The Swiss National Supercomputing Centre (CSCS) is rolling out the second half of its Cray XC30 supercomputer, the first to employ both Intel Xeon processors and NVIDA GPUs. With phase one up and running since April, “Piz Daint” is getting a power boost from NVIDIA GPUs, which not only provide CSCS researchers with more computing power – over a petaflops worth – but are also helping to scale back energy requirements.
CSCS Director Thomas Schulthess addressed the upgrade with this official statement:
“Given the ever growing demands of computer models, we can only contain energy consumption in supercomputing with a radical change in computer architecture,” he said.
The GPU-equipped Piz Daint will benefit a wide variety of disciplines, including climate science, geoscience, chemistry, physics, biology, and materials research.
According to Schulthess, the new and improved Piz Daint enables a weather prediction application to run three times faster with seven times less energy compared to its predecessor, the CPU-only “Monte Rosa.” It’s also worth noting that the upgraded Piz Daint is expected to improve energy-efficiency by a factor of three compared to GPU-less version.
CSCS undertook extensive testing before fielding the unit:
In all, the CSCS researchers compared four supercomputer systems, using full simulations codes from chemistry, materials and nanoscience, as well as regional climate simulations over the alpine region. The tests were first run with the original codes on all four machines, and then with newly implemented codes that had been designed especially for efficiency and to run on CPUs as well as graphic processors. The evaluations reveal that not only GPU processors but also improved algorithms influence energy efficiency and performance.
Schulthess discusses the way that GPU computing, which evolved out of the gaming and graphics industry, has been a game-changer for HPC. System designers were at first hesitant to deploy the uber-parallel coprocessors because of the programming challenges entailed by the hybrid architecture. But the promise of more powerful, energy-efficient, high-performance computing systems eventually won out. Over the last decade, research groups foreign and domestic have coalesced around general-purpose GPU computing. In Switzerland, the HP2C project is dedicated to advancing more efficient computing, and hybrid supercomputers – which use a mixture of conventional processors and accelerators – are a central element of the group’s strategy.
Previously code-named “Cascade” and based on the Aries interconnect, the Cray XC30 supercomputer was designed for a wide variety of applications. The hybrid, next-generation, high-performance computing platform accepts a wide variety of processor types, including Intel Xeon processors, Intel Xeon Phi coprocessors, and NVIDIA Tesla GPU accelerators.
In related news, Cray announced yesterday that the XC30 series will support the new Intel Xeon processor E5-2600 v2 product family, formerly code named “Ivy Bridge.” The new 12-core E5-2600 v2 variants are socket-compatible with the 8-core “Sandy Bridge” (Xeon E5-2670) processors that are currently used by XC30 systems like Piz Daint.