Abstract Preview

Here is the abstract you requested from the dpc_2018 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.

Thermal Co-Design of Exascale Computing System in Packages (SiPs)
Keywords: through-silicon via, liquid cooling, simulation study
Exascale computing power could lead to dramatic advances in various scientific fields, from precision medicine to regional climate, water use to materials science, atomic physics to deep learning, improving both our understanding of the world and how we live in it. Exascale computing entails the use of 3D stacked electronic devices in order to minimize data transfer latency, maximize power efficiency, and achieve high-density packaging to enable aggregation of the full large-scale system. Due to their significantly higher heat generation density and less access to the heat sink placed on top of the electronic package, 3D stacked electronic devices are much more susceptible than their 2D counterparts to overheating. The overheating problem is one of the issues to be addressed to realize an exascale computer. Although the thermal design of 3D electronics with through-silicon vias (TSVs) have been studied in general [3, 4, and 1], the thermal co-design of an exascale system in package (SiP) has not yet been extensively investigated and only briefly discussed in [5]. In this work, in order to keep the SiP component temperatures within the allowable limits, various package designs, module layouts, and liquid-cooled thermal management solutions using TSVs [2] have been investigated via simulation studies performed using COMSOL 5.2, a multi-physics modeling software package. We have identified package and module design decision trade-offs in order to minimize thermal resistance and pressure drop, thus maximizing the cooling solution efficiency. Specifically, we have shown the significant effect of module layout and cooling liquid flow direction on the pressure drop and thus the pump work, considerably reducing the operating cost of the system. Also, we have demonstrated that the relative sizes of the components stacked on top of one another play a key role in transferring the heat through the 3D stack up to the heat sink and thus reducing the component temperatures. Finally, we propose schemes to minimize the thermal resistance between the modules of highest power in a 3D stack and the heat sink, noticeably decreasing the 3D stack temperature. Aggregating all the individual findings of the study into an optimal, realistic cooling solution demonstrated the significant efficacy and efficiency of our proposed design recommendations, reliably and cost-effectively keeping the components temperatures within the allowable limits.
Koosha Nassiri Nazif, PhD student
Stanford University
Stanford, CA

  • Amkor
  • ASE
  • Canon
  • Corning
  • EMD Performance Materials
  • Honeywell
  • Indium
  • Kester
  • Kyocera America
  • Master Bond
  • Micro Systems Technologies
  • MRSI
  • Palomar
  • Promex
  • Qualcomm
  • Quik-Pak
  • Raytheon
  • Rochester Electronics
  • Specialty Coating Systems
  • Spectrum Semiconductor Materials
  • Technic