Search papers, labs, and topics across Lattice.
SNAP-V, a RISC-V based neuromorphic SoC, is introduced with two accelerator variants (Cerebra-S and Cerebra-H) designed for efficient small-scale SNN inference. The architecture integrates a RISC-V core with parallel processing nodes and distributed memory to mitigate memory-processor bottlenecks common in conventional SoCs. Experimental validation demonstrates accurate SNN inference with an average accuracy deviation of 2.62% and an energy efficiency of 1.05 pJ per synaptic operation in 45 nm CMOS.
Achieve energy-efficient, accurate spiking neural network inference on the edge with a RISC-V SoC that consumes just 1.05 pJ per synaptic operation.
Spiking Neural Networks (SNNs) have gained significant attention in edge computing due to their low power consumption and computational efficiency. However, existing implementations either use conventional System on Chip (SoC) architectures that suffer from memory-processor bottlenecks, or large-scale neuromorphic hardware that is inefficient and wasteful for small-scale SNN applications. This work presents SNAP-V, a RISC-V-based neuromorphic SoC with two accelerator variants: Cerebra-S (bus-based) and Cerebra-H (Network-on-Chip (NoC)-based) which are optimized for small-scale SNN inference, integrating a RISC-V core for management tasks, with both accelerators featuring parallel processing nodes and distributed memory. Experimental results show close agreement between software and hardware inference, with an average accuracy deviation of 2.62% across multiple network configurations, and an average synaptic energy of 1.05 pJ per synaptic operation (SOP) in 45 nm CMOS technology. These results show that the proposed solution enables accurate, energy-efficient SNN inference suitable for real-time edge applications.
