- Maybe you missed it? Review of the Banana Pi R4.
- Announcement of the BPI-CM6 DeepX AI Module: An Overview of Key Features.
SinoVoip launches BPI-CM6 powered with Deepx AI module
[4/11/2025] – Introducing the Banana Pi BPI-CM6, a cutting-edge industrial-grade RISC-V core board expertly crafted for edge computing and AI applications. Equipped with the powerful SpacemiT K1 octa-core 64-bit RISC-V AI CPU, it delivers an impressive 2.0 TOPS of AI computing capability. This makes it an ideal solution for a range of applications, including industrial control, robotics, and advanced edge AI computing.
Banana Pi BPI-CM6 + DeepX AI M.2 module card
Introducing DEEPX, a new player in the Korean AI chip design industry.
DEEPX is a South Korean company that focuses on the development of AI semiconductors and edge computing solutions. Their DeepX DX-M1 AI chip, found in an M.2 2280 module card, supports AI capabilities by allowing video analysis and multi-model processing. This module not only supports sophisticated AI applications but also does so in a compact and energy-efficient design.
Hailo Company introduces the Hailo-8, a cutting-edge AI processor chip that features similar technical specifications to its competitors. With both companies vying for dominance in this densely populated market, we have compiled a clear and concise comparison table based on our research findings for your convenience.
| Chip Name | TOPS | Power Consumption | Software Support |
| DeepX DX-M1 | 25 TOPS | 3-5 watts | DXNN® software development kit (SDK) key components: DX-COM: A model conversion tool to adapt AI models for the DX-M1. DX-SIM: A simulation tool for testing and optimizing models. DX-RT: Runtime software for executing AI models on the DX-M1. DX-NPU Driver: A device driver for seamless hardware integration. DX-APP: Sample code and demos for quick deployment of AI applications. |
| Hailo-8 | 26 TOPS | 2.5 watts | Dataflow Compiler: For offline compilation and optimization of deep learning models. HailoRT: A runtime software package for real-time inferencing on host processors like x86 and ARM. Model Zoo: A collection of pre-trained models compatible with frameworks like TensorFlow and ONNX. TAPPAS: Application examples showcasing pipeline elements and pre-trained AI tasks. |
Interfaces and Ports.
The BPI-CM6 supports a substantial memory capacity of up to 16GB of LPDDR4 and a generous 128GB of eMMC storage, making it a powerful option for various applications, including industrial. Additionally, it features a range of connectivity interfaces, including HDMI, USB, and MIPI CSI/DSI, which enhance its versatility for different uses. Its compatibility with open-source ecosystems further adds to its appeal, allowing for flexibility and facilitating the rapid deployment of AI models in multiple environments. This combination of features makes the BPI-CM6 an excellent choice for developers looking to create robust and efficient AI solutions.
BPI-CM6 (Core Board)
BPI-CM6 (Carrier Board)
Price and Availability
As the product is relatively new, we currently do not have the details available. However, please rest assured that we will promptly update this section with any new information as soon as it is released. In the meantime, feel free to explore the official SinoVoip store on AliExpress.
About
SinoVoip specializes in embedded software and hardware integration. They are known for the Banana Pi series of SBCs and products for IoT, AIoT, and edge computing, including mini-PCs, smart routers, AI modules, and industrial computers.
