GTC Asia’s second and final day drew to a rousing close Thursday in Bejing, with more than 40 workshops and lectures, and a closing keynote by Steve Scott, chief technologist for NVIDIA’s HPC business.
With more than 1,500 attendees from 40 countries, the event’s closing session focused on a dizzying range of topics, including modeling techniques for space exploration, monsoons and oil-reserve exploration. Each of these disciplines now uses heterogeneous computing to accelerate massively complicating computing tasks.
In a morning lecture, Dan Negrut, of University of Wisconsin, showcased his team’s work on granular terrain simulation. His team has used the GPU to more accurately represent terrain issues encountered by the Mars Rover, resulting in a 10x acceleration. Among the examples he showed were demonstrations involving the manipulation of thousands of M&Ms and golf balls, eachl individually manipulated by the GPU.
In a separate track, Anthony Lichnewsky, software architect at oil-giant Schlumberger, described the sobering race against time we’re facing to find more oil reserves. He noted that most of the world’s “easily-found” reserves have already been tapped, leaving only reserves in challenging spots characterized by deep water and harsh climates.
The previous CPU-based computing solutions his firm relied on can’t compute fast enough, getting hung up on compute-intensive ‘kernel’ work that chews up precious hours. Lichnewsky’s team has turned to GPU-enabled solutions to help explorers find oil more efficiently.
In the closing keynote, Scott, who came to NVIDIA earlier this year from Cray, where he had served as chief technology officer, focused on the future of high-performance computing, honing in on exaflop computing and power efficiency.
He reiterated a common theme over this week’s event, arguing that we will be severely constrained by power if we continue to follow the same computing path of decades past. Citing the many applications and industries that have recently adopted GPU computing, and the coming introduction of NVIDIA’s Kepler (2012) and Maxwell (2014) GPU architecture, Scott expressed his excitement for the HPC industry.
The closing keynote also honored 14 Chinese university students for their work in areas ranging from astronomy to video processing, within a CUDA coding contest, sponsored by NVIDIA. The 14 students received awards and plaques, including a first-place cash prize of RMB20,000.
Scott also presented Shanghai Jiao Tong University with a CUDA Center of Excellence award, recognizing its groundbreaking work using GPUs and CUDA and its plans to build the fastest supercomputer of any university in China.