Faculty Receives New GPUs for Fluid Dynamics and Machine Learning Research

Scientific computing and simulation require lots of power. This means students, faculty, and researchers at the School of Computational Science and Engineering (CSE) require the latest and greatest in computing technology to sustain their studies.

To help meet this need, CSE Assistant Professor Spencer Bryngelson recently received two NVIDIA A100-80GB graphic processing units (GPU) through NVIDIA’s Academic Hardware Grant Program. Accepted on behalf of the School in April, this gift upgrades the computing arsenal available to CSE faculty and students.

“We do research that relies on this type of hardware,” Bryngelson said. “With it, research happens much more quickly. We can solve bigger problems, we can solve smaller problems more quickly, and all sorts of things that were not available in the past.”

According to NVIDIA, the A100 provides up to 20 times higher performance compared to previous generations and the 80GB model is the world’s fastest memory bandwidth at over two terabytes per second.

Bryngelson and his research group intend to use the GPUs to further their research interests in multiphase fluid dynamics and machine learning.

Bryngelson described one example where models and numerical methods can test advances in ultrasound technology for lithotripsy procedures. These simulations analyze the effects of shock waves applied to kidney stones through computers in lieu of subjecting people to real-world tests. The A100 GPUs can improve accuracy and speed in those computer simulations.

“This type of hardware enables us to train models, to learn what an artificial neural network should look like to predict the things we care about,” Bryngelson said regarding application to machine learning. “A100s are the most powerful tool for the job for this kind of thing.”

Bryngelson’s gifted GPUs also represents Georgia Tech’s continued partnership with commercial providers, like NVIDIA, to equip the Institute with the tools to conduct meaningful research.

NVIDIA is one of many members of the College of Computing’s Corporate Affiliates Program (CAP). CAP is a recruiting pipeline that connects corporate partners with College of Computing students, meeting the industry at the point of need with talent straight from the College.

The School of CSE and NVIDIA certainly are no strangers to each other.

CSE Assistant Professor Srijan Kumar acquired a NVIDIA A100 GPU earlier this year through the Academic Grant Hardware Program. NVIDIA also gifted Kumar a Quadro RTX8000 graphics card in 2021.

In 2021, CSE Associate Professor Polo Chau and Prairie View A&M University Assistant Professor Xishuang Dong partnered with NVIDIA’s Deep Learning Institute to develop a data science teaching kit. The kit teaches students fundamental and advanced topics on accelerated data science, machine learning, data visualization, graph analytics, and more.

In 2019, the School of CSE earned a grant from NVIDIA’s Artificial Intelligence Lab (NVAIL). The NVAIL grant facilitated Chau’s DLI collaborations, as well as gifted the School a NVIDIA DGX station and $100,000 cash award for one year toward research into scalable graph algorithms.

With the addition of Bryngelson’s GPUs to these tools from NVIDIA, the School of CSE is fulfilling its mission of solving difficult problems through interdisciplinary cooperation and external partnerships.

“Georgia Tech and NVIDIA has had a long partnership and we gratefully acknowledge their contributions in our research and publications,” Bryngelson said.