SiFive and CEVA are joining forces to help customers create ultra-low-power domain-specific Edge AI processors, as announced at CES this week. The aim is to produce the right SoC mashup of commercial RISC-V CPU processor IP and CEVA’s digital signal processing cores and AI processors for each customer tailored to their specific markets without dark silicon lurking to drag the energy use up. “We are focusing more on the domain-specific design aspects,” said James Prior, head of global communications for SiFive, in an interview with Semiconductor Engineering. “We’ve got a concept design that could be further fine-tuned to a specific market application.” Both companies bring pre-proven IP that SiFive says can speed up the design and test process. The Edge AI SoCs are supported by CEVA’s CDNN Deep Neural Network machine learning software compiler that creates fully-optimized runtime software for the CEVA-XM vision processors, CEVA-BX audio DSPs and NeuPro AI processors.