Ever wonder what allows the AI you use to work so efficiently? Or how computers can process data and spit it back to you so quickly?
In this age of digital and technological renaissance, machine learning models, AI technologies and quantum computing, humanity has transformed the way we interact with computational models. One of these ways is with the creation of the GPU (graphical precision unit) computer, used for massive parallel computation and often used to generate AI language learning models.
Now, USC’s Daniel J. Epstein Department of Industrial and Systems Engineering will introduce GPU clusters with a new $700,000 grant from the Air Force Office of Scientific Research, or AFOSR. The money will go towards the purchase of GPU clusters, which will be housed in the new Industrial and Systems Engineering building on campus.
Andres Gomez, an assistant professor at the Daniel J. Epstein Department of Industrial and Systems Engineering, said he was excited to receive the funding.
“We are getting a state-of-the-art GPU cluster. Not a lot of universities have something like this as a part of their infrastructure,” he said. “If we didn’t have it, we’d have to rent the machines, which are expensive, difficult to acquire, and come with constraints.”
The Genius of GPUs
These computer clusters use massive parallel computation, a process that allows for multiple problems to be run at once on a single operating system. Comparing this to traditional CPU computer computation, GPUs allow for training machine-learning tools, and designing and solving problems on larger scales.
Giacomo Nannicini, a USC Viterbi associate professor with a specialty in quantum algorithms and optimization, said that GPUs have become increasingly popular because of their aid in the creation of machine learning systems. The parallel computation capabilities, which GPU servers primarily use, are perfect to run some computations in machine learning.
The new GPU cluster will be housed in the new ISE building on campus — the present site of Salvatori Hall, currently part of the Lord Department of Computer Science. Access to the GPU clusters for learning in related classes and for faculty to conduct important research will be available to both students and faculty alike.
Meisam Razaviyayn, Andrew and Erna Viterbi Early Career Chair and associate professor industrial and systems engineering, said he is thrilled to have access to the new GPU cluster: “Integrating these GPUs into our school’s infrastructure is a transformative step, empowering both our research and students. This will provide crucial hands-on training for our AI students in modern AI and large-scale optimization.”
Published on November 13th, 2023
Last updated on November 13th, 2023