AI accelerator DEEPX DX-M1 – As the demand for intelligent edge computing accelerates, Virtium Embedded Artists is breaking new ground with the launch of the iMX8M Mini DX-M1—hailed as the industry’s first System-on-Module (SOM) featuring an integrated DEEPX DX-M1 AI accelerator. This pioneering move eliminates the need for a separate AI module, giving OEMs and embedded system designers a streamlined path to deploy high-performance AI capabilities without overhauling existing hardware.
In this conversation, David Beasley, Executive VP of Sales and Marketing at Virtium and Anders Rosvall, CEO of Embedded Artists share the motivations behind this breakthrough, the engineering challenges their team tackled, and how this compact, power-efficient solution is enabling faster time-to-market, simplified system design, and scalable product strategies for edge AI applications across industries.
The iMX8M Mini DX-M1 is positioned as the industry’s first SOM with an onboard hardware AI accelerator. What drove the decision to integrate the AI accelerator chip directly onto the SOM, and what unique challenges did your team face during development?
David Beasley: Many of our customers are looking for ways to quickly integrate AI capabilities into their existing Edge compute platforms. The challenge is that many of these platforms weren’t originally designed with the flexibility to support AI acceleration hardware—adding it often requires a complete board redesign. To address this, we engineered a solution that helps customers overcome those limitations and significantly accelerate their time to market.
Anders Rosvall: We’ve dedicated extensive engineering effort to develop and verify this robust and scalable solution. It’s a highly integrated design with 900+ components with LPDDR4 & LPDDR5 memory subsystems and other high-speed interfaces.
As for the decision to integrate the AI accelerator on the SOM, it’s about scalability and ease of integration. One carrier board design can support a scalable solution – just mount a SOM with or without an AI accelerator, depending on your need for product differentiation.
With 25 TOPS of AI processing throughput and just 5W average power consumption, how does the DX-M1 NPU compare to existing solutions in edge AI performance and efficiency?
Anders Rosvall: To the best of our knowledge, the DEEPX DX-M1 NPU/AI booster really redefines the landscape of edge AI, achieving unparalleled cost-efficiency (inference/$), power-efficiency (TOPS/W), and performance efficiency (FPS/TOPS).
David Beasley: For customers aiming to rapidly integrate AI capabilities—especially around sensors and cameras—our solution delivering 25 TOPS at just 5W is an ideal fit when paired with the NXP i.MX8 Mini processor. It offers a compelling combination of power, performance, and efficiency. This not only enhances system reliability but also reduces the effort required to manage thermal issues, all while maintaining a lower overall cost compared to many alternatives on the market.
This SOM is available in two versions—one with the DEEPX AI accelerator and one without. How do you see OEMs deciding between these two configurations based on their target applications?
David Beasley: The NXP i.MX8 Mini application processor is a proven, cost-effective choice for many Edge compute platforms. By offering an optional AI accelerator, customers can create two distinct product tiers using the same hardware SOM footprint—one standard and one enhanced with AI capabilities. This approach enables greater flexibility without adding complexity to the BOM or supply chain, making it easier to scale and adapt to evolving market demands.
Space-saving and design simplification are major selling points of the iMX8M Mini DX-M1. Can you share some examples of how this SOM reduces the bill-of-materials and system complexity for developers?
Anders Rosvall: The alternative to buying our product is sourcing 900+ components. Getting to start a development project with a proven design that is also very well-supported reduce time to market as well as lower the project risk.
From your perspective, which industries or applications stand to benefit the most from this solution, and how have early adopters responded to the capabilities of the iMX8M Mini DX-M1 so far?
David Beasley: The solution is ideal for customers who manage hardware platforms in edge compute environments and where power efficiency is an important aspect of their design.
- Key application areas include:
- Edge Cameras & Vision Systems
- Drones and autonomous aerial systems
- Security and surveillance systems
- Industrial Computing & Smart Factories
- Automated inspection and industrial monitoring
- Smart transportation and fleet systems (using cameras and sensors)
- Autonomous Robotics
- Medical Technology (e.g., portable diagnostic equipment)
- Agriculture Technology (e.g., precision farming and crop monitoring)
- Edge Cameras & Vision Systems
