Exploiting process variability in voltage/frequency control

Sebastian Herbert, Siddharth Garg, Diana Marculescu

Research output: Contribution to journalArticle

Abstract

Fine-grained dynamic voltage/frequency scaling (DVFS) is an important tool in managing the balance between power and performance in chip-multiprocessors. Although manufacturing process variations are giving rise to significant core-to-core variations in power and performance, traditional DVFS controllers are unaware of these variations. Exploiting the different power profiles of the cores can significantly improve energy efficiency. Process variations do not significantly affect dynamic power, so less-leaky processing units are more energy-efficient than their leakier counterparts at a given supply voltage and frequency. Taking advantage of this observation, three existing DVFS control algorithms are modified to shift work from inefficient, leaky processing units to efficient, less leaky ones, maintaining performance while reducing total power consumption. This work-shifting is carried out both between dies in a given speed bin and between voltage/frequency islands on a given die. The gains enabled by incorporating variability-awareness into the three DVFS algorithms are demonstrated on both multithreaded and multiprogrammed workloads. For a baseline 16-core design with per-core voltage/frequency islands (VFIs) and a 4×4 mesh on-chip network, the aggregate power per squared throughput (power/throughput 2 or P/T 2) over all fabricated dies is reduced by 9.2%, 5.7%, and 7.7% for the three controllers. Chip multiprocessor designs using other VFI granularities and network topologies are also examined.

Original languageEnglish (US)
Article number5983425
Pages (from-to)1392-1404
Number of pages13
JournalIEEE Transactions on Very Large Scale Integration (VLSI) Systems
Volume20
Issue number8
DOIs
StatePublished - 2012

Fingerprint

Electric potential
Throughput
Controllers
Bins
Processing
Energy efficiency
Electric power utilization
Topology

Keywords

  • Chip-multiprocessor
  • dynamic voltage/frequency scaling (DVFS)
  • process variations

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Hardware and Architecture
  • Software

Cite this

Exploiting process variability in voltage/frequency control. / Herbert, Sebastian; Garg, Siddharth; Marculescu, Diana.

In: IEEE Transactions on Very Large Scale Integration (VLSI) Systems, Vol. 20, No. 8, 5983425, 2012, p. 1392-1404.

Research output: Contribution to journalArticle

@article{6fd5d4a958cb4343b4743c8170775513,
title = "Exploiting process variability in voltage/frequency control",
abstract = "Fine-grained dynamic voltage/frequency scaling (DVFS) is an important tool in managing the balance between power and performance in chip-multiprocessors. Although manufacturing process variations are giving rise to significant core-to-core variations in power and performance, traditional DVFS controllers are unaware of these variations. Exploiting the different power profiles of the cores can significantly improve energy efficiency. Process variations do not significantly affect dynamic power, so less-leaky processing units are more energy-efficient than their leakier counterparts at a given supply voltage and frequency. Taking advantage of this observation, three existing DVFS control algorithms are modified to shift work from inefficient, leaky processing units to efficient, less leaky ones, maintaining performance while reducing total power consumption. This work-shifting is carried out both between dies in a given speed bin and between voltage/frequency islands on a given die. The gains enabled by incorporating variability-awareness into the three DVFS algorithms are demonstrated on both multithreaded and multiprogrammed workloads. For a baseline 16-core design with per-core voltage/frequency islands (VFIs) and a 4×4 mesh on-chip network, the aggregate power per squared throughput (power/throughput 2 or P/T 2) over all fabricated dies is reduced by 9.2{\%}, 5.7{\%}, and 7.7{\%} for the three controllers. Chip multiprocessor designs using other VFI granularities and network topologies are also examined.",
keywords = "Chip-multiprocessor, dynamic voltage/frequency scaling (DVFS), process variations",
author = "Sebastian Herbert and Siddharth Garg and Diana Marculescu",
year = "2012",
doi = "10.1109/TVLSI.2011.2160001",
language = "English (US)",
volume = "20",
pages = "1392--1404",
journal = "IEEE Transactions on Very Large Scale Integration (VLSI) Systems",
issn = "1063-8210",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "8",

}

TY - JOUR

T1 - Exploiting process variability in voltage/frequency control

AU - Herbert, Sebastian

AU - Garg, Siddharth

AU - Marculescu, Diana

PY - 2012

Y1 - 2012

N2 - Fine-grained dynamic voltage/frequency scaling (DVFS) is an important tool in managing the balance between power and performance in chip-multiprocessors. Although manufacturing process variations are giving rise to significant core-to-core variations in power and performance, traditional DVFS controllers are unaware of these variations. Exploiting the different power profiles of the cores can significantly improve energy efficiency. Process variations do not significantly affect dynamic power, so less-leaky processing units are more energy-efficient than their leakier counterparts at a given supply voltage and frequency. Taking advantage of this observation, three existing DVFS control algorithms are modified to shift work from inefficient, leaky processing units to efficient, less leaky ones, maintaining performance while reducing total power consumption. This work-shifting is carried out both between dies in a given speed bin and between voltage/frequency islands on a given die. The gains enabled by incorporating variability-awareness into the three DVFS algorithms are demonstrated on both multithreaded and multiprogrammed workloads. For a baseline 16-core design with per-core voltage/frequency islands (VFIs) and a 4×4 mesh on-chip network, the aggregate power per squared throughput (power/throughput 2 or P/T 2) over all fabricated dies is reduced by 9.2%, 5.7%, and 7.7% for the three controllers. Chip multiprocessor designs using other VFI granularities and network topologies are also examined.

AB - Fine-grained dynamic voltage/frequency scaling (DVFS) is an important tool in managing the balance between power and performance in chip-multiprocessors. Although manufacturing process variations are giving rise to significant core-to-core variations in power and performance, traditional DVFS controllers are unaware of these variations. Exploiting the different power profiles of the cores can significantly improve energy efficiency. Process variations do not significantly affect dynamic power, so less-leaky processing units are more energy-efficient than their leakier counterparts at a given supply voltage and frequency. Taking advantage of this observation, three existing DVFS control algorithms are modified to shift work from inefficient, leaky processing units to efficient, less leaky ones, maintaining performance while reducing total power consumption. This work-shifting is carried out both between dies in a given speed bin and between voltage/frequency islands on a given die. The gains enabled by incorporating variability-awareness into the three DVFS algorithms are demonstrated on both multithreaded and multiprogrammed workloads. For a baseline 16-core design with per-core voltage/frequency islands (VFIs) and a 4×4 mesh on-chip network, the aggregate power per squared throughput (power/throughput 2 or P/T 2) over all fabricated dies is reduced by 9.2%, 5.7%, and 7.7% for the three controllers. Chip multiprocessor designs using other VFI granularities and network topologies are also examined.

KW - Chip-multiprocessor

KW - dynamic voltage/frequency scaling (DVFS)

KW - process variations

UR - http://www.scopus.com/inward/record.url?scp=84862686667&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84862686667&partnerID=8YFLogxK

U2 - 10.1109/TVLSI.2011.2160001

DO - 10.1109/TVLSI.2011.2160001

M3 - Article

VL - 20

SP - 1392

EP - 1404

JO - IEEE Transactions on Very Large Scale Integration (VLSI) Systems

JF - IEEE Transactions on Very Large Scale Integration (VLSI) Systems

SN - 1063-8210

IS - 8

M1 - 5983425

ER -