Ian Lee: Making the Most of your Hardware - Micro-Benchmarks on NVIDIA GPUs

Student's Name: 
Ian Lee
Advisor's Name: 
Jose Renau
Home University: 
University of Connecticut
PDF icon Lee_poster.pdf6.06 MB
Microsoft Office document icon Lee_Report.doc46 KB

National Science Foundation: Nugget

Ian Lee

Ian Lee worked in the Micro Architecture Santa Cruz (MASC) Research Lab under the advisement of Professor Jose Renau. A rising Senior at the University of Connecticut, Ian will be pursuing a PhD in Computer Engineering upon completing his Bachelors of Science in Engineering. The work performed required familiarizing himself with the CUDA device language, and refining each algorithm in order to make the most of the hardware that he was working with.

Mr. Lee’s work dealt with the use of NVIDIA GPUs (graphics cards) as an alternate means of computing data. The use of the GPUs allows for parallel computations, resulting in marked performance increases with respect to modern CPUs. Mr. Lee’s research focused primarily on benchmarking and development of standards to define the performances of these GPUs. Four benchmarks were developed in an attempt to target specific performance areas of the cards, such as transfer raters, and overhead. These benchmarks are substantial as there are currently no performance benchmarks available for GPUs. His research is ongoing, as he will be continuing his collaboration with the MASC lab in the fall upon his return to the University of Connecticut.