Tips
Hello everyone, I am the most chatty Xiaoyueyue from Chipboard Square. Last night I worked hard all night and finally wrote a ten-page PPT, ready to chat with global small and medium-sized enterprises and individual creators about how to layout the artificial intelligence industry. The boss said that Xiaoyueyue’s lunch chicken leg this month is linked to the sales of the boards, so whether Xiaoyueyue can eat well depends on everyone’s care.
Driven by key technologies such as 5G communication, big data, supercomputing cloud, and machine learning, artificial intelligence is accelerating and presenting new features such as deep learning, cross-border integration, human-machine collaboration, collective intelligence opening, and autonomous control, even affecting the economy, society, and international political and economic patterns. In China, artificial intelligence technology is leading the latest round of technological and industrial transformation, making companies like Baidu, Alibaba, Tencent, and Huawei rank among the world’s leading technology companies. Artificial intelligence, with its unique “leading goose effect,” promotes many small enterprises to transform and expand their technology business. Throughout human history, wherever there is a rise, there will inevitably be disputes between new and old forces. The old forces’ self-righteousness, arrogance, and extremism have led to many contradictions and conflicts in the world, and the process of conflict is always bloody, it is nothing more than who oppresses whom or who destroys whom. The United States is quickly advancing protectionism and unilateralism, stealing the fruits of labor from the working people of other countries, and severely promoting the deterioration of international trade, leading to intensified global trade and investment disputes. Thus, the world economy has once again come to a crossroads. Contemporary facts have proven that the development of domestic technology companies can no longer rely on the United States, and certainly cannot rely on fairy emperors. Now, we can only rely on the awakening of practitioners themselves.
Chipboard Square is a fair and just platform. We are the Chipboard Square for electronic enthusiasts all over the world, no matter what your skin color is, our culture only has differences in beauty, but absolutely no superiority or inferiority. We are definitely not “extreme patriots” or “extreme nationalists.” We have also recommended many evaluation boards based on American Qualcomm chips; however, constrained by objective barriers in the international trade situation, whether these boards can be supplied smoothly is not something we can subjectively control. With the development of the US trade war, many startups have come to Chipboard Square to recommend domestic alternatives, so today we summarize and discuss domestic artificial intelligence development boards.
Huawei HiSilicon Kirin 970 Chip Hikey970 Development Board
To be objective, Kirin 970 is introduced as the first product in this article for its objective necessity.
First of all, the technology of this chip itself is advanced enough. It is the world’s first mobile AI computing platform with an independent NPU (Neural Processing Unit) built-in. It uses TSMC’s 10nm process and has 4-core Cortex-A73 2.36GHz and 4-core Cortex-A53 1.8GHz.
Secondly, it can effectively help startups promote the layout of smart robots and smart city project industries. HiKey970 integrates Huawei’s innovative design HiAI framework and other mainstream neural network frameworks, supporting CPU, GPU, AI computing, and also supports hardware acceleration for neural network computing based on NPU, with energy efficiency and performance reaching 50 times and 25 times that of CPU computing respectively.
Click the mini program to purchase Hikey970 AI deep learning development board
Firefly Core-3399Pro-JD4 Development Board
Firefly Core-3399Pro-JD4 development board is based on the domestic Rockchip high-performance AI processor RK3399Pro chip.This chip, like RK3399, adopts the ARM dual-core Cortex-A72 + quad-core Cortex-A53 big.LITTLE architecture, with a main frequency of up to 1.8GHz for the A72 core, and integrates Mali-T860MP4 quad-core graphics processor, which has powerful general computing performance.The Pro suffix indicates that it has been upgraded based on RK3399, with more integrated AI neural network processor NPU, supporting 8Bit/16Bit computing, with computing power reaching 3.0 Tops, meeting the needs of various AI applications such as vision and audio.Firefly Core-3399Pro-JD4 is compatible with various AI frameworks, supporting TensorFlow Lite/Android NNAPI.
AI software tools support the import and mapping, optimization of Caffe/TensorFlow models, allowing developers to easily utilize AI technology. It is suitable for the development of intelligent medical health, edge computing and other devices.
Click the mini program to purchase Firefly Core-3399Pro-JD4
TB-RK3399Pro
Rockchip’s Toybrick series development board TB-RK3399Pro is the original factory public version solution from Rockchip. It is equipped with Rockchip’s latest AI processor RK3399Pro. It adopts the big.LITTLE CPU architecture with dual-core Cortex-A72 + quad-core Cortex-A53, leading in overall performance and power consumption; the quad-core ARM high-end GPU Mali-T860 integrates more bandwidth compression technology, with excellent overall performance, and its on-chip NPU (neural network processor) computing performance reaches 3.0 TOPs. Like the aforementioned Firefly RK3399Pro, the TB-RK3399Pro development board has rich external interfaces, making it an All In One (AIO) board.
