As a Chilean scientist living abroad, I’m always thinking about how my work can be relevant in Latin America. Research funds can be scarce, even if demand for consumer products such as personal computers and mobile phones is surging there.
So the idea that consumer technology can be used to drive scientific research—one long championed by NVIDIA Chief Executive Jen-Hsun Huang—appeals to me. Even in the United States, scientists have a large unfulfilled demand for compute time for research. It’s even bigger in middle-income countries.
The price-performance and power-performance ratios of GPU clusters can help close this gap, making it possible for scientists to own and operate powerful research systems, which they couldn’t do before. That’s why GPU computing is a source of much excitement among Latin American researchers.
Right now, I’m back in Chile organizing an advanced institute on the simulation of tsunamis and storm surges, a topic of particular relevance for the country. One of my co-organizers here recently got a research grant from the Chilean government to develop and build a database of pre-modeled tsunami scenarios using GPU computing. Meanwhile, the host university just completed its first year as a CUDA Teaching Center, and at least five graduate students are doing their theses using GPUs.
In other parts of Latin America, GPU computing has also made strides. Brazil always does things in a big way, and the first GPU Computing Developer Forum there last July attracted more than 1,500 students, researchers and developers from industry. Brazil is also home to the first CUDA Center of Excellence in the region, at Rio de Janeiro’s Universidad Federal Fluminense.
My colleagues report similar enthusiasm for GPU computing elsewhere in the region. “I was amazed with the large crowds showing up,” CUDA Fellow Manuel Ujaldón told me while on a worldwide CUDA teaching tour this year. “After giving about 20 seminars this year—including in Europe, Africa, Oceania—I can attest to unequaled enthusiasm in Latin America.”
That kind of enthusiasm shows GPU computing offers more than just greater efficiency for scientists struggling for time on the world’s most advanced supercomputers. GPU computing is widening access to computational science for researchers in Latin America and other regions.