A new NSF $400k grant, "CC* Compute: GPU-based Computation and Data Enabled Research and Education (G-CoDERE) at PSU", has been awarded to an interdisciplinary group of PSU faculty led by Feng Liu (CS) and comprising Christof Teuscher (ECE), Jay L Nadeau (Physics), Bruno Jedynak (Math+Stat), Steve L Reichow (Chemistry), and systems architect William Garrick (OIT Research Computing Manager and Architect).
The funds facilitate the purchase of graphics processing units (GPUs) that support large-scale parallel computing necessary for modern data-driven research and education. The addition of these resources at Portland State could propel scientific and technological advances in computer vision, graphics and deep learning, new computing paradigms, the understanding of complex biological phenomena and advanced microscopy data analysis and acquisition.
“Scientists and engineers are increasingly working with large, computationally-intensive models and datasets. The addition of GPU nodes to Portland State’s existing research computing infrastructure will enhance the university’s capacity for technological innovation and scientific discovery,” said Feng Liu, the principal investigator on the grant and an associate professor of computer science.
“The improvements in research computing will also provide students with training in cutting-edge computing and data analysis techniques that will be valuable when they begin their careers.”
GPUs that excel at large-scale parallel computing are critical for analyzing and visualizing massive quantities of research data and developing data-driven technologies. The parallel processing power of GPUs boosts the performance of regular central processing units (CPUs) by orders of magnitude for many AI and scientific computing tasks.
According to Will Garrick, who manages research computing at PSU, the university’s research computing infrastructure currently includes two high-performance computing clusters that provide many CPU cores, ample memory, and standard data storage. But, there are currently only two GPUs available on these systems. The grant will add a significant number of GPUs and high-performance storage, greatly increasing the computational capability available to the campus community.
A few projects the new research computing infrastructure will support include:
- The development of new technologies capable of creating ultra-realistic computer images and videos and detecting fake visual media known as “deep fakes.”
- The 3-D modeling and study of the behavior of intercellular proteins that choreograph complex biological phenomena such as the contraction of our heart and the detection of sensory data.
- The design and study if next-generation computing models and architectures.
- The evaluation of data generated by holographic microscopes used to search for life in extreme environments here on earth and beyond.
- The application of statistical learning to computer vision to study species and environments; and to computational medicine to study neurodegenerative diseases.
“The addition of GPUs to Portland State’s research computing infrastructure is really exciting for all of us,” said Jay Nadeau, a co-principal investigator on the grant and an associate professor of physics. “This new hardware will reduce the time we spend analyzing computationally intensive datasets. It also provides the opportunity to integrate GPU techniques into the curriculum and opens the door to new collaborations with researchers at OHSU and other institutions.”
In addition to supporting faculty research and student education, PSU’s new research computing infrastructure will be available to researchers at other institutions and industry partners. A recent $1.4M grant awarded to Oregon State University, the University of Oregon, and Portland State will establish a team of data analytics and training professionals with expertise indata integration, data analytics and machine learning at the three universities. They will help facilitate novel approaches to research problems. Combined, these grants will establish Oregon universities’ place as powerhouses of data science in the Pacific Northwest.