Thanks to Frontier, the Department of Energy’s first $600 million exascale supercomputer — which is located at Oak Ridge National Laboratory — the United States ranks first in the Top500 supercomputer list for the number of calculations performed per second. can be performed. But Jack Dongarra noted that China, which has two supercomputers in sixth and ninth place on the list, could overtake the United States’ computing capacity for solving scientific and technological problems.
Dongarra, who holds positions at the University of Tennessee, ORNL, and the University of Manchester in the United Kingdom, is this year’s winner of the prestigious AM Turing Award from the Association of Computing Machinery, the equivalent of the Nobel Prize in computer science. It has a price of $1 million.
Dongarra recently spoke with Friends of ORNL (FORNL).
In 1993, he explained, he provided a software standard or benchmark for evaluating the relative performance of the world’s top 500 supercomputers twice a year by having them solve a problem of solving linear equations. He and two others manage the Top500 list.
Dongarra told FORNL that Border performed calculations at a rate of 1.1 exaflops, or 1.1 trillion calculations per second — that is, a billion times a billion floating point operations per second, or FLOPS (e.g., adding and multiplying numbers with decimal points). Frontier is five times faster than the most powerful supercomputers in use today.
To help his audience understand the power of Frontier, which takes up the space of two tennis courts, Dongarra suggested we imagine the UT has 60 Neyland stadiums, each with 100,000 filled seats. To do the number of calculations per second you can get on Frontier, you have to give each person a laptop with a capacity of 166 billion FLOPS and connect the laptops in all stadiums.
China
Dongarra has not had any contact with Chinese computer scientists for the past two years, but he has heard rumors that China may have built two exascale computers, but chooses to keep his achievement a secret. He said Taiwan Semiconductor Manufacturing Co. provides US and Chinese companies with state-of-the-art semiconductor fabrication for supercomputers. Taiwan is an independent democracy that has a strong partnership with the US and opposes being part of communist China.
Dongarra noted that China is the world leader with 173 supercomputers used for science. The US comes in second with 126 supercomputers. One of the American supercomputers is Summit at ORNL, which once ranked first and now fourth in the Top500. Dongarra said it will be dismantled when it is five years old because maintenance costs will be too high.
Costs
The US Department of Energy’s Exascale Computing program will cost taxpayers $3.6 billion over seven years. Next year, in addition to Frontier at ORNL, two more $600 million DOE exascale supercomputers — Aurora at Argonne National Laboratory and El Capitan at Lawrence Livermore National Laboratory — should be up and running, with 21 science applications.
These exascale supercomputers will test mathematical models that simulate complex physical phenomena or designs, such as climate and weather, evolution of the cosmos, small modular reactors, new chemical compounds that can be used in vaccines or therapeutic drugs, power grids, wind turbines and combustion engines.
While fast calculations can be made using cloud computing, Dongarra said large, powerful supercomputers are needed to provide “better reliability and more accuracy in our calculations” and to get “a three-dimensional, completely realistic implementation of what scientists are trying to achieve.” modeling, as well as better resolution because they discover at a deeper level what is going on on a finer scale.”
The equations that are solved are based on the laws of physics and most supercomputers are programmed with the languages Fortran and C++.
Another goal of larger supercomputers, he added, is to optimize “a model” of, say, an internal combustion engine by “running thousands of different models with modified parameters” to identify an engine design that uses fuel with maximum efficiency and minimal emissions of pollutants .
Dongarra said computational simulations are the third pillar of science after theory and experiments. He noted that researchers are using computer models to answer some questions because, for example, it’s too difficult to build large wind tunnels and too dangerous to try a new chemical on humans to see if it’s an effective drug. .
We’d have to wait too long to find out how much the climate would change if we doubled our fossil fuel burn, he said.
Frontier is built by Hewlett Packard Enterprises (HPE) using an interconnect from Cray (which HPE recently purchased), and nearly eight million computer processing units (CPUs) and graphics processing units (GPUs) made by Advanced Micro Devices (AMD) to be coordinated. GPUs, which are used for video games and video editing and which have more transistors than CPUs, perform complex mathematical and geometric calculations necessary for graphics rendering.
“The GPUs provide 98% of Frontier’s performance capabilities,” said Dongarra. He added that the computational speed could be quadrupled if the size of the numbers with decimals were represented in a compressed manner (for example, in the case of pi, from 3.14159265 to 3.14). That could boost Frontier’s peak performance from 2 exaflops for modeling and simulation to 11.2 exaflops.
To explain the performance of supercomputers, Dongarra used a car analogy.
“If you’re driving your car, your speedometer may say you can go up to 160 miles per hour. That’s your car’s theoretical potential. But the highest speed you can safely reach on the road is 120 mph. That’s what we’re trying to measure: the achievable speed of execution in supercomputer performance,” he said.
In his summary, Dongarra said that supercomputer hardware is constantly changing, so programmers must continue to develop algorithms and design software that fits the capabilities of the hardware. He added that a major revolution in high-performance computing will be the imminent increase in the use of artificial intelligence (AI), the ability of a computer program or machine to learn and think like humans by finding patterns in a vast array. database.
One example is the supercomputers behind weather forecasts, Dongarra said.
“Weather forecasting starts with creating a three-dimensional model of today’s weather. AI learns by sifting through all-weather data from the past 100 years to gain insights. Based on knowledge of those conditions and of physical laws, it can accurately predict what the weather will be like in the coming days,” he said.
At ORNL, Summit has already used AI for some of its simulations.
The future is here.