World’s top supercomputer ranks 5th in power efficiency
Summit is eight times more powerful than ORNL’s previous Titan system and is built with 4,608 IBM compute servers, each containing two 22-core IBM Power9 processors and six NVIDIA Tesla V100 graphics processing unit accelerators, interconnected with dual-rail Mellanox EDR 100Gb/s InfiniBand. Summit also has 10 Pbytes of memory paired with fast, high-bandwidth links for efficient data movement. All this provides a peak performance of 215 petaflops compared to Titan which delivered 27 petaflops in 2012, and makes Summit currently the fastest supercomputer.
However, power consumption is the limiting factor for such machines, and the power consumption of 15MW for Summit puts the system at 14.3 Gflops/watt, which ranks it at number 5 in the current Green 500 listing from November 2017. That list is being revised at the moment, so the ranking may fall further as more power efficient systems are evaluated.
The most power efficient supercomputers are all developed by PEZY Computing in Japan. The Shoubu system B installed at RIKEN’s Advanced Center for Computing and Communication in Japan achieved a power efficiency of 17.0 gigaflops/watt, while the Suiren2 cluster at the High Energy Accelerator Research Organization/KEK, also in Japan, reached 16.8 gigaflops/watt while PEZY’s own Sakura system achieved 16.6 gigaflops/watt. All of these top three systems are positioned in the bottom half of the TOP500 performance with Shoubu at 258, Surien2 at 306 and Sakura at 275.
The closest comparison is at number four in the list is the DGX SaturnV Volta system at NVIDIA’s headquarters in San Jose, California which achieved 15.1 gigaflops/watt, using similar GPU technology to Summit and Xeon processors from Intel, but this comes in at number 149 on the TOP500 performance list.
“Summit takes accelerated computing to the next level, with more computing power, more memory, an enormous high-performance file system and fast data paths to tie it all together. That means researchers will be able to get more accurate results faster,” said Jeff Nichols, ORNL associate laboratory director for computing and computational sciences. “Summit’s AI-optimized hardware also gives researchers an incredible platform for analyzing massive datasets and creating intelligent software to accelerate the pace of discovery.”
Summit will be open to select projects this year while ORNL and IBM work through the acceptance process for the machine. In 2019, the bulk of access to the system will go to research teams selected through DOE’s Innovative and Novel Computational Impact on Theory and Experiment, or INCITE, program.
The Mont Blanc project in Europe aims to use a new generation of ARM-based processors for a power efficient, high performance supercomputer.
Related articles:
IBM, DOE unveil ‘world’s fastest’ AI supercomputer
THE TECH BEHIND EUROPE’S €1BN SUPERCOMPUTER
SUPERCOMPUTERS SEE POWER CHALLENGES IN NEW LIST
€1BN TO BUILD EUROPEAN SUPERCOMPUTER WITH NEW LOW POWER PROCESSORS
US LAB TO BUILD ARM-BASED SUPERCOMPUTER
RESEARCHER BUILDS SUPERCOMPUTER WITH RASPBERRY PI BOARDS
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
