Saturday, February 15, 2025

AI Chip For Robots, AI Boxes, And City Security

- Advertisement -

The edge AI chip decodes 12 video streams, processes AI models, and supports on-device GenAI processing for robots, security, and data applications.

Ambarella Expands N1 Edge GenAI Family With SoC Targeted at On-Premise Multi-Channel VLM and NN Processing in Under 20 Watts
Ambarella Expands N1 Edge GenAI Family With SoC Targeted at On-Premise Multi-Channel VLM and NN Processing in Under 20 Watts

Ambarella, Inc., an edge AI semiconductor company, announced at CES the N1-655 edge GenAI system-on-chip (SoC). This chip can decode 12 simultaneous 1080p30 video streams while processing the video and running multiple vision-language models (VLMs) alongside CNNs. It supports multimodal VLMs and large-language models (LLMs) while consuming 20 watts, which is less than cloud processors. The N1-655 can run models like Phi, Gemma, LLaVA-OneVision, and Llama without an internet connection, handling inputs such as visuals and speech for applications in AI boxes, autonomous mobile robots (AMRs), and security video recorders.

Potential users include developers of autonomous mobile robots (AMRs), businesses implementing on-premise AI boxes for tasks like security or data processing, and providers of smart security video recorders.

- Advertisement -

After introducing the N1 SoC last year, Ambarella is expanding its edge GenAI SoC lineup for tasks beyond on-camera processing. Both current models run GenAI models locally, enhancing privacy while reducing power use and total cost compared to cloud-based processing.

There is a growing shift towards performing more GenAI processing at the edge, driven by the costs of cloud-based bandwidth, service fees, high power usage, and concerns about data security and privacy. Ambarella uses advanced fine-tuning methods like QLoRA to reduce the size of popular GenAI models, enabling their use in edge applications. The company leads this shift with the N1-655, offering high performance and low power consumption.

“The N1-655 continues Ambarella’s industry-leading levels of AI performance per watt, this time for AMRs and on-premise AI box applications that require the latest edge GenAI features while keeping power consumption under 20 watts,” said Fermi Wang, President and CEO of Ambarella. “We tailored this new SoC to run faster, while fine-tuning the most popular LLMs and VLMs to make them even smaller, eliminating the need for high-bandwidth and cost-intensive cloud processing in this class of applications.”

For more information, click here.

Nidhi Agarwal
Nidhi Agarwal
Nidhi Agarwal is a Senior Technology Journalist at EFY with a deep interest in embedded systems, development boards and IoT cloud solutions.

SHARE YOUR THOUGHTS & COMMENTS

EFY Prime

Unique DIY Projects

Electronics News

Truly Innovative Electronics

Latest DIY Videos

Electronics Components

Electronics Jobs

Calculators For Electronics