Co-Optimization of Power Delivery Network Design for 3-D Heterogeneous Integration of RRAM-Based Compute In-Memory Accelerators
Three-dimensional heterogeneous integration (3D-HI) offers promising solutions for incorporating substantial embedded memory into cutting-edge analog compute-in-memory (CIM) AI accelerators, addressing the need for on-chip acceleration of large AI models. However, this approach faces challenges with...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal on Exploratory Solid-State Computational Devices and Circuits |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10854426/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Three-dimensional heterogeneous integration (3D-HI) offers promising solutions for incorporating substantial embedded memory into cutting-edge analog compute-in-memory (CIM) AI accelerators, addressing the need for on-chip acceleration of large AI models. However, this approach faces challenges with power supply noise (PSN) margins due to <inline-formula> <tex-math notation="LaTeX">$V_{\text {DD}}$ </tex-math></inline-formula> scaling and increased power delivery network (PDN) impedance. This study demonstrates the necessity and benefits of 3D-HI for large-scale CIM accelerators, where 2-D implementations would exceed manufacturing reticle limits. Our 3-D designs achieve 39% higher energy efficiency, <inline-formula> <tex-math notation="LaTeX">$8\times $ </tex-math></inline-formula> higher operation density, and improved throughput through shorter vertical interconnects. We quantify steady-state IR-drop impacts in 3D-HI CIM architectures using a framework that combines PDN modeling, 3D-HI power, performance, area estimation, and behavioral modeling. We demonstrate that a drop in supply voltage to CIM arrays increases sensitivity to process, voltage, and temperature (PVT) noise. Using our framework, we model IR-drop and simulate its impact on the accuracy of ResNet-50 and ResNet-152 when classifying images from the ImageNet 1k dataset in the presence of injected PVT noise. We analyze the impact of through-silicon via (TSV) design and placement to optimize the IR-drop and classification accuracy. For ResNet architectures in 3-D integration, we demonstrate that peripheral TSV placement provides an optimal balance between interconnect complexity and performance, achieving IR-drop below 10% of <inline-formula> <tex-math notation="LaTeX">$V_{\text {DD}}$ </tex-math></inline-formula> while maintaining high classification accuracy. |
|---|---|
| ISSN: | 2329-9231 |