AI-2025 World Electronics Achievement Awards-EE Times

Product：BG22L & BG24L Bluetooth LE SoC

Recommended reasons： Optimized for common Bluetooth applications like asset tracking tags and small appliances, the BG22L brings the most competitive combination of security, processing power, and connectivity for high-volume, cost-sensitive, and low-power applications. The BG24L SoC includes the Silicon Labs accelerator for AI/ML applications and support for Bluetooth Channel Sounding for asset tracking and geofencing, even in the most crowded areas like packed warehouses and multi-family housing units. "We know that our customers are always looking for ways to keep pace with the demands of the evolving IoT market while reducing costs," said Ross Sabolcik, senior vice president of the Industrial and Commercial business unit at Silicon Labs. "The BG22L and BG24L expand our portfolio to match those customer needs and concerns by providing an optimized set of industry-leading Bluetooth features with our signature IoT capabilities like high RF sensitivity, low power, robust security, and powerful compute."

More Vote

Telink Semiconductor（Shanghai） Co., Ltd.

Product：Telink TL-EdgeAI Development Platform

Recommended reasons： (1) Superior Chip Foundation Telink's new-generation chips underpin the TL-EdgeAI platform. The TL721X, upgraded from TLSR9, offers a low working current of 1 mA for high-performance IoT devices. The TL751X, featuring a multi-core design and HiFi-5 DSP, supports diverse applications. (2) Significant Low-Power Advantage As one of the world's lowest-power smart IoT connectivity platforms, it is particularly suitable for battery-powered products. It effectively addresses the challenge of device endurance, ensuring continuous and stable operation of various intelligent devices. (3) Powerful AI Capabilities The TL-EdgeAI platform supports LiteRT, TVM, and other models, and can convert models from TensorFlow, PyTorch, and JAX. It enables autonomous learning and supports large and small models, expanding application possibilities. (4) Convenient and Efficient Development The TL-EdgeAI platform facilitates quick model porting, lowers development costs, and accelerates time-to-market. Using Telink's ML/AI SDK, users can easily integrate trained models into products, especially with LiteRT.

More Vote

Beijing Smart-chip Microelectronics Technology Co., Ltd.

Product："Falcon" Artificial Intelligence Visual Integrated Chip

Recommended reasons： The "Falcon" visual-integrated artificial intelligence chip is an industrial-grade AI chip specifically designed for ultra-high-definition 4K intelligent vision application scenarios. It integrates functions such as AI recognition, image capture, long/short video processing, and audio playback. With an on-chip neural network accelerator, it can support computing acceleration for multiple types of neural networks. Its unique RCU (Reconfigurable Computing Unit) enables reconfigurable computing, achieving technological breakthroughs in algorithm compatibility, computational energy efficiency, imaging quality, and stability and reliability. Boasting a computing power of 4 Tops, its energy efficiency is 40% higher than that of similar products. It can operate stably within a working temperature range of -40°C to 85°C, and can perform targeted optimizations for functions like depth of field, night vision, backlight handling, wide-angle shooting, and recognition algorithms required in scenarios such as power transmission and transformation, thus realizing domestic substitution for similar products.

