Cray has shown off its fastest supercomputer yet, which puts a petaflop of computing power in a single cabinet. The company also says it is adding machine-learning capabilities across its systems.
The XC50 supercomputer is designed for the most demanding high-performance computing (HPC) users, according to Cray. It said supercomputing applications are evolving to include more deep-learning algorithms, and as a result the uses of the GPUs in its systems are increasingly "enabling our customers to use new analytics techniques to gain insight from increasingly large and complex data".
The company said the new machine provides the "highest performance density" of any Cray supercomputer and will allow customers to take on larger, more complex workloads.
The Swiss National Supercomputing Centre (CSCS) in Lugano is currently upgrading its Cray XC30 supercomputer, nicknamed Piz Daint, to a Cray XC50 system and will combine it with the Centre's Cray XC40 supercomputer.
"Our new Cray XC50 supercomputer will significantly accelerate our computational research capabilities allowing our users to perform more advanced, data-intensive simulations, visualizations and data analyses across a wide array of scientific studies," said professor Thomas Shulthess, director of the CSCS.
"Cray's next-generation supercomputer and its continued integration of GPU acceleration has created a powerful and efficient hybrid multi-core system for addressing our current and future HPC workloads."
Features of the Cray XC50 supercomputer include the Aries network interconnect, which is designed for datacenter GPU-accelerated applications, where high node-to-node communication performance is critical, "innovative" cooling systems to lower customers' total cost of ownership, and support for the NVIDIA Tesla P100 GPU accelerator as well as for next-generation Intel Xeon and Intel Xeon Phi processors.
Cray also said it is adding deep-learning capabilities across its supercomputers and cluster systems.
"The convergence of supercomputing and big-data analytics is happening now, and the rise of deep-learning algorithms is evidence of how customers are increasingly using high-performance computing techniques to accelerate analytics applications," said Steve Scott, senior vice president and chief technology officer at Cray.
"Training problems look very much like classical supercomputing problems."
Marine geophysical company PGS is running machine-learning algorithms on its Cray XC40 supercomputer, nicknamed Abel.
Machine-learning technologies such as regularization and steering can be applied to a significant computational problem in seismic exploration, full waveform inversion or FWI, which is being used to find a high-resolution, high-fidelity representation of the subsurface in the ultra-deep Gulf of Mexico.
"This class of problems is notoriously hard," said Dr Sverre Brandsberg-Dahl, global chief geophysicist for Imaging and Engineering, at PGS.
"It is a multidimensional ill-posed optimization problem that is far from automated and requires lots of skilled resources' intervention, sometimes more art than science in many cases," he said.
"Our Cray XC40 system was able to learn how to best steer refracted and diving waves for deep model updates and how best to reproduce the sharp salt boundaries in the Gulf of Mexico. Machine learning at scale on our Cray supercomputer showed dramatic improvement in the quality of the inversion process as compared with current state-of-the-art FWI."