Mumtahina Islam Sukanya1, Aniket Bharamgonda2, Abhijit Dasgupta2, and Ankur Srivastava1
1Department of Electrical and Computer Engineering, University of Maryland, College Park, MD, USA
2Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park, MD, USA
For more information about this article and related research, please contact Prof. Abhijit Dasgupta.
Abstract:
This cross-domain study presents an end-to-end digital twin system for heterogeneous integrated(HI) processors. This simulation framework is built to bridge the gap between architectural-level modeling and package-level physical design for HI systems. The framework combines architectural performance, power consumption, thermal profiles, and mechanical stress analysis into a unified digital twin environment that allows the system to be examined by various cross-domain metrics. Such an integrated approach facilitates detailed multi-physics reliability assessment and supports comprehensive evaluation of design trade-offs. Representative case studies are presented in this work to demonstrate how the digital twin captures the impact of varying internal system modifications or external input variations. This internal variation was performed by modifying the architectural parameter, core-memory communication bandwidth, and the external change was modeled by changing the system’s benchmark workload input pattern. The results indicate that the framework provides physically grounded and extensive insight into the architectural-thermal-mechanical trade-offs associated with these changes. Moreover, a mitigative co-design example presented illustrated how the information generated by the digital twin can be used not only for system evaluation but also for thermal and mechanical management of the system through dynamic architectural adjustments. Overall, the framework offers a comprehensive perspective into design exploration and reliability assessment of HI systems.
This article is available online here and to CALCE Consortium Members for personal review.