DynaPlasia: Reconfigurable Spatial Accelerator with eDRAM-IMC
본문
Overview
We present an eDRAM-based CIM processor called DynaPlasia with a novel triple-mode 3T2C cell and a dynamic reconfigurable core architecture that enables high system efficiency for ML workloads. For ResNet-18 (ImageNet dataset), DynaPlasia achieves system energy efficiency of 37.2TOPS/W and compute density of 2.03TOPS/mm2 at 1.0V and 250MHz for INT4/INT5 activation/weight precision
Implementation results
Performance Comparison
Architecture
Features
- Dynamic Reconfigurable Core Architecture
- Triple-mode Cell
- Hierarchical In-memory ADC
- Signed-input Signed-weight IMC Operation
- Leakage Tolerant Computing
Related Papers
- ISSCC 2023