Characterizing GPU Overclocking Faults

Eldad Zuberi*, Avishai Wool

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Graphics Processing Units (GPUs) are powerful parallel processors that are becoming common on computers. They are used in many high-performance tasks such as crypto-mining and neural-network training. It is common to overclock a GPU to gain performance, however this practice may introduce calculation faults. In our work, we lay the foundations to exploiting these faults, by characterizing their formation and structure. We find that temperature is a contributing factor to the fault rate, but is not the sole cause. We also find that faults are a byte-wide phenomenon: individual bit-flips are rare. Surprisingly, we find that the vast majority of byte faults are in fact byte-flips: all 8 bits are simultaneously negated. Finally, we find strong evidence that faults are triggered by memory-remnant reads at an alignment of a 32 byte memory transaction size.

Original languageEnglish
Title of host publicationComputer Security – ESORICS 2021 - 26th European Symposium on Research in Computer Security, Proceedings
EditorsElisa Bertino, Haya Shulman, Michael Waidner
PublisherSpringer Science and Business Media Deutschland GmbH
Pages110-130
Number of pages21
ISBN (Print)9783030884178
DOIs
StatePublished - 2021
Event26th European Symposium on Research in Computer Security, ESORICS 2021 - Virtual, Online
Duration: 4 Oct 20218 Oct 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12972 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference26th European Symposium on Research in Computer Security, ESORICS 2021
CityVirtual, Online
Period4/10/218/10/21

Keywords

  • CUDA
  • DFA
  • Fault injection
  • GPU
  • Nvidia
  • Overclocking

Fingerprint

Dive into the research topics of 'Characterizing GPU Overclocking Faults'. Together they form a unique fingerprint.

Cite this