications and technologies that will drive the next wave of computational scientific research and discovery.

At first, the lab will focus on six scientific domain areas: meteorology and climate; geophysics; materials science and nanotechnology; life science; astro- and plasma-physics; and biomedical engineering and fluid mechanics.

It will support a number of scientific applications that are important to Swiss computational scientists, including the COSMO weather simulation application, the CP2K, GROMACS, and NAMD molecular dynamics applications, the SPECFEM3D seismology and geodynamics application, as well as a host of other codes.

The Swiss National Supercomputing Center (CSCS) in Lugano will collaborate closely with the lab, supplying computing resources and technical expertise to help domain scientists get running quickly and efficiently on these GPU-accelerated systems.

Adding to the available hybrid multicore computing infrastructure, the newly announced CSCS “Piz Daint” GPU-accelerated supercomputer will be available to researchers through the center’s normal allocation process when it opens to the public in spring 2014.

Expected to expand the scope of the center’s research and discovery significantly, Piz Daint is using NVIDIA Tesla K20X GPU accelerators to deliver new levels of energy-efficient compute performance.

In fact, it likely will be the fastest GPU-based supercomputer in Europe.

Researcher Workshop – Dec. 11, 2013

To introduce researchers to lab personnel and provide information on available training materials and supercomputing resources, the lab will host an on-site workshop on Wednesday, Dec. 11, 2013.

To be held at the ETH Zurich/Hoenggerberg campus, Room HPT C103, the workshop is free to attend, and open to researchers in Switzerland and Central Europe.

To learn more or to register, visit