Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

Joel Öhrling; Sébastien Lafond; Dragos Truscan

doi:10.48550/arXiv.2112.10121

Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

Informationsteknologi

Forskningsoutput: Kapitel i bok/konferenshandling › Publicerad konferensartikel › Vetenskaplig › Peer review

1 Citeringar (Scopus)

96 Nedladdningar (Pure)

Sammanfattning

In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of difierent configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core.
We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.

Originalspråk	Engelska
Titel på värdpublikation	Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021.
Förlag	Springer
Sidor	144-160
Antal sidor	17
ISBN (tryckt)	978-3-031-04579-0
DOI	https://doi.org/10.48550/arXiv.2112.10121 https://doi.org/10.1007/978-3-031-04580-6_10
Status	Publicerad - 2022
MoE-publikationstyp	A4 Artikel i en konferenspublikation
Evenemang	21st International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation, SAMOS 2021 - Virtual, Online Varaktighet: 4 juli 2021 → 8 juli 2021

Publikationsserier

Namn	Lecture Notes in Computer Science
Förlag	Springer
Volym	13227
ISSN (tryckt)	0302-9743

Konferens

Konferens	21st International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation, SAMOS 2021
Ort	Virtual, Online
Period	04/07/21 → 08/07/21

Finansiering

Part of this work was carried out with financial support from the Nordic Master programme (contract NMP-2016/10169) and ECSEL-JU AIDOaRt project (grant agreement No. 101007350).

Åtkomst av dokument

Camera ready versionAccepterat manuskript från författare, 755 KBLicens: Publisher rights policy

https://urn.fi/URN:NBN:fi-fe2023042739008

Andra filer och länkar

Author version, Arxiv.org

AIDOaRT
Porres Paltor, I. (Ansvarig forskare), Truscan, D. (CoPI), Nybom, K. (CoI), Logacheva, E. (CoI), Winsten, J. (CoI) & Peltomäki, J. (CoI)
European Commission
01/04/21 → 30/09/24
Projekt: EU

Citera det här

Öhrling, J., Lafond, S., & Truscan, D. (2022). Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs. I Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021. (s. 144-160). ( Lecture Notes in Computer Science; Vol. 13227). Springer. https://doi.org/10.48550/arXiv.2112.10121, https://doi.org/10.1007/978-3-031-04580-6_10

@inproceedings{2cacc30d0fbe4c94bc6ff164a283e777,

title = "Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs",

abstract = "In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of difierent configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core. We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.",

keywords = "energy dissipation, Temperature, Multicore processing, System identification",

author = "Joel {\"O}hrling and S{\'e}bastien Lafond and Dragos Truscan",

note = "Funding Information: Part of this work was carried out with financial support from the Nordic Master programme (contract NMP-2016/10169) and ECSEL-JU AIDOaRt project (grant agreement No. 101007350). Publisher Copyright: {\textcopyright} 2022, Springer Nature Switzerland AG.; 21st International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation, SAMOS 2021 ; Conference date: 04-07-2021 Through 08-07-2021",

year = "2022",

doi = "10.48550/arXiv.2112.10121",

language = "English",

isbn = "978-3-031-04579-0",

series = " Lecture Notes in Computer Science",

publisher = "Springer",

pages = "144--160",

booktitle = "Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021.",

}

Öhrling, J, Lafond, S & Truscan, D 2022, Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs. i Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021. . Lecture Notes in Computer Science, vol. 13227, Springer, s. 144-160, 21st International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation, SAMOS 2021, Virtual, Online, 04/07/21. https://doi.org/10.48550/arXiv.2112.10121, https://doi.org/10.1007/978-3-031-04580-6_10

Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs. / Öhrling, Joel; Lafond, Sébastien ; Truscan, Dragos.
Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021. . Springer, 2022. s. 144-160 ( Lecture Notes in Computer Science; Vol. 13227).

Forskningsoutput: Kapitel i bok/konferenshandling › Publicerad konferensartikel › Vetenskaplig › Peer review

TY - GEN

T1 - Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

AU - Öhrling, Joel

AU - Lafond, Sébastien

AU - Truscan, Dragos

N1 - Funding Information: Part of this work was carried out with financial support from the Nordic Master programme (contract NMP-2016/10169) and ECSEL-JU AIDOaRt project (grant agreement No. 101007350). Publisher Copyright: © 2022, Springer Nature Switzerland AG.

PY - 2022

Y1 - 2022

N2 - In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of difierent configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core. We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.

AB - In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of difierent configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core. We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.

KW - energy dissipation

KW - Temperature

KW - Multicore processing

KW - System identification

UR - https://arxiv.org/abs/2112.10121

U2 - 10.48550/arXiv.2112.10121

DO - 10.48550/arXiv.2112.10121

M3 - Published conference proceeding

AN - SCOPUS:85129800655

SN - 978-3-031-04579-0

T3 - Lecture Notes in Computer Science

SP - 144

EP - 160

BT - Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021.

PB - Springer

T2 - 21st International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation, SAMOS 2021

Y2 - 4 July 2021 through 8 July 2021

ER -

Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

Sammanfattning

Publikationsserier

Konferens

Finansiering

Åtkomst av dokument

Andra filer och länkar

Fingeravtryck

Projekt

AIDOaRT

Citera det här