With the rapid development of industry informatization and the diversification of application requirements in different industries, Techvision Intelligent Technology Co., Ltd launched the first SMARC2.0 computer module withRK3399 multi-core processor series TVI2317A. The module based on 64 bit dual core Cortex-A72 + Quad coreCortex-A53 is a high-performance module designed for low-power embedded computer. Users can develop carrier boards with different functions according to the application requirements of different industries, and carry out rapid secondary development. They can also select different core boards in the same model, and distinguish products of different grades in terms of performance to meet the application of different industries.
The calculation module can select commercial version and industrial version according to the application environment of customers. The commercial version is 0 ℃ ~ 80 ℃, which can be used for general indoor application; the industrial version is - 20 ℃ ~ 85 ℃, which can be used for outdoor harsh environment, such as traffic control, industrial equipment, etc. In addition, it can provide 10 year life cycle supply.
Powerful operation performance, CPU: 64 bit dual core Cortex-A72 + Quad core Cortex-A53, with the highest main frequency of 1.8G, GPU：RK3399 Mali-T864 GPU,Support OpenGL ES3.2, Vulkan 1.0, OpenCL 2.0
*The scores of different versions of Antutu will be different;
*Different hardware environments will be different;
Excellent multimedia performance: it supports 4K 10-bit H264 / H265 decoding, 4K VP9 decoding supports multi-channel video Co coding and co decoding; it can be applied to conference system and electronic whiteboard, one channel coding and four channel decoding; it can also be used for UAV coding and transmission to the ground at the same time.
In terms of embedded AI,*Test condition: Android calls MPP (JNI native) to decode local bitstream file
1. CPU / GPU computing unit is universal and has strong compatibility. GPU can provide 102.4 GFLOPS computing power
2. Tensorflow LITE / Android NN with optimization (Android 8.1)
3. GPU supports Native FP16 operation: double operation performance, reduce bandwidth and energy consumption, and the model does not need additional processing
4. GPU performance allocation mechanism: assign the priority within the application, and allocate the GPU rendering / operation ratio at the system level to meet the requirements of visual, audio and other entry-level AI capabilities: > 20fps image recognition, > 10fps universal target detection; for example, offline translator, access control gate, face recognition and human flow statistics, Rockchip provides technical support: model performance analysis and suggestions:
Rich and varied IO expansion performance: the module has 2 USB 3.0, 2 USB 2.0, PCIe expansion bus, I²C expansion bus, I²S bus, SPI bus, URAT signal, GPIO signal, 1 Mipi_DSI, 1 Mipi_CSI, 1 Mipi_DSI / CSI optional, 1 HDMI 2.0 signal, 1 Gigabit Ethernet signal.In terms of display performance, it can support two-way and three-way asynchronous display: the maximum of dual screen different display can support 4K + 2K different display; the maximum of three-way asynchronous display can support 1080p + 1080p + 2.5k; the dual screen / three screen splicing display supports horizontal splicing, and the proportion of vertical splicing is flexibly configured;
On the system, Linux supports Debian version and Android 7.1 system. The original chip manufacturer ROCKCHIP will maintain the Android version for a long time. The system software can provide some customized features to meet the needs of applications in different industries.
1. Mic array support (8CH IIS, supporting audio 3A algorithm)
2. Multi camera framework support (2-way Mipi CSI, 8-way USB camera and above *)
3. Multi screen splicing display，and support multi screen display ( each display can play different content )
4. Single / dual HDMI in
5. Dual Ethernet (dual network card bridging, external network access), static IP settings
6. Dual Wi-Fi (AP + station, AP + direct)
7. Upgrade security (RK AB update, AB update)
8. Fast reverse processing, cold start for 2.5s, hot start for 0.03s
9. Trapezoid correction algorithm
10. Camera start speed (fastest ~ 1s)