Sleepless in Salt Lake: Tesla P100 Powers Winning Team in Student Cluster Competition

by Roy Kim

Simple lesson: If you want to win the granddaddy of student supercomputing competitions, sleep is optional. So, occasionally, is power. Our NVIDIA Tesla Accelerated Computing Platform isn’t.

Thanks to a system design that put more NVIDIA GPUs into play than any of its competitors, a team from the University of Science and Technology of China (USTC) grabbed the top spot Thursday in the annual Student Cluster Competition at the SC16 supercomputing show in Salt Lake City, Utah.

The team also generated the competition’s top Linpack score — 31.15 teraflops — a new record and more than triple the competition’s previous top mark. And its efficiency would have ranked it among the top machines on the annual Green500 list of most efficient supercomputers.

The winning team. Image credit: SC16.
The winning team. Image credit: SC16.

The victory for Team SwanGeese, from one of China’s top research universities, is just the latest for Tesla in a major international student supercomputing competition.

USTC’s six-member team relied on a system powered by 16 NVIDIA GPUs.

The team was one of 14 finalists from around the globe — nine powered by NVIDIA Tesla GPUs — to make the cut to compete in the finals.

Power Outage

First held at SC07 in Reno, Nevada, the annual Supercomputing Cluster Competition has inspired other student competitions in China, Germany and South Africa.

But the one held every year at the annual SC conference is different. Teams build the most powerful cluster possible under the competition’s 3,120-watt (26-amp) power limit and then work for 48 hours straight to run them through various real-world applications.

As if it wasn’t challenging enough to work through two nights — even after the show floor was closed to the conference’s 14,000 other attendees — the competition’s organizers cut the power, briefly, Tuesday, to test how well the team’s could recover from an unscheduled outage.

The loss of electricity caused weary laughter, groans and cries of “oh no” to resound throughout the competition’s corner of the show floor, which was littered with almost as many empty pizza boxes as green NVIDIA logos.

Sleep Impact: Where Tesla Meets Tired

Two members of USTC's team during their winning run.
Two members of USTC’s team during their winning run.

By the last day of the competition, the students on each of the teams were noticeably weary. Some rubbed their eyes. Others gnawed on pizza crusts and sipped coffee to stay alert.

“I haven’t gone back to my hotel since yesterday,” admitted one competitor from Taiwan’s National Tsing Hua University as he talked about his team’s system, which was powered by eight Tesla P100 GPUs.

But while all these teams proved they could do without sleep, most couldn’t do without NVIDIA GPUs.

Asked for the biggest lesson he would take from this competition, Mingyao Shen, from China’s Huazhong University of Science and Technology, had a simple answer: “If we had more GPUs, we could have done better,” he said with a laugh.

Learn more about Team SwanGeese’s winning entry.