Watch all the Transform 2020 sessions on-demand here.
TOP500 released an update to its list of the fastest supercomputers in the world, with the U.S. Department of Energy’s Oak Ridge National Laboratory leading the way. In its debut earlier this month, Summit clocked in at 122 petaflops of compute power on High Performance Linpack (HPL), a benchmark used to rank supercomputers ranked on the TOP500 list.
Summit uses more than 27,000 Nvidia graphics processing unit chips (GPU), and five of the seven fastest supercomputers in the world utilize Nvidia GPUs — like the Tesla V100, which first made its debut in May 2017. Summit has already been used to do things like apply machine learning in the search for genetic links between diseases or explore materials that can be used for superconductors.
“When we first started talking about the original Tesla K80 back in 2015, we were only contributing about 11 percent of the list that year, if I add up all the computational horsepower on the top of the list,” Nvidia VP Ian Buck told VentureBeat. “This year, the majority of 56 percent of the computation on the list is coming from GPUs, and this really talks to the adoption of accelerated computing, of using GPUs for solving the kinds of problems and building the kinds of systems that are necessary to advance computing.”
Also new to the list is Sierra. Housed at Lawrence Livermore National Laboratory, Sierra is now ranked the world’s third-fastest supercomputer, with 71 petaflops of compute power.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
Both Summit and Sierra were built by IBM and include IBM Power9 CPUs.
The TOP500 updates its ranking of top supercomputers every six months.
The new rankings were announced today at the International Supercomputing Conference being held this week in Frankfurt, Germany.
Also announced today, Nvidia released nine new GPU Cloud computing containers to make it easier to work with deep learning frameworks.
The United States regains the title of owning the word’s fastest supercomputer after years of Chinese dominance.
China’s Sunway TaihuLight supercomputer and its 93 petaflops of compute power are powered by the Sunway SW26010 processor. Based at the National Supercomputing Center in Wuxi, China, the supercomputer had been the top ranked supercomputer in the world for the past two years.
Titan, which also uses Nvidia GPUs at Oak Ridge National Laboratory, comes in at seventh-fastest in the world. Titan was ranked the fastest supercomputer in the November 2012 edition of the TOP500, but that title belonged to Chinese supercomputers from 2013 until now — including Tianhe-2, a supercomputer developed by China’s National University of Defense Technology.
For the first time ever, in TOP500 rankings last fall, China surpassed the United States in total number of ranked supercomputers, 202 to 143.
That trend continued in the latest report, with the number of ranked U.S. supercomputers falling to 126 as China’s total climbs to 206.
China and the United States are followed in the largest number of ranked supercomputers by 36 systems in Japan, 22 in the United Kingdom, and 21 in Germany.
Many of the world’s supercomputers are being used by governments to do things like carry out high performance computing and train deep AI models with massive amounts of data. For example, the world’s fifth-fastest supercomputer and fastest in Japan is named Artificial Intelligence Bridging Cloud (AIBC).
Updated 3:35 a.m. to include additional information from TOP500 about top ranked nations.
