MediaTek unveiled the high-performance Genio 720 and Genio 520 edge-AI IoT systems, enabling popular IoT applications with a potent Neural Processing Unit for cutting-edge edge AI. The newest generative AI models, human machine interface (HMI), multimedia, and networking features for smart home, retail, industrial, and commercial IoT devices are supported by these new Genio series updates.
With a built-in MediaTek 8th-generation Neural Processing Unit (NPU) that provides complete hardware acceleration for transformer and convolutional neural network (CNN) models, the Genio 720 and Genio 520 offer exceptional edge performance, supporting up to 10 TOPS. With up to 16GB of LPDDR5 memory, the Genio 720 and Genio 520 also provide improved memory performance to enable data-intensive edge-optimized large language models (LLMs) like Llama, Gemini, Phi, and DeepSeek and speed up generative AI activities. With the support of MediaTek’s vast global AI ecosystem, developers may quickly bring products to market by utilising common frameworks and top global language models to effectively implement multi-modal generative AI applications.
With strong yet effective generative AI capabilities on-device, the Genio 720 and Genio 520 will unleash a new era of Internet of Things innovation, offering a seamless user experience that also ensures data privacy, “Developers can scale their designs across various products and form factors, from smart retail displays to complex industrial HMI applications, with integration of NVIDIA TAO and widely-adopted AI models for Edge AI applications.”
To maximise speed and power efficiency, the Genio 720 and Genio 520, which are built on a 6nm production node, combine an octa-core CPU with two Arm Cortex-A78 processors and six Cortex-A55 processors. Because to their low power consumption optimisation, these incredibly effective IoT systems are appropriate for mobile devices that need battery-operated operation and fanless enclosures.
The Genio 720 and Genio 520 are perfect for commercial displays, smart retail devices, HMI applications, and other multi-window or interactive applications because of their many sophisticated multimedia features. Developers can easily code once and have it run anywhere because to these IoT platforms’ unified hardware and software design, which also provides customisation options to suit particular application requirements.
The Open Standard Module (OSM) development cycle for users is significantly shortened by Genio 720/520’s support for OSM, which includes reference designs that guarantee power and signal integrity. In the second half of the year, MediaTek partners want to introduce OSM solutions built on the Genio 720/520, which can speed up time-to-market for a variety of applications.
MediaTek Genio 720
Broad peripheral connections and an integrated Neural Processing Unit (NPU) make this high-performance Gen-AI platform suitable for a variety of smart edge devices. A potent Internet of Things platform with sophisticated multimedia features and generative AI capabilities is the MediaTek Genio 720. It is compatible with Android, Yocto, and Ubuntu operating systems and offers sophisticated interactive HMI, making it adaptable to a wide range of applications in the retail, commercial, industrial, and smart home sectors.
MediaTek Genio 520
In addition to its advanced multimedia capabilities, intelligent interactive HMI, and compatibility for Android, Yocto, and Ubuntu operating systems, the MediaTek Genio 520 offers a high-performance, generative AI-capable Internet of Things platform. It is made for a variety of uses, including as retail, commercial, industrial, and smart home applications.
Highlights of the Genio 720 and Genio 520 platforms
- Incredibly effective 6nm SoC
- An octa-core CPU with two Arm Cortex-A78 processors operating at 2.6 GHz
- Best-in-class 10 TOPS performance (NPU+CPU+GPU) is possible with Gen-AI.
- LPDDR5 memory up to 16GB, perfect for LLMs
- Single or twin active displays with up to 5K high-detail
- Decoding and encoding 4K HEVC video
- Six FHD30 cameras with MIPI virtual channels are possible.
- High-speed, adaptable I/O interfaces
- MediaTek Wi-Fi 6/6E solutions that are already integrated, with support for Wi-Fi 7 and 5G Redcap modules
- Options for Ubuntu, Yocto Linux, and Android
- Open Standard Module (OSM) design assistance
- Broad support for temperatures (-40 to 105°C)
- Extended commitment to product life
Important Features of the Genio 720 and 520
Generative AI NPU
Full hardware acceleration for Transformer and Convolutional Neural Networks (CNN) models is provided by the 8th Generation NPU, and data-intensive Small Language Model (SLM) and Large Language Model (LLM) activities are supported by improved memory performance.
