Argonne National Laboratory

GPGPUs: How to Combine High Computational Power with High Reliability

TitleGPGPUs: How to Combine High Computational Power with High Reliability
Publication TypeConference Paper
Year of Publication2014
AuthorsGomez, LABautist, Cappello, F, Carro, L, Debardeleben, NA, Fang, B, Gurumurthi, S, Pattabiraman, K, Rech, R, M. Reorda, S
Conference Name2014 Design Automation and Test in Europe Conference and Exhibition
Conference LocationDresden, Germany
Other NumbersANL/MCS-P5068-0114
AbstractGPGPUs are increasingly used in several domains, from gaming to different kinds of computationally intensive applications. In some cases, their reliability is becoming a serious issue and several research activities are focusing on its evaluation. This paper aims at overviewing some major results in the area. First, it shows and analyzes the results of some experiments aiming at assessing the GPGPU reliability in HPC datacenters. Secondly, it provides recent results about the reliability of some GPGPUs, derived from radiation experiments. Finally, it describes the characteristics of an advanced fault injection environment allowing one to effectively evaluate the resiliency of applications running on GPGPUs.