Faculty Sponsor
Nicholas Rosasco
College
College of Arts & Sciences (CAS)
Department/Program
Commputing And Information Sciances
Presentation Type
Poster Presentation
Symposium Date
Summer 7-23-2025
Abstract
GPUs are used for both recreational purposes and extremely important simulations for things like earthquake prediction in addition to mathematical modeling. In all cases it is critical that the GPUs used are chosen effectively, as these tend to be very expensive components. By benchmarking a variety of GPUs running on GNU/Linux with a roster of different benchmarks, this project provides information of the capabilities of GPUs running with GNU/Linux, which is common in research settings. This is compared to the expected performance levels of each GPU as stated in selected online sources and manufacturer claims, which are typically geared towards Microsoft Windows environments. The GPUs tested on linux devices included mostly NVIDIA GPUs, as they are the current dominant maker. The benchmarks used in the first phase were 3D renderings and focused on measuring clock speed of each GPU. In a secondary round of tests GPU performance was scored with Large Language Models (LLM) tools. The principal tool used was Meta’s PyTorch suite; the Time to First Token (TTFT) as well as the Token per Second (TPS) were chosen as metrics. The findings of the 3D rendering benchmarks did stay consistent with expected rankings of each GPU. Additionally the recorded Clock speed of each GPU was not reduced from the GNU/Linux. The limitations of this design were caused by each GPU being in its own slightly different computer and each benchmark not sharing its exact methods and internals.
Recommended Citation
Haines, Nathanael W., "GPU Performance in a Linux Enviroment" (2025). Summer Interdisciplinary Research Symposium. 262.
https://scholar.valpo.edu/sires/262

Biographical Information about Author(s)
Nathanael was only able to attend Valparaiso University thanks to the scholarship granted by EPIC from the National Science Foundation. He initially became interested in Computer science from learning Minecraft command blocks which is a type of sudo code that can change the parts of the game. That passion was able to bring him all the way to choosing his major.