New US supercomputer is fastest on Earth

June 29, 2018
by Mark Ollig

Since 2016, China has been ranked the world’s leader in supercomputer processing speed.

China’s Sunway Taihulight supercomputer processes data at 125 petaflops, or 125 quadrillion (125,000 trillion) floating-point operations (calculations) per second.

One quadrillion is a 1 with 15 zeros (yes, I looked it up).

This month, the US Department of Energy’s Oak Ridge National Laboratory (ORNL) publicly announced its newest scientific supercomputer, called Summit.

The Summit supercomputer is an AC922 (Accelerated Compute Server 922) system, built by IBM.

The AC922 supports cutting-edge artificial intelligence and data-intensive applications.

ORNL also announced Summit has taken over the number-one position in global supercomputing processing speed, as it can process data an incredible 200 quadrillion floating-point operations per second (200 petaflops).

Summit’s 200 petaflops of processing power allows it to analyze massive amounts of data – quickly.

It can be mind-bending, trying to understand this kind of computer processing speed. I read, Summit processes data a million times faster than the best laptop computer.

This processing ability will be used for energy research, and to create advanced new materials, including quantum substances.

New fission, fusion, and bioenergy; artificial intelligence innovations, and what had been considered improbable scientific discoveries will become possible using Summit.

Physically, Summit is made up of many individual rows of computer cabinet bays, containing shelves filled with individually-wired computing cards.

The cabinet bay backplanes are linked together, using fiber-optic and electrical cabling.

Approximately 4,000 gallons of water per minute is circulated among the cabinet bays, via an insulated-tube cooling system, to keep the electronic computing components from overheating.

For those wanting some in-depth computing details, the following description of the Summit supercomputer is from the US government’s ORNL website.

“Summit consists of 4,608 computer servers, each containing two 22-core IBM Power9 processors and six NVIDIA Tesla V100 graphics processing unit accelerators. Summit is interconnected with a dual-rail Mellanox EDR InfiniBand network, which provides, overall, 200 gigabits per second throughput to each computer server, and runs a standard Linux operating system from Red Hat in Red Hat Enterprise Linux. The system also possesses more than 10 petabytes of memory paired with high-bandwidth pathways for efficient data movement.”

The Summit supercomputer has some nice hardware and software, huh, folks?

Indeed, we have come a long way from the days of those early 1980s IBM personal computers, with their disk operating system (DOS), 256kB of random access memory (or 512kB if one added another chip as I did), 20 MB hard drive, 4.77 MHz Intel processor, color graphics array, cathode ray tube monitor, floppy disks, and full mechanical keyboard.

Ah, nostalgia.

But, I digress; back to the Summit supercomputer.

“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,” said Jeff Nichols, associate laboratory director for computing and computational sciences at ORNL.

Computing processing firsts at the Oak Ridge National Laboratory include:

first gigaflops calculations in 1988;

first teraflops calculations in 1998;

first petaflops calculations in 2008; and

first exaflops calculations in 2018.

FLOPS are floating-point operations per second. It’s a measurement of a computer’s performance; I tend to think of it as analogous to the horsepower in a car, if you will.

One exaflop equals one quintillion calculations per second. That’s a 1 followed by 18 zeros.

The first exascale operations per second (exaops; sometimes seen as an exaflop) computing calculation barrier was reached this year, when the Summit supercomputer obtained a staggering 1.8 quintillion calculations in one second, while analyzing millions of genome organism DNA sets as part of its health and human research.

The Summit supercomputer can obtain processing speed within the exascale range only for specific scientific applications.

The Department of Energy’s Exascale Computing Project (ECP) is researching the possible uses for exascale computing.

“The mission of the US Department of Energy’s Exascale Computing Project is to accelerate delivery of a capable exascale computing ecosystem that will provide breakthrough solutions to address America’s most critical challenges in scientific discovery, energy assurance, economic competitiveness, and national security,” per the statement on the ECP’s webpage, located at https://www.exascaleproject.org.

Exascale computing will also be used for health and environmental research, creating new superconducting materials, space exploration, and building better aircraft.

I was surprised to learn the Summit supercomputer was able to reach a maximum processing speed of 3.3 exaops, which is a significant step toward achieving full exascale computing in the very near future.

In fact, the ECP’s goal is to have exascale supercomputers processing data by 2021.

Take a visual walk through the rows of the Summit supercomputer bays at https://bit.ly/2K9Zgbw.

Other videos of the Oak Ridge National Laboratory Summit supercomputer are available on the ORNL Dropbox at https://bit.ly/2tFNN8p.

Be sure to visit the Bits & Bytes weblog at https://bitscolumn.blogspot.com.

Advertise in over
250+ MN newspapers