A leading PC OEM builds its next on-device AI compute architecture on Ceva-NeuPro

Customer case study | Ceva-NeuPro Max and Ceva-NeuPro Prime

One of the world’s leading PC original equipment manufacturers selected Ceva NeuPro NPUs as foundational IP for its next-generation on-device AI compute architecture. The design targets AI PCs, tablets and personal AI experiences & AI agents that run locally. 

The selected approach uses two NPU sizes to address different processing requirements within the same product strategy, while aligning with a roadmap intended to support evolving AI and large-language-model workloads. One NPU is used for running advanced LLM for gen-AI and AI agents while the other is running always on power efficient models such as face recognition and gesture tracking models.  

Customer’s target markets & applications

  • AI PCs 
  • Tablets 
  • Personal AI  
  • On-device generative AI 

The challenge

As AI perfoliate across devices and use cases and moves to the edge, the customer needed to bring on-device generative AI to PCs and tablets, power personal AI experiences locally, and keep pace with rapidly changing AI models. 

Why Ceva

  • Performance for advanced on-device AI: Ceva provides the AI processing capabilities required to support next-generation personal AI experiences on PCs and tablets. 
  • Scalability across product requirements: Two NPU sizes allow the customer to address different performance requirements while maintaining a common architectural foundation. 
  • Alignment with the customer’s AI roadmap: Ceva’s roadmap supports the continued development of the customer’s on-device AI compute architecture. 
  • Support for evolving AI workloads: The architecture is designed to address increasingly demanding AI models and on-device generative AI applications. 

Implementation Highlights

  • Architecture license 
  • Ceva-NeuPro Max (32K) 
  • Ceva-NeuPro Prime (2K)  
  • Architecture built for on-device generative AI