Exploiting 2.5D/3D Heterogeneous Integration for AI Computing

Zhenyu Wang; Jingbo Sun; Alper Goksoy; Sumit K. Mandal; Yaotian Liu; Jae Sun Seo; Chaitali Chakrabarti; Umit Y. Ogras; Vidya Chhabria; Jeff Zhang; Yu Cao

doi:10.1109/ASP-DAC58780.2024.10473875

Exploiting 2.5D/3D Heterogeneous Integration for AI Computing

Zhenyu Wang, Jingbo Sun, Alper Goksoy, Sumit K. Mandal, Yaotian Liu, Jae Sun Seo, Chaitali Chakrabarti, Umit Y. Ogras, Vidya Chhabria, Jeff Zhang, Yu Cao

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

3 Scopus citations

Abstract

The evolution of AI algorithms has not only revolutionized many application domains, but also posed tremendous challenges on the hardware platform. Advanced packaging technology today, such as 2.5D and 3D interconnection, provides a promising solution to meet the ever-increasing demands of bandwidth, data movement, and system scale in AI computing. This work presents HISIM, a modeling and benchmarking tool for chiplet-based heterogeneous integration. HISIM emphasizes the hierarchical interconnection that connects various chiplets through network-on-package. It further integrates technology roadmap, power/latency prediction, and thermal analysis together to support electro-thermal co-design. Leveraging HISIM with in-memory computing chiplets, we explore the advantages and limitations of 2.5D and 3D heterogenous integration on representative AI algorithms, such as DNNs, transformers, and graph neural networks.

Original language	English (US)
Title of host publication	ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	758-764
Number of pages	7
ISBN (Electronic)	9798350393545
DOIs	https://doi.org/10.1109/ASP-DAC58780.2024.10473875
State	Published - 2024
Externally published	Yes
Event	29th Asia and South Pacific Design Automation Conference, ASP-DAC 2024 - Incheon, Korea, Republic of Duration: Jan 22 2024 → Jan 25 2024

Publication series

Name	Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Conference

Conference	29th Asia and South Pacific Design Automation Conference, ASP-DAC 2024
Country/Territory	Korea, Republic of
City	Incheon
Period	1/22/24 → 1/25/24

Keywords

2.5D
3D
Chiplet
Heterogeneous Integration
ML accelerators
Performance Analysis

ASJC Scopus subject areas

Electrical and Electronic Engineering
Computer Science Applications
Computer Graphics and Computer-Aided Design

Access to Document

10.1109/ASP-DAC58780.2024.10473875

Cite this

Wang, Z., Sun, J., Goksoy, A., Mandal, S. K., Liu, Y., Seo, J. S., Chakrabarti, C., Ogras, U. Y., Chhabria, V., Zhang, J., & Cao, Y. (2024). Exploiting 2.5D/3D Heterogeneous Integration for AI Computing. In ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings (pp. 758-764). (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASP-DAC58780.2024.10473875

Exploiting 2.5D/3D Heterogeneous Integration for AI Computing. / Wang, Zhenyu; Sun, Jingbo; Goksoy, Alper et al.
ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings. Institute of Electrical and Electronics Engineers Inc., 2024. p. 758-764 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Wang, Z, Sun, J, Goksoy, A, Mandal, SK, Liu, Y, Seo, JS, Chakrabarti, C, Ogras, UY, Chhabria, V , Zhang, J & Cao, Y 2024, Exploiting 2.5D/3D Heterogeneous Integration for AI Computing. in ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings. Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, Institute of Electrical and Electronics Engineers Inc., pp. 758-764, 29th Asia and South Pacific Design Automation Conference, ASP-DAC 2024, Incheon, Korea, Republic of, 1/22/24. https://doi.org/10.1109/ASP-DAC58780.2024.10473875

Wang Z, Sun J, Goksoy A, Mandal SK, Liu Y, Seo JS et al. Exploiting 2.5D/3D Heterogeneous Integration for AI Computing. In ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings. Institute of Electrical and Electronics Engineers Inc. 2024. p. 758-764. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). doi: 10.1109/ASP-DAC58780.2024.10473875

