KL730 AI Soc
The KL730 chip integrates the most advanced third-generation KDP series reconfigurable NPU architecture, providing up to 8 TOPS of effective computing power.
The new NPU architecture not only offers higher computational efficiency for the latest CNN network architectures but also performs well in transformer applications, significantly reducing the demand for DDR bandwidth. The KL730 has powerful video processing capabilities, supporting 4K 60FPS output. Leveraging the team's 10+ years of experience in the ISP field, the KL730 demonstrates exceptional performance in areas such as noise reduction, wide dynamic range, fisheye correction, and low-light imaging. It supports a wide range of applications in the following areas: intelligent security, autonomous driving, video conferencing, industrial cameras, commercial robots, and CMS electronic rear-view mirrors.
The new NPU architecture not only offers higher computational efficiency for the latest CNN network architectures but also performs well in transformer applications, significantly reducing the demand for DDR bandwidth. The KL730 has powerful video processing capabilities, supporting 4K 60FPS output. Leveraging the team's 10+ years of experience in the ISP field, the KL730 demonstrates exceptional performance in areas such as noise reduction, wide dynamic range, fisheye correction, and low-light imaging. It supports a wide range of applications in the following areas: intelligent security, autonomous driving, video conferencing, industrial cameras, commercial robots, and CMS electronic rear-view mirrors.
- Quad ARM® Cortex™ A55 CPU.
- Built-in DSP for accelerating AI model post-processing and speech processing.
- Linux and RTOS, built on TSMC's 12-nanometer process.
- Up to 4K@60fps resolution, seamless RGB Bayer interface with mainstream sensors, up to 4 channels of image interfaces.
- Up to 3.6eTOPS@int8 / 7.2eTOPS@int4.
- Supports frameworks like Caffe, TensorFlow, TensorFlow Lite, PyTorch, Keras, and ONNX.
- Compatible with various AI models such as CNN, Transformer, RNN Hybrid, offering higher processing efficiency and accuracy.
KL630 AI SoC
The KL630, Kneron’s next generation AI chip, leads the way with cutting edge NPU architecture. It is the first solution in the industry to support Int4 precision as well as transformer neural networks, enabling exceptional compute performance efficiency and low energy cost. The KL630 is suitable for diverse range of use case scenarios.
- Based on ARM® Cortex™ A5 CPU, enabling high compute efficiency and low energy expenditure.
- Kneron’s third generation NPU, with compute power up to 0.5eTOPS@int8/1eTOPS@int4.
- Supports all major AI frameworks including Caffe、Tensorflow、Tensorflowlite、Pytorch、Keras、ONNX.
- Superior ISP supports 5M@30FPS，120db HDR, low light conditions, anti-blur, as well as panaroma fisheye correction.
- Suitable for various edge AI devices and wide-angle camera scenarios including smart security, cities, autonomotives.
KL530 AI SoC
The KL530 is Kneron’s most up-to-date heterogeneous AI chip enabled with a brand new NPU architecture. It is the first chip in the industry to commercially support INT4 precision and transformers. Compared to other chips, it possesses higher computing efficiency and lower power consumption. AIoT and other scenarios.
- Low-power design enabled by ARM Cortex M4 CPU.
- The computing power of the KL530 runs up to 1 TOPS@INT 4, and its processing efficiency is up to 70% higher than that of INT 8 under the same hardware conditions.
- Support for various AI models such as CNN, Transformer, RNN Hybrid.
- The chip’s smart ISP can optimize image quality through AI inferencing; and its powerful codec can achieve high-efficiency multimedia compression.
- The chip’s cold start time is less than 500 ms, and its average power consumption is less than 500 mW.
KL720 AI SoC
With up to 0.9 TOPS per watt, our latest chip redefines the limits of edge AI performance. Engineered for real-world intelligence at the edge, it unleashes blazing-fast AI processing wherever efficiency, scalability, and power matter most—turning every device into a smart, autonomous powerhouse.
Delivering 2–4× greater power efficiency than competing solutions at a more accessible cost, the KL720 is Kneron’s next-generation edge AI SoC, designed for flagship devices with demanding AI workloads.
Key Features:
All of these capabilities and more can run simultaneously on the KL720, delivering powerful AI performance directly at the edge while maintaining exceptional energy efficiency.
Key Features:
- Low-power architecture powered by an ARM Cortex‑M4 CPU, enabling efficient on-device AI processing
- Ideal for high-end devices, including IP cameras, smart TVs, AI glasses and headsets, and AIoT gateways
- Advanced vision processing, capable of handling 4K images, Full HD video, and 3D sensing for highly accurate facial recognition and intuitive gesture control, perfect for gaming systems, retail kiosks, and interactive devices
- Natural Language Processing (NLP) support for translators, voice assistants, and conversational AI applications
All of these capabilities and more can run simultaneously on the KL720, delivering powerful AI performance directly at the edge while maintaining exceptional energy efficiency.
KL520 AI SoC
Our first-generation chip put us on the map with a combination of size, power savings, and performance not seen in the edge AI market. The proof is in a number of customers going to market with KL520-powered products.
The KL520 with its dual ARM Cortex M4 CPU can operate as a host or companion AI co-processor enabling edge AI on devices like smart locks, security cameras, drones, smart home appliances, and robotics. It is compatible with various 3D sensor technologies such as Structured Light, dual-cameras, ToF camera, and Kneron’s own exclusive 3D sensing technology.
- This chip is perfect for smart home devices like smart door locks, smart doorbells, IP cams, robot vacuums, etc
- Small power footprint allows some devices to run on 8 AA batteries for fifteen months