Edge AI Platforms Expand IoT Compute Capabilities

MediaTek has introduced several new processors in its Genio family at Embedded World, expanding its edge AI and embedded computing portfolio. The new platforms—Genio Pro, Genio 420 and Genio 360—target a wide range of IoT applications, from high-performance robotics and machine vision systems to smart home, retail and industrial devices.

The platforms combine CPU, GPU and neural processing capabilities designed to enable on-device artificial intelligence and real-time data processing at the edge.

High-Performance Edge AI with Genio Pro
The Genio Pro series represents MediaTek’s high-performance offering for IoT and embedded applications. Built on a 3 nm process from TSMC, the platform integrates an Arm v9.2 CPU architecture consisting of one Cortex-X925 core, three Cortex-X4 cores and four Cortex-A720 cores.

This configuration delivers up to 260,000 DMIPS of compute performance. Graphics processing is handled by the Arm Immortalis-G925 GPU, providing approximately 3.1 TFLOPS of compute capability.

The platform includes MediaTek’s eighth-generation neural processing unit (NPU) with more than 50 TOPS of AI acceleration. This allows the platform to support machine vision workloads as well as generative AI models with up to seven billion parameters, generating up to 23 tokens per second in large language model inference tasks.

The processor also supports complex sensor and multimedia systems, including connections for up to 16 cameras and up to three 4K displays.

Multimedia and Robotics Development
Genio Pro supports Linux-based operating systems including Yocto, Debian and Ubuntu distributions. It also supports the Robot Operating System 2 (ROS 2) framework, which is widely used in robotics development.

The platform includes support for high-speed memory such as LPDDR5x up to 8533 Mbps, 8K30 video encoding and decoding, and multiple camera pipelines enabling advanced sensor fusion. Connectivity and expansion options include PCIe Gen 4 interfaces, USB connectivity and dual 2.5 Gb Ethernet controllers.

These capabilities enable developers to build systems such as autonomous mobile robots, commercial drones, machine vision platforms and logistics automation equipment.

Mid-Range Edge AI with Genio 420
The Genio 420 platform targets mid-tier IoT systems requiring AI acceleration and multimedia support with lower system cost. Built on a 6 nm process, the platform delivers up to 7.2 TOPS of AI processing capability.

It supports up to 16 GB of LPDDR5X memory and offers flexible display configurations including dual 2.5K displays or a single ultra-wide 5K display. The platform is pin-to-pin compatible with MediaTek’s Genio 720 and Genio 520 processors, allowing device manufacturers to scale products across performance tiers without redesigning hardware.

Edge AI for Value-Tier Devices
MediaTek also introduced the Genio 360 and Genio 360P platforms for value-tier IoT devices requiring local AI inference. The Genio 360 delivers approximately 6 TOPS of AI performance, while the Genio 360P provides up to 8.5 TOPS.

Both processors can support generative AI models with up to two billion parameters, enabling on-device inference for applications such as computer vision, speech recognition and intelligent user interfaces.

The Genio 360 series supports configurations including a single 4K display or dual Full HD displays and uses up to 8 GB of LP4X-3733 memory.

Expanding Edge AI Infrastructure
With the introduction of these platforms, MediaTek is extending its Genio processor lineup across multiple IoT performance tiers. By integrating CPU compute, GPU graphics and neural processing into a single platform, the systems aim to support increasingly complex AI workloads executed directly at the edge.

Edited by Industrial Journalist, Romila DSilva – AI Powered