Some time ago, as it was believed that the giants were heading toward the next traffic-level entrance, the "smart speaker" became a hot topic in the industry. Echo and Google Assistant took the lead abroad, while domestic players like Xiaodu and others launched their own products, driving the wave of this boom. With over 100 "boxes" on the market, the smart speaker frenzy was in full swing.
However, after the initial hype, the market's attention shifted to providing AI chips for smart speakers. On October 31st, Hangzhou Guoxin, a veteran chip manufacturer known for set-top box chips, held a conference to launch the GX8010 AI chip equipped with an NPU. The key application areas include smart speakers, voice interfaces, and educational toys for preschoolers.
As the ecosystem is handed over to the big players, AI chips are gaining ground through their specialized capabilities.
[Image: A visual representation of the GX8010 chip]
Hangzhou Guoxin CEO Huang Zhijie shared insights during an interview, stating that the company aims to transition from "digital life" to "intellectual life." In early 2016, they established an AI division and began developing AI chips. Technically, some of the communication, signal processing, and video coding technologies used in traditional set-top box chips are also applicable to the new AI chip, offering technical synergy.
With deep learning making major breakthroughs across various fields, the demand for higher computing power has increased. As a result, many companies are focusing on building AI-specific chips. While traditional CPUs can handle linear problems, they are not well-suited for deep neural networks. Guoxin, like many other vendors, is working on creating custom chips to improve efficiency.
The newly released GX8010 features an independently developed NPU called gxNPU, tailored for artificial intelligence. It addresses the inefficiencies of traditional chips in neural network operations, offering optimized performance for such tasks.
Additionally, to address the small memory bandwidth in IoT devices, Guoxin designed a neural network compression engine called NCompressor. This allows for one-button compression of neural networks, significantly reducing memory usage and bandwidth requirements while improving operational speed.
So, what sets the gxNPU apart from Google’s TPU or Huawei’s Kirin 970 NPU? Google TPU is primarily used in servers and is considered cloud intelligence, focusing on massive computing power rather than cost or power consumption. In contrast, the gxNPU is designed for IoT devices, emphasizing terminal intelligence. It includes a neural network compression engine, which reduces memory and bandwidth needs, consumes less power, and is more suitable for various IoT applications.
What does "high performance and low power consumption" mean for terminal intelligence? Guoxin offers solutions like "multi-core heterogeneity" and "multi-level wake-up" to manage dynamic power consumption and standby issues.
Guoxin Product Manager RobotLing explained that multi-core heterogeneity involves optimizing the operating frequency and timing of each module, enabling on-demand use and efficient service. For example, in voice interaction scenarios, the GX8010 operates at 100-200 MHz for offline speech recognition, while the DSP runs at 300-400 MHz for multi-microphone array processing. The CPU dynamically adjusts system load, ensuring efficient operation with minimal power consumption.
The standby issue has long been a challenge for voice-enabled devices. The GX8010’s multi-level wake-up mechanism uses hardware-based detection to determine if there's sound, voice, or a keyword. During standby, the chip employs VAD (Voice Activity Detection) to monitor microphone input. Once a voice command is detected, the DSP initiates noise reduction, the NPU activates word recognition, and if a keyword is identified, the entire system is awakened. This step-by-step approach ensures real-time response while extending battery life.
Leveraging partnerships, Guoxin is targeting the IoT market. At the conference, numerous collaborating companies were present, including Alibaba, Tencent, HKUST, Sibischi, Yunzhisheng, and others. Rokid and Sibischi founders also gave keynote speeches and participated in post-conference visits.
Huang Zhijie emphasized that the chip should be the result of collaboration. Misa, founder of Rokid, noted, “We don’t make chips ourselves, but we don’t need all the technology from top to bottom.â€
In terms of cooperation, Guoxin will work with Rokid on smart speakers and collaborate more closely with Sibischi on smart home and automotive projects.
Huang Zhijie stated that the goal for this chip is to achieve revenue in 2018.
The Hangzhou Guoxin team is composed of alumni from Zhejiang University. CEO Huang Zhijie graduated from Zhejiang University, and RobotLing, the product manager of the chip, is his younger brother from the same university.
In the California Gold Rush of the 19th century, the most profitable people weren’t the gold diggers, but those who made shovels. Similarly, in the current wave of smart speakers and smart homes, the biggest winners may be those producing the “brain†of AI chips.
Fiber Optic Distribution Box,Fiber Optic Breakout Box,Fibre Optic Breakout Box,Fibre Break Out Box
Cixi Dani Plastic Products Co.,Ltd , https://www.danifiberoptic.com