Diverse Display Support
The platform is perfect for commercial displays, smart retail, HMI, and other interactive or multi-window applications since it supports a wide range of display outputs, including dual active displays, LVDS (FHD), MIPI (2.5K), eDP 1.4 (4K60), and DP over Type-C (5K60 ultrawide).
Advanced Connectivity
For Internet of Things devices, multi-gigabit, tri-band Wi-Fi with the option of a 5G RedCap module provides greater bandwidth, dependability, and the advantage of always-connected wireless.
Extremely Power Efficient
Because to its low power consumption architecture, this effective platform is appropriate for mobile devices that need battery operation and fanless enclosures.
Genio 720 and 520 Specifications
| Specification | Genio 720 | Genio 520 |
|---|---|---|
| CPU | 2x Arm Cortex-A78 (up to 2.2GHz) 6x Arm Cortex-A55 (up to 2.0GHz) | 2x Arm Cortex-A78 (up to 2.2GHz) 6x Arm Cortex-A55 (up to 2.0GHz) |
| Memory Support | Up to 16GB LPDDR4X (up to 4266Mbps) LPDDR5 (up to 6400Mbps) | Up to 16GB LPDDR4X (up to 4266Mbps) LPDDR5 (up to 6400Mbps) |
| Storage Options | UFS 3.1 (2-lane) eMMC 5.1 | UFS 3.1 (2-lane) eMMC 5.1 |
| AI Capabilities | Integrated 8th Generation NPU Supports up to 10 TOPS Full hardware acceleration for transformer and convolutional neural network (CNN) models | Integrated 8th Generation NPU Supports up to 10 TOPS Full hardware acceleration for transformer and convolutional neural network (CNN) models |
| Graphics Processor | Arm Mali-G57 MC2 (up to 880MHz) Supports OpenGL, OpenCL, Vulkan | Arm Mali-G57 MC2 (up to 880MHz) Supports OpenGL, OpenCL, Vulkan |
| Display Support | Supports one 4K/5K ultrawide display or dual 2.5K displays | Supports one 4K/5K ultrawide display or dual 2.5K displays |
| Video Encoding | 4K at 30fps (HEVC/H.264) | 4K at 30fps (HEVC/H.264) |
| Video Decoding | 4K at 60fps (HEVC/H.264) | 4K at 60fps (HEVC/H.264) |
| Peripheral Interfaces | 2x USB 3.2 Gen-1 (Host, 1 shared with DP) 2x USB 2.0 (Host) 1x USB 2.0 (Host/Device) 1x Gigabit Ethernet MAC 1x PCIe Gen2 4x UART 9x I2C 6x SPI 4x PWM | 2x USB 3.2 Gen-1 (Host, 1 shared with DP) 2x USB 2.0 (Host) 1x USB 2.0 (Host/Device) 1x Gigabit Ethernet MAC 1x PCIe Gen2 4x UART 9x I2C 6x SPI 4x PWM |
| Audio Interfaces | Audio-In: 2x I2S (2ch, M+M/S), 1x PCM (2ch), 2x DMIC (2ch) Audio-Out: 2x I2S (2ch, M), 1x PCM (2ch) | Audio-In: 2x I2S (2ch, M+M/S), 1x PCM (2ch), 2x DMIC (2ch) Audio-Out: 2x I2S (2ch, M), 1x PCM (2ch) |
| Wireless Connectivity | Integrated Wi-Fi 6 1×1 / 6E 2×2 Wi-Fi 7 module available Bluetooth 5.3 | Integrated Wi-Fi 6 1×1 / 6E 2×2 Wi-Fi 7 module available Bluetooth 5.3 |
| Camera ISP | Internal single ISP Supports up to 16MP + 16MP at 30fps | Internal single ISP Supports up to 16MP + 16MP at 30fps |
| Operating System Support | Android Yocto Ubuntu | Android Yocto Ubuntu |
| Package Type | VFBGA 13.8×12.8mm Ball pitch: 0.4mm | VFBGA 13.8×12.8mm Ball pitch: 0.4mm |
