Information and Technology

Game the System with GPUs

Story posted November 15, 2011

Any gamer knows the importance of maximizing computational power, particularly when it comes to graphics. Who wants to wait for the scene to change when you're flying to your next quest? Gamers invest in the best video cards, also known as graphics cards. Inside these graphics cards are special processors optimized for 3D graphics; working together to handle the mathematical calculations required to provide a  seamless, stutter free gaming experience. 

These processors, referred to as GPUs (Graphics Processing Units), have been compared to mathematical laborers, many units honed to perform the same type of calculation at rapid speeds. CPUs (Central Processing Units), on the other hand, are more like executives, decision makers who can perform many different types of calculations, at slower speeds. Research problems can require an enormous amount of computational power to build a model or analyze vast amounts of data. The Bowdoin Computing Grid, a set of mulit-processor Linux computers, supports faculty, students and staff whose work requires high performance computing power serving people representing 9 academic departments on campus.  

This fall an additional node was added to the Bowdoin Computing Grid. The new GPU added the ability to calculate hundreds of times faster than the existing CPUs. The GPU is made up of 2 Nvidia GeForce GTX 580 video cards, each having 3 GB of RAM and 1024 GPU processing units. This prototype GPU node can be leveraged by MatLab and custom code to process calculations at rocket speeds. In addition to the GPU, we have added 2 very powerful regular Grid nodes each having 128 GB of RAM and 24 CPU cores.   

If you have a computer job that takes more than a few hours to run, consider taking advantage of the Bowdoin Computing Grid. Learn more at: http://www.bowdoin.edu/it/what-we-offer/linux-computational-computing.shtml or by contacting Dj Merrill at deej@bowdoin.edu.