MediaTek and Rockchip Unveil New Chips to Enhance Edge AI Capabilities

According to reports from Electronic Enthusiasts (by Zhang Ying), since February 2025, DeepSeek has emerged, and the explosion of open-source models (such as DeepSeek and Mistral) has reduced the cost of AI inference at the edge for enterprises and accelerated iteration. Qiu Xiaoxin, Chairman of Aixin Yuan Zhi, stated that edge AI chips are a crucial pivot for leveraging the intelligent world, and in the era of edge intelligence, economical, efficient, and environmentally friendly AI chips will attract more attention from enterprises.Recently, MediaTek, Yuntian Lifa, and Rockchip have successively released the latest edge AI chips and application cases. This article will summarize the latest technological trends and application hotspots represented by these chip products.

MediaTek Launches 3nm AI PC Chip and the World’s First 4K 165HDR TV AI Chip

Recently, international research firm Canalys released a report indicating that 2024 marks a significant shift from traditional PCs to AI PCs, estimating that global AI PC shipments will reach 48 million units this year, accounting for 18% of total PC shipments. By 2025, global AI PC shipments will exceed 100 million units, capturing 40% of total PC shipments. By 2028, global AI PC shipments are expected to reach 205 million units, with a compound annual growth rate of 44% from 2024 to 2028.In the AI PC processor market, the main players currently include Intel, AMD, Qualcomm, and Apple. MediaTek has recently made a strong entry with new products, presenting itself as a new competitor in the industry. It is widely believed that AI PC processors are evolving towards enhanced computing power, reduced power consumption, architectural innovation, integration, and miniaturization, driving advancements in process technology, multi-core and heterogeneous computing, and instruction set optimization at the chip design level.On April 7, MediaTek announced the launch of its first 3nm AI PC chip, the Chromebook processor Kompanio Ultra, fully entering the AI PC field. The product’s biggest selling point is its AI computing power of up to 50 TOPS, which can promote the implementation of generative AI experiences at the edge.MediaTek and Rockchip Unveil New Chips to Enhance Edge AI CapabilitiesIt is reported that the Kompanio Ultra features MediaTek’s eighth-generation NPU, employs a full big-core CPU architecture, and includes an Arm Cortex-X925 processor with a maximum frequency of 3.62GHz, providing excellent multitasking capabilities. MediaTek stated that the Kompanio Ultra fully demonstrates the company’s commitment to breaking through mobile computing performance and efficiency over the years. MediaTek has closely collaborated with Google to ensure that the latest Chromebook Plus enjoys next-generation edge AI capabilities, outstanding performance per watt, and immersive multimedia features.MediaTek indicated that Chromebooks equipped with the Kompanio Ultra are expected to be launched in the coming months. Additionally, MediaTek’s layout in the AI PC field includes co-developing the personal supercomputer Project DIGITS with NVIDIA, which later became NVIDIA’s DGX personal AI supercomputer DGX Spark, touted as the world’s smallest AI supercomputer.On April 9, at the China Electronics Show held at the Shenzhen Convention and Exhibition Center, MediaTek showcased the world’s first 4K 165HDR single-chip TV, the MediaTek Pentonic 800, an AI chip aimed at high-end 4K smart TVs and display devices.

MediaTek and Rockchip Unveil New Chips to Enhance Edge AI Capabilities

Image: MediaTek Pentonic 800Compared to the previous generation, the AI performance of the MediaTek Pentonic 800 has improved by 50%, and memory bandwidth usage has decreased by 60%. The Pentonic 800 chip features MiraVision Pro smart display technology, providing a top-notch visual experience with a variable refresh rate (VRR) of up to 165Hz, supporting various AI image quality enhancement technologies, including AI Super Resolution (AI-SR 3.0), AI-Contrast 2.0, AI scene recognition quality monitoring, and AI object recognition quality enhancement. It can be widely applied in smart TVs, smart projectors, smart displays, commercial displays, and embedded large displays.

Rockchip Launches Integrated Cabin Solution for Large Models at the Edge

Recently, Rockchip showcased its RK3588M integrated edge large model cabin solution to automotive industry partners, along with the RK2118M/RK3308M industry-innovative automotive AI audio processor platform and the innovative combination of RK751M four-channel Class D digital audio amplifier.Rockchip’s automotive business head, Zhu Liang, stated that the 8nm process smart cockpit SoCs RK3588M and RK3576M, combined with mainstream edge large models, provide stronger localized AI processing capabilities for smart vehicles, driving upgrades in in-car multimodal interaction, scene perception, and personalized services. This system achieves three major breakthroughs:1. The multimodal interaction system supports deep integration of voice, vision, and touch, allowing for precise recognition of driving and riding intentions; 2. The scene perception engine realizes personalized service proactive push through environmental semantic understanding; 3. The heterogeneous computing architecture supports seamless switching between QNX and Android dual systems, significantly improving resource scheduling efficiency.

Yuntian Lifa: DeepEdge 10 Chip Platform Empowers Smart Parking

It is reported that Yuntian Lifa’s new generation of self-controlled edge AI inference chip, the DeepEdge10 series, has a maximum computing power of 48 TOPS and supports D2D/C2C Mash interconnection expansion, meeting the deployment needs of large models with hundreds of billions of parameters.Through the powerful multimodal large model and neural network processor platform capabilities, Yuntian Lifa has developed several edge AI products, including smart AI panoramic cameras and smart parking holographic bases, and based on this, created the “Holographic Parking Lot Solution,” providing real-time occupancy status and abnormal occupancy detection, as well as precise recognition of license plates and vehicle types, providing scientific support for parking space allocation.This solution is built on the self-developed DeepEdge10 chip platform, supporting efficient inference of large models on edge devices. In parking lots of the same scale, the required edge devices cost less than half of those from similar large manufacturers, significantly reducing deployment costs.

MediaTek and Rockchip Unveil New Chips to Enhance Edge AI Capabilities

Disclaimer: This article is original from Electronic Enthusiasts, please indicate the source above when reprinting. For group communication, please add WeChat elecfans999,for submission of interview requests, please send an email to [email protected].

More Hot Articles to Read

  • Electronic Skin Breaks Out, Igniting Humanoid Robots

  • The World’s First Two-Dimensional RISC-V MPU! Named “Wuji”, from the Fudan Team

  • Humanoid Robot Perception Revolution! Innovative Machine Vision Sensors Emerge

  • Entering the Thousand Yuan Level! Exoskeleton Robots Sold Out, These Motor Companies Accelerate Positioning

  • Global Foundries and UMC Plan Merger! Challenging Samsung, the Global Wafer Foundry Landscape May Change

Click to Follow and Star Us

Don’t miss any updates by marking us as a star!MediaTek and Rockchip Unveil New Chips to Enhance Edge AI Capabilities

Leave a Comment