An AI-embedded terminal is a compact edge computing platform designed to provide local processing and AI inference capabilities for intelligent terminal, machine vision, and other edge AI applications. The TVD3376R is built around the Rockchip RK3588 processor and integrates the Rockchip RK1828 AI co-processor to provide additional AI inference acceleration. The RK3588 combines an octa-core CPU with an Arm Mali-G610 MP4 GPU and a 6 TOPS@INT8 NPU, while the RK1828 provides 20 TOPS@INT8 AI computing power and integrates 5GB of high-bandwidth DRAM. The system supports Android and Linux operating systems and combines computing, AI acceleration, display, video input, networking, storage, and peripheral interfaces in a passive-cooling enclosure measuring 200×122×57mm.

Key Features

The TVD3376R uses the RK3588 as its main processor, featuring four Cortex-A76 cores running at up to 2.4GHz and four Cortex-A55 cores running at up to 1.8GHz. Its Arm Mali-G610 MP4 GPU and integrated NPU provide processing resources for edge computing workloads. In addition, the RK1828 AI co-processor delivers 20 TOPS@INT8 AI computing power and includes 5GB high-bandwidth DRAM, with support for inference of models at the 7B level. This combination allows the terminal to handle AI inference tasks locally while using the RK3588 as the primary system processor.

The terminal provides flexible system memory and storage configurations. System memory is available as LPDDR5/5X in 4GB, 8GB, or 16GB configurations, while eMMC storage is available in 32GB, 64GB, or 128GB capacities. The listed ordering configurations include TVD3376R-8G64G and TVD3376R-8G128G, both with 8GB system memory and an operating temperature range of 0-60°C.

Interfaces and System Connectivity

The TVD3376R is designed with a wide range of interfaces for connecting displays, cameras or video sources, network equipment, storage devices, and peripherals. For display output, it provides one HDMI OUT and one DP interface, while one HDMI IN interface is available for video input. This configuration supports applications that require both local display output and video source input.

For network connectivity, the terminal provides two 10/100/1000Mbps RJ45 Ethernet ports. It also includes one SIM interface and reserves one M.2 Key-B interface for optional communication modules. Peripheral expansion includes one TF card slot, two USB 3.0 Type-A ports, two USB 2.0 Type-A ports, and one 3.5mm audio jack. These interfaces allow the system to connect with external devices according to the requirements of the target application.

AI Processing and Working Method

The system combines general-purpose computing with dedicated AI acceleration. The RK3588 acts as the main SoC, handling the core computing functions of the terminal, while the RK1828 serves as an additional AI inference processor. By integrating these two processing platforms, the TVD3376R can be used for local AI processing rather than relying exclusively on remote computing resources.

The RK1828 provides 20 TOPS@INT8 AI computing power and has 5GB of integrated high-bandwidth DRAM, supporting inference of 7B-level models according to the product information. This architecture is suitable for applications where AI inference needs to be performed at the edge, particularly machine vision and intelligent terminal scenarios. The actual workload and model compatibility depend on the software and application environment used with the system.

Applications

The TVD3376R is intended for edge AI, machine vision, and intelligent terminal applications. In machine vision systems, its computing and video interfaces can be integrated with video input equipment and used as part of an edge processing platform. The HDMI IN interface provides a video input channel, while HDMI OUT and DP can be used for display connectivity.

For intelligent terminal applications, the combination of RK3588 computing resources, RK1828 AI acceleration, network connectivity, storage, USB interfaces, and Android/Linux support provides a foundation for developing and deploying edge computing systems. Dual Gigabit Ethernet ports can also support applications requiring multiple network connections.

The compact enclosure and fanless passive cooling design are useful for deployments where a small form factor and reduced mechanical cooling complexity are required. The product supports 12V DC power and has a specified operating temperature range of 0-60°C, making these conditions important considerations when integrating the terminal into an application.

Product Benefits

The main value of the TVD3376R comes from combining a high-performance embedded processor with a dedicated AI co-processor in a compact terminal. Instead of using a separate computing platform and AI accelerator, the system integrates both processing functions into one device. Its 20 TOPS@INT8 RK1828 AI capability, 5GB high-bandwidth DRAM, and RK3588 processing platform provide an integrated hardware foundation for edge AI workloads.

The fanless passive cooling structure eliminates the need for a cooling fan, while the 200×122×57mm enclosure provides a compact form factor. Android and Linux support gives developers flexibility in selecting the operating environment for different edge applications. The combination of network, video, storage, USB, audio, and communication expansion interfaces further simplifies integration with external equipment.

Selection and Installation Considerations

When selecting the TVD3376R configuration, users should consider the required system memory, storage capacity, communication expansion, operating environment, and application workload. The available system memory options are 4GB, 8GB, and 16GB, while eMMC storage options range from 32GB to 128GB. Optional communication modules can be installed through the reserved M.2 Key-B interface according to the required configuration.

The terminal uses 12V DC power and measures 200×122×57mm. Since it uses passive cooling, installation should provide suitable environmental conditions within the specified 0-60°C operating temperature range. The selection of interfaces and expansion modules should also be based on the actual devices and connectivity requirements of the final application.

Conclusion

The TVD3376R provides a compact AI-embedded terminal platform that combines Rockchip RK3588 computing with RK1828 dedicated AI acceleration. With 20 TOPS@INT8 AI computing power from the RK1828, 5GB high-bandwidth DRAM, multiple memory and storage configurations, dual Gigabit Ethernet, HDMI and DP display interfaces, HDMI video input, USB expansion, Android/Linux support, and fanless passive cooling, it provides an integrated hardware platform for edge AI, machine vision, and intelligent terminal applications.