Last week Nvidia said that its Tesla Kepler GPU accelerators are powering the world's two most efficient heterorgeneous supercomputers: the "Eurora" system at CINECA, Italy's largest supercomputing center located in Casalecchio di Reno, and the "Aurora Tigon" system at the Salex ES facilities in Chieti, Italy. These top two systems are powered by advanced Eurotech high-performance Aurora servers, equipped with Nvidia Tesla K20 GPU accelerators.
According to Nvidia, "Eurora" delivers 3,208.83 MFlops per watt, making it 2.6 times more energy efficient than the best system using Intel CPUs alone (at Météo France). It also passed "Beacon", the most efficient Intel Xeon Phi accelerator-based system which is located at the National Institute for Computational Sciences, at the University of Tennessee.
The two Tesla-based systems claimed the top two "most energy efficient" spots on the Green500 list of June 2013. The Intel-based "Beacon" came in third followed by "SANAM" (King Abdulaziz City for Science and Technology), "BlueGene/Q" (IBM Thomas J. Watson Research Center), "Cetus" (DOE/SC/Argonne National Laboratory), "CADMOS BG/Q" (Ecole Polytechnique Federale de Lausanne), "BlueGene/Q" (Warsaw) and "BlueGene/Q" (University of Rochester).
"Kepler GPU accelerator-based systems in the top 100 spots on the Green500 list are 50 percent more energy efficient on average than the latest Intel Xeon Phi-based systems," Nvidia said. "By delivering thousands of small energy efficient cores operating in parallel, GPU accelerators are considerably more energy efficient than standard CPUs. These GPU cores are optimized to run the compute-intensive portions of applications, while general-purpose CPU cores, which process in serial, are highly inefficient when running compute-intensive applications."
The "Aurora Tigon" supercomputer, second on the list and located in Italy, delivers 3,179.88 MFlops per watt. Both Tesla-based machines were built by Eurotech, and according to Green500, improves upon the previous greenest supercomputer in the world by nearly 30 percent. The "Beacon" system, third on the list, is based on Intel Xeon Phi 5110P coprocessors while the #4 system, "SANAM", is based on AMD FirePro S10000 GPUs.
"The current fastest supercomputer in the world, Tianhe-2, uses heterogeneous computing elements based on Intel Xeon Phi. It delivers 1.9 gigaflops/watt, which is commensurate with the majority of Xeon Phi systems that are ranked between No. 30 and 35, inclusive," Green500 said.
Issued twice annually, the Green500 list rates the 500 most energy efficient supercomputers based on performance achieved relative to power consumed.
Kevin Parrish is a contributing editor and writer for Tom's Hardware,Tom's Games and Tom's Guide. He's also a graphic artist, CAD operator and network administrator.
See here for all of Kevin's Tom's IT Pro articles.