AIG-CIM: A Scalable Chiplet Module with Tri-Gear Heterogeneous Compute-in-Memory for Diffusion Acceleration
By Yiqi Jing1, Meng Wu1, Jiaqi Zhou1, Yiyang Sun1, Yufei Ma1,2, Ru Huang1 , Le Ye1,3 , Tianyu Jia1
1 School of Integrated Circuits, Peking University, Beijing, China, 2 Institute for Artificial Intelligence, Peking University, Beijing, China, 3 Advanced Institute of Information Technology of Peking University, Hangzhou, China
ABSTRACT
The emergence of Diffusion models has gained significant attention in the field of Artificial Intelligence Generated Content. While Diffusion demonstrates impressive image generation capability, it faces hardware deployment challenges due to its unique model architecture and computation requirement. In this paper, we present a hardware accelerator design, i.e. AIG-CIM, which incorporates tri-gear heterogeneous digital compute-in-memory to address the flexible data reuse demands in Diffusion models. Our framework offers a collaborative design methodology for large generative models from the computational circuit-level to the multi-chip-module system-level. We implemented and evaluated the AIG-CIM accelerator using TSMC 22nm technology. For several Diffusion inferences, scalable AIG-CIM chiplets achieve 21.3× latency reduction, up to 231.2× throughput improvement and three orders of magnitude energy efficiency improvement compared to RTX 3090 GPU.
To read the full article, click here
Related Chiplet
- Automotive AI Accelerator
- Direct Chiplet Interface
- HBM3e Advanced-packaging chiplet for all workloads
- UCIe AP based 8-bit 170-Gsps Chiplet Transceiver
- UCIe based 8-bit 48-Gsps Transceiver
Related Technical Papers
- Why package lithography matters in heterogeneous chiplet integration
- A Heterogeneous Chiplet Architecture for Accelerating End-to-End Transformer Models
- Heterogeneous Integration - Chiplets
- Challenges and Opportunities to Enable Large-Scale Computing via Heterogeneous Chiplets
Latest Technical Papers
- Taking 3D IC Heterogeneous Integration Mainstream
- CATCH: a Cost Analysis Tool for Co-optimization of chiplet-based Heterogeneous systems
- Three-dimensional photonic integration for ultra-low-energy, high-bandwidth interchip data links
- Defect Analysis and Built-In-Self-Test for Chiplet Interconnects in Fan-out Wafer-Level Packaging
- ARCAS: Adaptive Runtime System for Chiplet-Aware Scheduling