A workload-aware neuromorphic controller for dynamic power and thermal management

Saurabh Sinha, Jounghyuk Suh, Bertan Bakkaloglu, Yu Cao

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

A workload-aware low-power neuromorphic controller for dynamic power and thermal management in VLSI systems is presented. The neuromorphic controller predicts future workload and temperature values based on the past values and preemptively regulates supply voltage and frequency. Our specific contributions include: (1) implementation of a digital and analog version of the controller in 45nm CMOS technology, resulting in 3% performance hit with a power overhead in the range of 10-150 microwatts from the controller circuit, (2) higher prediction accuracy compared to a software based OS-governed DVFS scheme, reducing wasted power and improving error margins, (3) power savings of up to 52% and improvement of up to 15% compared to the OS based scheme, (4) implementing DVFS when the processor is memory bandwidth limited for additional savings and (4) accurate temperature prediction to proactively implement DVFS and prevent CPU shutdown.The digital design has minimal power overhead and is more reconfigurable, while analog design is better suited for nonlinear and complex computational tasks.

Original languageEnglish (US)
Title of host publicationProceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011
Pages200-207
Number of pages8
DOIs
StatePublished - 2011
Event2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011 - San Diego, CA, United States
Duration: Jun 6 2011Jun 9 2011

Other

Other2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011
CountryUnited States
CitySan Diego, CA
Period6/6/116/9/11

Fingerprint

Temperature control
Controllers
Program processors
Bandwidth
Data storage equipment
Temperature
Power management
Networks (circuits)
Electric potential

ASJC Scopus subject areas

  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

Sinha, S., Suh, J., Bakkaloglu, B., & Cao, Y. (2011). A workload-aware neuromorphic controller for dynamic power and thermal management. In Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011 (pp. 200-207). [5963936] https://doi.org/10.1109/AHS.2011.5963936

A workload-aware neuromorphic controller for dynamic power and thermal management. / Sinha, Saurabh; Suh, Jounghyuk; Bakkaloglu, Bertan; Cao, Yu.

Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011. 2011. p. 200-207 5963936.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Sinha, S, Suh, J, Bakkaloglu, B & Cao, Y 2011, A workload-aware neuromorphic controller for dynamic power and thermal management. in Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011., 5963936, pp. 200-207, 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011, San Diego, CA, United States, 6/6/11. https://doi.org/10.1109/AHS.2011.5963936
Sinha S, Suh J, Bakkaloglu B, Cao Y. A workload-aware neuromorphic controller for dynamic power and thermal management. In Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011. 2011. p. 200-207. 5963936 https://doi.org/10.1109/AHS.2011.5963936
Sinha, Saurabh ; Suh, Jounghyuk ; Bakkaloglu, Bertan ; Cao, Yu. / A workload-aware neuromorphic controller for dynamic power and thermal management. Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011. 2011. pp. 200-207
@inproceedings{fad140e5645d40339c377379958f1c85,
title = "A workload-aware neuromorphic controller for dynamic power and thermal management",
abstract = "A workload-aware low-power neuromorphic controller for dynamic power and thermal management in VLSI systems is presented. The neuromorphic controller predicts future workload and temperature values based on the past values and preemptively regulates supply voltage and frequency. Our specific contributions include: (1) implementation of a digital and analog version of the controller in 45nm CMOS technology, resulting in 3{\%} performance hit with a power overhead in the range of 10-150 microwatts from the controller circuit, (2) higher prediction accuracy compared to a software based OS-governed DVFS scheme, reducing wasted power and improving error margins, (3) power savings of up to 52{\%} and improvement of up to 15{\%} compared to the OS based scheme, (4) implementing DVFS when the processor is memory bandwidth limited for additional savings and (4) accurate temperature prediction to proactively implement DVFS and prevent CPU shutdown.The digital design has minimal power overhead and is more reconfigurable, while analog design is better suited for nonlinear and complex computational tasks.",
author = "Saurabh Sinha and Jounghyuk Suh and Bertan Bakkaloglu and Yu Cao",
year = "2011",
doi = "10.1109/AHS.2011.5963936",
language = "English (US)",
isbn = "9781457705984",
pages = "200--207",
booktitle = "Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011",

}

TY - GEN

T1 - A workload-aware neuromorphic controller for dynamic power and thermal management

AU - Sinha, Saurabh

AU - Suh, Jounghyuk

AU - Bakkaloglu, Bertan

AU - Cao, Yu

PY - 2011

Y1 - 2011

N2 - A workload-aware low-power neuromorphic controller for dynamic power and thermal management in VLSI systems is presented. The neuromorphic controller predicts future workload and temperature values based on the past values and preemptively regulates supply voltage and frequency. Our specific contributions include: (1) implementation of a digital and analog version of the controller in 45nm CMOS technology, resulting in 3% performance hit with a power overhead in the range of 10-150 microwatts from the controller circuit, (2) higher prediction accuracy compared to a software based OS-governed DVFS scheme, reducing wasted power and improving error margins, (3) power savings of up to 52% and improvement of up to 15% compared to the OS based scheme, (4) implementing DVFS when the processor is memory bandwidth limited for additional savings and (4) accurate temperature prediction to proactively implement DVFS and prevent CPU shutdown.The digital design has minimal power overhead and is more reconfigurable, while analog design is better suited for nonlinear and complex computational tasks.

AB - A workload-aware low-power neuromorphic controller for dynamic power and thermal management in VLSI systems is presented. The neuromorphic controller predicts future workload and temperature values based on the past values and preemptively regulates supply voltage and frequency. Our specific contributions include: (1) implementation of a digital and analog version of the controller in 45nm CMOS technology, resulting in 3% performance hit with a power overhead in the range of 10-150 microwatts from the controller circuit, (2) higher prediction accuracy compared to a software based OS-governed DVFS scheme, reducing wasted power and improving error margins, (3) power savings of up to 52% and improvement of up to 15% compared to the OS based scheme, (4) implementing DVFS when the processor is memory bandwidth limited for additional savings and (4) accurate temperature prediction to proactively implement DVFS and prevent CPU shutdown.The digital design has minimal power overhead and is more reconfigurable, while analog design is better suited for nonlinear and complex computational tasks.

UR - http://www.scopus.com/inward/record.url?scp=80052123455&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80052123455&partnerID=8YFLogxK

U2 - 10.1109/AHS.2011.5963936

DO - 10.1109/AHS.2011.5963936

M3 - Conference contribution

AN - SCOPUS:80052123455

SN - 9781457705984

SP - 200

EP - 207

BT - Proceedings of the 2011 NASA/ESA Conference on Adaptive Hardware and Systems, AHS 2011

ER -