CEVA's NeuPro-M NPUs: AI Revolution!

CEVA, Inc., a leading provider of advanced wireless technology and custom AI solutions, has unveiled its upgraded NeuPro-M NPU series. Designed to meet the high-performance needs of next-gen Generative AI, these NPUs excel in both power and efficiency. From cloud computing to edge devices, they seamlessly handle various AI tasks. This series supports transformer networks, in addition to other neural networks, enabling efficient development and operation of AI applications across a wide range of scenarios.

Ran Snir, Vice President and General Manager of CEVA's Vision Business Unit, explains, "Our NeuPro-M NPU IP is purpose-built to efficiently handle current and future AI workloads, delivering exceptional performance across different use cases, from edge devices to cloud computing."

ABI Research predicts substantial growth in Edge AI, driven by the power of Generative AI. While AI's potential to transform industries is vast, deploying it on edge devices presents unique challenges like performance and power consumption.

The enhanced NeuPro-M architecture, equipped with a versatile vector processing unit (VPU), offers scalability and supports various network layers. It boasts exceptional processing performance and energy efficiency, making it a standout NPU IP product in the industry. The new NPM12 and NPM14 NPU cores provide enhanced AI capabilities, catering to diverse AI needs.

CEVA's comprehensive development toolkit, built around their renowned CDNN neural network AI compiler, enhances the NeuPro-M experience. It maximizes the potential of NeuPro-M's processing engines, ensuring optimal AI application performance. The CDNN software package also supports popular open-source frameworks, enhancing compatibility and flexibility.