At Embedded World 2025, NXP showcased the integration of the DeepSeek R1 8B model with its i.MX 8M Plus application processor, augmented by Kinara’s Ara-2 neural processing unit (NPU). This configuration exemplifies the synergy between NXP’s processing capabilities and Kinara’s specialized AI acceleration.
—
Synaptics is my Embedded World 2025 video coverage sponsor, check out my Synaptics videos here: https://www.youtube.com/playlist?list=PL7xXqJFxvYvhAbQoe9YN4c84SqXxIY3fQ
—
DeepSeek R1 is an open-source reasoning model designed to emulate human-like cognitive processes. Before providing responses, it undergoes iterative self-assessment to ensure accuracy and reliability. When queried about appropriate actions following a car accident, DeepSeek R1 generated a comprehensive, step-by-step guide, reflecting its advanced reasoning capabilities.
The integration of Kinara’s Ara-2 NPU, capable of delivering up to 40 TOPS (Tera Operations Per Second), significantly enhances the system’s AI performance. This hardware acceleration facilitates efficient processing of complex models like DeepSeek R1, enabling rapid and accurate responses to intricate queries.
In another demonstration, NXP presented a multimodal application utilizing the Large Language and Vision Assistant (LAVA) model. This model combines a vision encoder with a large language model (LLM) to analyze and interpret visual data. Operating entirely at the edge, this setup ensures data privacy, low latency, and reduced power consumption.
When tasked with describing a challenging image, the system processed the visual input and, within approximately seven seconds, provided a detailed description. This rapid analysis underscores the efficacy of edge computing in handling complex AI workloads without relying on cloud-based resources.
The system also demonstrated real-time video analysis capabilities. By capturing frames from a live video stream and processing them through the LAVA model, it generated accurate descriptions of dynamic scenes. This functionality has potential applications in security monitoring, event summarization, and automated content generation.
NXP’s recent agreement to acquire Kinara for $307 million underscores its commitment to advancing edge AI technologies. This strategic move aims to enhance NXP’s processing portfolio with Kinara’s high-performance, energy-efficient NPUs, thereby expanding the possibilities for intelligent edge applications across industrial and automotive sectors.
The combination of NXP’s processors and Kinara’s AI acceleration hardware offers a scalable platform for deploying sophisticated AI models at the edge. This integration facilitates the development of intelligent systems capable of real-time data processing, decision-making, and autonomous operation, marking a significant step forward in edge computing.
Industry observers speculate that NXP’s acquisition of Kinara could lead to the development of more energy-efficient and high-performance AI solutions, particularly in the realms of autonomous vehicles and smart industrial applications. This move positions NXP to better compete with other semiconductor companies investing in edge AI capabilities.
Check out all my Embedded World 2025 videos in this playlist: https://www.youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC using the dual wireless DJI Mic 2 microphones with the DJI lapel microphone https://amzn.to/3XIj3l8 ), watch all my DJI Pocket 3 videos here https://www.youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Join https://www.youtube.com/charbax/join for Early Access to my videos and to support my work, or you can Click the “Super Thanks” button below the video to send a highlighted comment!
