Graphics processing units are revolutionizing the supercomputing industry by providing a combination of performance and efficiency that is unattainable with heterogeneous CPU systems. To get in touch with the rapidly developing GPU movement in high-performance computing, check out the NVIDIA® GTC conference (GPU Technology Conference) that will take place March 18-21 in San Jose, California.
An example of this trend comes in the form of a project led by Alan Gray of the University of Edinburgh on the physics of soft matter. According to an InsideHPC report, Gray and his team are working to study ice cream at its molecular level to improve its texture and taste. Initially, the scientists performed this research with a large supercomputer that required more than 200 cabinets. After looking into Cray’s GPU-enabled options, the organization realized it could get the same performance more efficiently using a GPU-accelerated solution.
The conference will feature discussions of some of the most advanced GPU solutions in the world, and usage models will describe how they can be leveraged to support sophisticated computing functionality. As one of the leaders in using GPUs for supercomputing, Cray will have a major presence at the conference, with Cray customers and employees giving various presentations.
Cray customers at GTC
Oak Ridge National Lab: ORNL will showcase how it has used GPUs in the Titan supercomputer to achieve high performance standards in an incredibly efficient system architecture. The paper from the ORNL will be presented on March 19 at 14:00 hours PDT.
The University of Illinois: The Blue Waters supercomputer at the University of Illinois is used to run a highly parallel molecular dynamics code that supports advanced research efforts. This presentation will take place at 17:00 hours on March 19.
Cray at GTC
Cray will also make a few presentations of its own at the GTC event. These include an in-depth look at programming techniques used for GPU-based supercomputers. Three papers will be showcased to detail how different programming techniques can maximize the performance and efficiency of supercomputers built on GPU architectures.
Cray will also present a paper on using the OpenACC specification in the Cray compilation environment at 17:00 hours on March 19. This discussion is titled “OpenACC Support and Extensions in the Cray Compilation Environment.”
Information about using programming to enable better performance in hybrid HPC solutions will also be presented at the conference. The discussion, “Portable and Productive Performance with OpenACC Compilers and Tools,” will take place at 09:00 hours on March 20.
Another Cray presentation will focus on using OpenACC on the Blue Waters Cray XK7 supercomputer. This session, titled as ” Experiences Using OpenACC to port the CCSD(T) Computational Chemistry Method in GAMESS on Blue Waters,” begins at 17:30 hours on March 20.
NVIDIA® GTC will be an exciting opportunity to learn more about how GPUs are being used in supercomputing. To view the full list of Cray-related presentations, follow this link.