Click the mini program to purchase TB-RK3399Pro development board
Maix Go K210 Embedded AI Development Board
MaxiGo is the proud work of the famous domestic AI company Sipeed under the MAIX series. The board is equipped with a Sipeed MAIX module, with the core chip being K210, a dual-core 64-bit RISC-V core. It is suitable for machine vision and microphone array signal processing. Among them, in Sipeed’s latest demo, the face detection speed has reached up to 100 frames per second.GO means to set off immediately; the design idea of this development board is to be ready to use; although the price is cheap, it is fully functional; very suitable for startups to purchase for project prototype development.
Click the mini program to purchase Maix Go K210 embedded development board
Maixduino AI
Maxiduino is equipped with the same Sipeed MAIX core AI module as MaxiGo, suitable for machine vision and microphone array signal processing.
The Maxiduino board adopts an Arduino form factor, equivalent to a different scheme evaluation board of Sipeed MAIX, with an onboard ESP32 wireless module, and supports camera, display, and MEMS microphone expansion.This board is suitable for individual electronic enthusiasts for AI development, with Arduino UNO compatible interfaces that are very suitable for geek users to expand their existing Arduino projects.
Click the mini program to purchase Maixduino AI
Nvidia Jetson Nano
We have said that Chipboard Square is not “extreme nationalists” and will not only push domestic products. In our view, these artificial intelligence development boards embody the sweat and wisdom of the working people; the fruits of labor only have colorful differences, but no national differences. However, whether Trump is an “extreme nationalist” is not for us to decide, let history prove it. Investment has risks, and solution selection must be cautious. Do not have the intention to harm others, but be wary of others. Those who choose Nvidia solutions should always pay attention to the supply chain, keep inventory, and remain highly vigilant.
The next product to discuss is from the American Nvidia, Jetson Nano. The Nvidia Jetson series has always had a good reputation among global AI practitioners. Whether for large state-owned enterprises, small and medium-sized enterprises, or individual researchers, it can provide solutions that meet specific performance and budget requirements. All of these solutions use the same architecture and SDK, allowing for a unified code base and seamless deployment across the entire product portfolio. Among them, Jetson Nano is based on the Maxwell architecture, with 128 CUDA cores, making it a low-cost AI computer with ultra-high performance and energy efficiency, capable of running modern AI workloads, running multiple neural networks in parallel, and simultaneously processing data from multiple high-definition sensors. This makes it an ideal entry-level option for adding advanced AI to embedded products.
Click the mini program to purchase Nvidia’s Jetson Nano
Nvidia Jetson TX2
The Jetson TX2, suitable for edge device AI applications, provides users with excellent speed and energy efficiency for embedded AI computing devices. It adopts NVIDIA Pascal GPU (equipped with 256 NVIDIA CUDA cores), up to 8GB of memory, 59.7GB/s memory bandwidth, and provides a wealth of standard hardware interfaces, perfectly adapting to various products and form factors, achieving a truly meaningful AI computing terminal.
Click the mini program to purchase Nvidia Jetson TX2
Googel Edge TPU Dev Board
The Edge TPU development board is a single-board computer equipped with Edge TPU, from the famous Google. The Edge TPU has been optimized for Tensorflow Lite, supporting INT8 data format, with very little precision loss, and is much faster than FP32 or FP16. In Tensorflow computing, the professionally optimized TPU has much higher performance than the more general GPU, making it the highest performance and fastest solution for running Tensorflow Lite in the embedded field.
Click the mini program to purchase Googel Edge TPU Dev Board
Google AIY Vision Kit
This little toy comes from Google, composed of Raspberry Pi Zero and Vision Bonnet. Among them, the Vision Bonnet is based on Intel Movidius MA2450 artificial neural network processor chip, which can run face recognition at a speed of up to 30 FPS.
Click the mini program to purchase Google AIY Vision Kit
UDOOBolt V8
UdooBolt V8 is based on the AMD Ryzen V1605B chip, which is a powerful APU chip, integrating multi-core Zen processors and Vega GPU. Thanks to the 3D graphics rendering capability of the high-performance Vega GPU, its main application area is the VR and AR branches of artificial intelligence. If your team’s development project is mobile AR and requires control of some external devices, then choosing this platform as an evaluation board is very appropriate.
The AMD Ryzen V1605B used in UdooBolt V8 is not an ARM architecture processor but an x86 architecture processor. The UdooBolt V8 board is equipped with an Arduino microcontroller, allowing users to develop various programs under the x86 operating system and use Arduino to receive data detected by external sensors or control external motors and other devices, providing considerable flexibility in use.
Click the mini program to purchase UDOOBolt V8
****************************
If you have any questions about the ten AI boards
Scan the code to add Chipboard Square customer service WeChat for consultation
****************************
*********************************************