In part one of this blog series, we covered how NXP’s i.MX 95 family brings the latest functional safety features to edge applications, from industrial environments to vehicles. In this next part of the series, we explore how the i.MX 95 family delivers advanced AI capabilities to the edge.
Over the past few years, AI innovation has grown exponentially. While a lot of the hype is centered around cloud-based generative AI applications like ChatGPT, the reality is that these applications are extremely costly to run and use massive amounts of energy. Edge processing offers a compelling solution to manage information quickly, efficiently and securely by keeping data and private information on-device. NXP has a long history of delivering edge solutions, and our latest i.MX 95 family takes it to a new level.
Our i.MX 95 applications processors deliver performance, security and more to the edge. Learn why the i.MX 95 family is right for your next AI design.
The i.MX 95 family delivers serious processing power for both conventional machine learning (ML) and generative AI applications at the edge. The i.MX 95 family combines NXP’s eIQ® Neutron neural processing unit (NPU) for machine learning acceleration with immersive 3D graphics powered by an Arm® Mali graphics processing unit (GPU) and an NXP-developed image signal processor. This powerful configuration is scalable for everything from video conferencing systems to large machines in factories.
Here are a few AI/ML applications for which the i.MX 95 microprocessor unit (MPU) is well-suited:
What really differentiates use cases built with the i.MX 95 family is not only the performance of the processors, but also their energy efficiency. The eIQ® Neutron NPU is a capable and efficient AI inference engine, providing competitive TOPS-per-watt at the edge. Our energy should flex multi-domain architecture has separate processing domains for safety, low-power and real-time applications. This offers fine-grained control of dynamic power and low-power operating modes, so the processors can still operate when a large portion of the SoC is powered off.
The NPU, image signal processor (ISP) and video processing unit (VPU) in the i.MX 95 SOC all work together to offer impressive machine vision capabilities at the edge. From home control panels and video surveillance of entryways to factory automation, our eIQ Neutron NPU supports the latest video processing features. Users can run advanced AI models including scene segmentation, live video upscaling and denoising to provide high-quality video footage on the NPU.
The i.MX 95 family can also process multiple streams of video footage at the same time, without having to send information to the cloud. The built-in image signal processor (ISP) supports up to eight raw camera sensors via 2x mobile industry processor interface-camera serial interface (MIPI-CSI) 4-lane interface with MIPI virtual channels. To provide better image quality in low-light conditions and enhance facial recognition, the ISP supports red green blue-infrared (RGB-IR) cameras, which capture both visible and infrared light simultaneously.
To help our customers simplify the development process, our eIQ® AI software development environment supports popular platforms such as:
The platform also supports different types of neural network structures including convolutional neural network (CNN), multilayer perceptron (MLP), recurrent neural network (RNN), long short-term memory (LSTM), temporal convolutional network (TCN) and large language model/small language model (LLM/SLM).
To further simplify neural network deployment, our eIQ AI/ML development software offers NVIDIA TAO Toolkit application programming interface (API) integration so developers have access to a vast library of pre-trained and fine-tuned AI models to deploy on our i.MX 95 processors.
NXP is also making it easier for developers to customize generative AI models and optimize them for edge applications. We are adding the GenAI Flow to the eIQ development software, which lets customers securely train models on domain-specific knowledge and private data. With this capability, customers can train an LLM on a user manual, for example, so a chatbot can answer users’ questions about specific products.
From the extensive AI capabilities of the i.MX 95 processors to our comprehensive developer tools, the i.MX 95 series unleashes powerful machine vision capabilities at the edge in the smart home, the smart office, vehicles, industrial environments, hospitals and beyond.
If you’re interested in using the i.MX 95 series for your next project and want to learn more, read more details on the i.MX 95 product page.
Also, stay tuned for part three of this blog series where we’ll highlight how the i.MX 95 family offers best-in-class connectivity and security features.
Toby Foster is a senior product marketing manager in NXP’s Business Line Secure Connected Edge, focusing on networking applications and management of the Layerscape family of processors. He has also held roles as system architect and applications engineering manager for network processors. Mr. Foster holds an M.S. and B.S. in electrical engineering from Harvey Mudd College.