According to official news from Rockchip: On March 6, the international forum “AI in 5G – Leading the New Era Forum” was held in Shenzhen, co-hosted by the International Telecommunication Union and the China Academy of Information and Communications Technology, with the Artificial Intelligence Industry Development Alliance (AIIA) and ZTE Corporation as the organizers. At the meeting, AIIA officially released the results of the first round of evaluations for the “AIIA DNN benchmark V0.5”, with the Rockchip RK3399 passing the evaluation as a product in the first round.
The AIIA DNN benchmark V0.5 testing evaluation mainly targets edge devices, based on the authoritative testing platform of AIIA, objectively reflecting the performance of processors with deep neural network acceleration capabilities in completing inference tasks across four typical application scenarios. AIIA’s overall team leader, Sun Mingjun, stated that the first round of evaluations includes four typical scenarios and two major categories of evaluation indicators, which include speed (fps) and algorithm performance, such as top1, top5, mAP, mIoU, PSNR, etc. This is the first benchmark in the field of deep learning processors to distinguish between integer and floating-point comparisons.
The 28nm Rockchip RK3399 AI chip participated in the first round of evaluations, and the evaluation results showed that with a maximum accuracy loss of 1% in int8 computation, floating-point models did not require retraining for quantization, achieving twice the performance compared to floating-point computation.
In addition, in the Interpretation evaluation, AIIA attempted for the first time to separate the evaluation of quantized and floating-point models in the benchmark tests. The Rockchip RK3399 development board also achieved several excellent results, leading the industry. The robust performance of the Rockchip RK3399 development board is also attributed to Tengine, a lightweight modular high-performance neural network inference engine developed by OPEN AI LAB, specifically optimized for Arm embedded devices, and does not rely on third-party libraries, supporting cross-platform use for Android and Linux.
In the innovation and research and development of AI devices such as IoT devices, smart interactive devices, and robots, Rockchip has demonstrated a leading technological advantage and ranks in the Top 20 of the global AI chip list. In addition to the RK3399, it also has several high-performance, low-power ARM chips such as RK3288 and RK3128.
Based on Rockchip’s strong R&D capabilities, the chip’s high performance, low power consumption, high stability, and market acceptance, as well as the future direction of AI industrialization development. Since its establishment, Guangzhou Xiangcheng Electronics has been committed to providing embedded industry Android solutions for Rockchip products. After several years of effort, it has formed a comprehensive Android product solution series based on RK3128, RK3288, and RK3399, widely used in:
Smart Security: Intelligent security, facial recognition turnstiles, smart parking lots, video (capture) monitoring, visitor machines, facial access control.
Commercial Display: Advertising machines, queue machines, digital nameplates, vending machines, hotel self-service integrated machines, self-service inquiry machines, medical department queue machines, ordering machines, cash registers, identity verification, ticket machines.
Medical Equipment: Medical imaging, POCT instruments, fluorescence immunoassay analyzers.
Multimedia Audio-Visual: Smart conference systems (paperless conference machines, conference projectors, remote conference machines, recording and broadcasting integrated machines), thin client machines.
Gaming and Animation: Animation, gaming motherboards (dart machines, online claw machines, Honor of Kings), virtual reality (VR), augmented reality (AR), video live streaming, karaoke machines.
Our product advantages:
1. Product advantages: A. The motherboard hardware and system support Alipay facial recognition cameras. While others are still in the exploratory phase, we already have applications, choosing us puts you one step ahead. B. Wired network supports routing functions (LAN functionality), 4G and RJ45 can be used simultaneously. Network connections are more flexible, and network monitoring cameras can be connected freely. C. Supports Android 5.1, Android 6.0, and Android 7.1 systems.
2. Rich functional interfaces: For example, the XC3288v1.3 has up to 4 serial ports, 7 USB ports, wired, Wi-Fi, LVDS, HDMI, and other interfaces, making it easy to expand peripherals.