More Vote

Witmem

Product：Witmem CIM Chip WTM8600

Recommended reasons： With the advancement of edge-side AI technology, AI/VR/XR glasses are no longer simple display devices but have evolved into intelligent terminals equipped with on-board AI processing capabilities. As a result, consumer expectations regarding the AI performance of such devices have significantly increased. On one hand, the required AI computing power has risen from an initial 1–2 TOPS to approximately 10 TOPS in order to support local inference for larger and more complex algorithm models. On the other hand, display technologies continue to improve, with VR devices now commonly featuring resolutions approaching 2K or 4K and refresh rates ranging between 90Hz and 120Hz. However, mainstream solutions based on the traditional von Neumann architecture have increasingly revealed performance limitations due to the growing volume of data being transferred back and forth, which negatively impacts both latency and power consumption. To overcome this challenge, Witmem Technology has developed the world's first display enhancement coprocessor chip—WTM8600—based on 3D CIM architecture, with sampling scheduled for 2024. This solution offers significant advantages in terms of computing power, energy efficiency, and cost-effectiveness. While typical market solutions provide less than 10 TOPS of computing power, the WTM8600 delivers up to 24 TOPS. When integrated with standard control chips and running multiple algorithms—including AI super-resolution and AI frame interpolation—the system maintains efficient computing performance while keeping power consumption within approximately 1W (depending on the specific algorithm). By offloading tasks from the main control chip and reducing overall power consumption, the WTM8600 significantly enhances visual performance, doubling the display frame rate from 60FPS to 120FPS and upgrading resolution from 2K to 4K. Product Specifications: • Computing Power: Up to 24 TOPS @ INT8 • Computing Energy Efficiency Ratio: >40 TOPS/W • Storage-Computation Array Size: 32MB (supports storage of multiple and larger AI models) • Computing Precision: Supports 8-bit, 10-bit, and 12-bit fixed-point operations • Package and Dimensions: FCBGA package, 15mm x 15mm, utilizing 3D packaging technology • Supported AI Models: AI super-resolution, AI frame interpolation, resampling, YoloV5, ResNet, and others

More Vote

Quectel Wireless Solutions Co., Ltd.

Product：AIFex AI Algorithm Platform

Recommended reasons： In 2024, Quectel launched its new industrial intelligence brand, Provectron, along with its core products: the AI algorithm platform AIFex and AI visual solutions. This marks Quectel's accelerated efforts to drive the comprehensive upgrading of AI technology empowerment in industrial intelligence. AIFex is a deep learning-based AI algorithm model training platform that integrates full-process algorithm development functions, including image upload, annotation, model training, testing, and deployment. It focuses on solving problems like complex defect detection and classification, and is applicable to various vertical application scenarios. Its core algorithm functions include image detection, image segmentation, image classification, OCR, and positive sample learning. The platform is integrated with an AIGC defect generator, which can quickly produce high-simulation defect images in batches based on real defects, with automatic annotation. This helps solve issues such as difficulties in model training and verification, and poor model indicators caused by insufficient defect data, improves the generalization ability of the model, and reduces the data cost of obtaining high-quality models. The model supports functions such as format conversion across multiple platforms and one-click deployment to inference terminals. It is also compatible with various mainstream x86 and ARM architecture inference terminals on the market, greatly lowering the threshold for AI algorithm development and application, and accelerating project implementation. At present, the Provectron AIFex algorithm platform has been widely applied in various scenarios such as AI visual inspection, AI sorting, and AI behavior monitoring.

More Vote

XMOS

Product：XMOS xcore Edge AI solution with AI, DSP, MCU and I/O integrated in a single chip

Recommended reasons： XMOS的xcore^®系列软件定义SoC拥有完善的架构，可提供高性能DSP、AI加速器、MCU和I/O。这些高性能、低延迟和低功耗处理器可以根据客户的需求进行定制化，以适用于多样化的边缘AI场景，其最终可配置能够完全通过软件来打造，可用于消费电子、工业和汽车等领域。目前，XMOS已利用其xcore技术推出了全面的边缘AI音视频解决方案。 ·XMOS的3D空间音频技术可在任何设备上都提供即插即用、比特精确、丰富的3D沉浸式音频体验，不仅可以支持设备间的传输，而且减少了一半的耳鼓压力并实现所有延迟超低。该技术与所有的操作系统和耳机都可无缝兼容，同时满足游戏对超低延迟的要求。 ·XMOS的AI降噪技术采用了深度神经网络（DNN）算法，以实时方式智能化地去除背景噪声，从而确保在极具挑战性的环境中也能够清晰地捕获音频。具有的特性包括1-4个麦克风阵列远场语音、可提供超级指向性自适应波束形成、回声消除、回声与噪声抑制、人工智能降噪和强大的去混响能力、可选的热词侦测器。 ·XMOS的多模态AI模型处理功能支持音频、图像、视觉和其他多种传感信号，可在最具挑战性的应用场景中，实现实时、持续工作的AI应用，同时保护个人数据和隐私；也可以根据需要，灵活地作为大模型、云和网络的接口，提供传感器信息预处理，可用于人脸检测、特征提取、身份验证、图像分类、离线本地自主运行、智能传感器接口。

More Vote