SYNtzulA: Open-Source Hardware for Energy-Efficient Spiking Neural Network Inference

Spiking Neural Networks (SNNs) represent a promising class of neural networks that emulate the behavior of biological neurons,
offering significant advantages in terms of energy efficiency and computational power. These networks achieve optimal performance on neuromorphic processors, however, such hardware remains constrained by prohibitive costs and limited accessibility. To address this gap, open source Process Design Kits (PDKs) and Electronic
Design Automation (EDA) tools can be used to facilitate broader access to hardware development. In this work, we present SYNtzulA (SYNtzulu on ASIC), a system-on-chip that integrates a RISC-V softcore with a dedicated accelerator for SNNs. The chip layout was developed using the open-source IHP-SG13G2 PDK and the OpenROAD flow, reducing development costs and promoting wider accessibility to neuromorphic hardware solutions. SYNtzulA occupies an area of approximately 5.2 mm2 and operates at a maximum frequency of 125 MHz. The system is capable of processing 109 synapses per second at its maximum frequency, while maintaining a power consumption during inference that scales linearly with the operating frequency, dissipating approximately 632 𝜇W/MHz.