Thermal designs, models and optimization for three-dimensional integrated circuits

Hwang, Leslie K.

Permalink

https://hdl.handle.net/2142/102951 Copy

Description

Title

Thermal designs, models and optimization for three-dimensional integrated circuits

Author(s)

Hwang, Leslie K.

Issue Date

2018-12-07

Director of Research (if dissertation) or Advisor (if thesis)

Wong, Martin

Doctoral Committee Chair(s)

Wong, Martin

Committee Member(s)

• Chen, Deming
• Hwu, Wen-Mei
• Miljkovic, Nenad

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• 3D IC
• thermal design
• TTSV
• thermal fin
• microchannel
• optimization

Abstract

Three-dimensional integrated circuits (3D ICs), a novel packaging technology, are heavily studied to enable improved performance with denser packaging and reduced interconnects. Despite numerous advantages, thermal management is the biggest bottleneck to expanding the applications of this device stacking technology. In addition to implementing the thermal-aware designs of existing methodologies, it is necessary to implement new features to dissipate heat efficiently. This work presents two main aspects of thermal designs: on-chip level and package level. First, we propose a novel thermal-aware physical design on chip between devices. We aim to mitigate localized hotspots to ensure the functionality by adding thermal fin geometry to existing thermal through- silicon via (TTSV). We analyze design requirements of thermal fin for single TTSV as well as TTSV cluster designs with the goal of maximizing heat dissipation while minimizing the interference with routing and area consumption. An analytical model of the three-dimensional system and thermal resistance circuit is built for accurate and runtime-efficient thermal analysis. In terms of high-performance computing systems in 3D ICs, thermal bottle- necks are much more challenging with merely on-chip design solutions. Inter- tier liquid cooling microchannel layers have been introduced into 3D ICs as an integrated cooling mechanism to tackle the thermal degradation. Many existing research works optimize microchannel designs based on runtime-intensive numerical methods or inaccurate thermo-fluid models. Hence, we propose an accurate but compact closed-form model of tapered microchannel to capture the relationship between the channel geometry and heat transfer performance. To improve the accuracy, our correlations are based on the developing flow model and derived from numerical simulation data on a sub- set of multiple channel parameters. Our model achieves 57% less error in Nusselt number and 45 % less error in pressure drop for channels with inlet width 100-400 μm compared to a commonly used approximate model on fully developed flow. Next, we present the correlations for diverging channels as well as complete correlations that extend to any linearly tapering channel models, that include diverging shape, uniformly rectangular shape and converging shape. The complete models provide the flexibility to analyze and optimize any arbitrary geometry based on the piecewise linear channel wall assumption. Finally, we demonstrate the optimized channel designs using the derived correlations. Tapered channel models provided the flexibility to incorporate any arbitrary shapes and explore the advanced geometries during the optimization. The microchannel is divided into small segments in axial direction from inlet to outlet and piecewise optimized. The simulated annealing method is applied in our optimization, and channel width at one randomly chosen segment interface is altered to evaluate the design at each iteration. The objective is to minimize the overall thermal resistance while pressure drop is maintained less than a threshold value and channel widths have minimum and maximum boundaries. We compare the designs with the optimization based on fully developed flow models and verify the channel performance through numerical simulations. To guarantee optimality, accurate analysis is crucial. Our proposed models have significantly improved the accuracy by applying the appropriate flow assumption. However, many opportunities exist to increase the design flexibility and the accuracy. Fluid conditions, such as coolant material and varying volumetric flow rate, can also be part of the optimization parameters to expand the design scope. Moreover, physical phenomena, such as reduced friction on the channel walls or a vortex created on abrupt angle changes, can be considered to improve the accuracy in the closed-form models.

Graduation Semester

2018-12

Type of Resource

text

Permalink

http://hdl.handle.net/2142/102951

Copyright and License Information

Copyright 2018 Leslie Hwang

Owning Collections