Energy Efficiency of Parallel Computations under Voltage Scaling

Volker Strumpen

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

Abstract

Since VLSI technology has reached the power limit, energy efficient computing is of growing interest. Yet there is little consensus about how to link electrical properties of energy and power consumption to algorithmic properties of computations. In this article we study the energy and time complexity of parallel computations as a function of supply voltage. To that end, we derive a voltage scaling model for multicore processors, and show that the asymptotic energy complexity of parallel computations equals their work complexity. We can maximize the savings of energy and time simultaneously by lowering the supply voltage to near- threshold voltage, and show that the maximum savings are bound by the parallelism of the executed algorithm. We conclude that energy savings of parallel computations by voltage scaling can be practically significant despite being asymptotically irrelevant
Original languageEnglish
Title of host publicationAustrochip Workshop on Microelectronics
Editors Gerald Hilber, Timm Ostermann, Andreas Rauchenecker
Place of PublicationLinz
PublisherInstitut für Integrierte Schaltungen
Pages41-46
Number of pages6
Publication statusPublished - 2013

Fields of science

  • 102 Computer Sciences
  • 101015 Operations research
  • 102003 Image processing
  • 102023 Supercomputing
  • 102007 Computer animation
  • 102009 Computer simulation
  • 102026 Virtual reality
  • 102025 Distributed systems
  • 202031 Network engineering

JKU Focus areas

  • Computation in Informatics and Mathematics

Cite this