@inproceedings{15d83fb22cff43fd935109740e5c8315,

title = "Exploiting 2.5D/3D Heterogeneous Integration for AI Computing",

abstract = "The evolution of AI algorithms has not only revolutionized many application domains, but also posed tremendous challenges on the hardware platform. Advanced packaging technology today, such as 2.5D and 3D interconnection, provides a promising solution to meet the ever-increasing demands of bandwidth, data movement, and system scale in AI computing. This work presents HISIM, a modeling and benchmarking tool for chiplet-based heterogeneous integration. HISIM emphasizes the hierarchical interconnection that connects various chiplets through network-on-package. It further integrates technology roadmap, power/latency prediction, and thermal analysis together to support electro-thermal co-design. Leveraging HISIM with in-memory computing chiplets, we explore the advantages and limitations of 2.5D and 3D heterogenous integration on representative AI algorithms, such as DNNs, transformers, and graph neural networks.",

keywords = "2.5D, 3D, Chiplet, Heterogeneous Integration, ML accelerators, Performance Analysis",

author = "Zhenyu Wang and Jingbo Sun and Alper Goksoy and Mandal, {Sumit K.} and Yaotian Liu and Seo, {Jae Sun} and Chaitali Chakrabarti and Ogras, {Umit Y.} and Vidya Chhabria and Jeff Zhang and Yu Cao",

note = "Publisher Copyright: {\textcopyright} 2024 IEEE.; 29th Asia and South Pacific Design Automation Conference, ASP-DAC 2024 ; Conference date: 22-01-2024 Through 25-01-2024",

year = "2024",

doi = "10.1109/ASP-DAC58780.2024.10473875",

language = "English (US)",

series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "758--764",

booktitle = "ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings",

}

TY - GEN

T1 - Exploiting 2.5D/3D Heterogeneous Integration for AI Computing

AU - Wang, Zhenyu

AU - Sun, Jingbo

AU - Goksoy, Alper

AU - Mandal, Sumit K.

AU - Liu, Yaotian

AU - Seo, Jae Sun

AU - Chakrabarti, Chaitali

AU - Ogras, Umit Y.

AU - Chhabria, Vidya

AU - Zhang, Jeff

AU - Cao, Yu

PY - 2024

Y1 - 2024

N2 - The evolution of AI algorithms has not only revolutionized many application domains, but also posed tremendous challenges on the hardware platform. Advanced packaging technology today, such as 2.5D and 3D interconnection, provides a promising solution to meet the ever-increasing demands of bandwidth, data movement, and system scale in AI computing. This work presents HISIM, a modeling and benchmarking tool for chiplet-based heterogeneous integration. HISIM emphasizes the hierarchical interconnection that connects various chiplets through network-on-package. It further integrates technology roadmap, power/latency prediction, and thermal analysis together to support electro-thermal co-design. Leveraging HISIM with in-memory computing chiplets, we explore the advantages and limitations of 2.5D and 3D heterogenous integration on representative AI algorithms, such as DNNs, transformers, and graph neural networks.

AB - The evolution of AI algorithms has not only revolutionized many application domains, but also posed tremendous challenges on the hardware platform. Advanced packaging technology today, such as 2.5D and 3D interconnection, provides a promising solution to meet the ever-increasing demands of bandwidth, data movement, and system scale in AI computing. This work presents HISIM, a modeling and benchmarking tool for chiplet-based heterogeneous integration. HISIM emphasizes the hierarchical interconnection that connects various chiplets through network-on-package. It further integrates technology roadmap, power/latency prediction, and thermal analysis together to support electro-thermal co-design. Leveraging HISIM with in-memory computing chiplets, we explore the advantages and limitations of 2.5D and 3D heterogenous integration on representative AI algorithms, such as DNNs, transformers, and graph neural networks.

KW - 2.5D

KW - 3D

KW - Chiplet

KW - Heterogeneous Integration

KW - ML accelerators

KW - Performance Analysis

UR - http://www.scopus.com/inward/record.url?scp=85189295237&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85189295237&partnerID=8YFLogxK

U2 - 10.1109/ASP-DAC58780.2024.10473875

DO - 10.1109/ASP-DAC58780.2024.10473875

M3 - Conference contribution

AN - SCOPUS:85189295237

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 758

EP - 764

BT - ASP-DAC 2024 - 29th Asia and South Pacific Design Automation Conference, Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 29th Asia and South Pacific Design Automation Conference, ASP-DAC 2024

Y2 - 22 January 2024 through 25 January 2024

ER -

Exploiting 2.5D/3D Heterogeneous Integration for AI Computing

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this