Yesterday, Ziguang Zhanrui officially released the new 5G SoC T820 chip, which adopts 6nm technology, octa-core CPU architecture, built-in financial-level security solutions, supports 5G high-speed connection, and 5G dual-card dual-standby.

In terms of performance, Ziguang Zhanrui T820 is built by TSMC's 6nm EUV process. It has a 1+3+4 three-cluster eight-core CPU. 2.3GHz, the four small cores are 2.1GHz Arm Cortex-A55 cores. The GPU is Arm Mali-G57MC4 running at 850MHz.

In terms of imaging, the T820 adopts a quad-core ISP architecture and is equipped with UNISOC’s sixth-generation Vivimagic image engine. It supports 1.6-megapixel shooting per second, a 108-megapixel camera, and full-channel 4K high-definition video recording. Combined with AI algorithms, it also supports AI noise reduction, AI-HDR, AI-Bokeh depth of field blur, FaceID unlocking, AI face detection, etc. Support 14bit RAW processing, combined with the new "heterogeneous image processing engine" architecture, XDR dedicated night scene multi-frame, multi-exposure shooting capabilities, it can restore more dark details when shooting night scenes.

In terms of security, the T820 adopts a trusted execution environment, an independent hardware encryption and decryption engine, and integrates a financial-grade iSE security unit. Compared with an external SE solution, it is more difficult to attack and locate, and has higher security. It can support hundreds of bank cards, digital currencies, etc. Applications in financial scenarios. Compared with the external SE solution, the computing power has also been increased by 2 times, and it supports high computing power security requirements such as video encrypted calls, international mainstream algorithms, and the storage capacity has also increased by 20 times, and can support hundreds of applications at the same time. In addition, high integration can also greatly reduce the difficulty of PCB design, overall machine design and manufacturing costs.

In terms of AI, the 820 adopts a new generation of high-efficiency NPU+VDSP architecture, and the computing power of the NPU reaches 8TOPS, which is 67% higher than the previous generation. A brand-new AI computing platform that can provide a full-stack AI solution integrating software, hardware, and algorithms, and supports low-bit quantization of model weight values and activation values, including 4bit, 8bit, 10bit, 12bit, 16bit, etc. Different precision ;Support access to mainstream training frameworks in the industry; support advanced weight compression technology, making it possible to deploy "big models" on the device side; support 200+ mainstream neural network operators, provide convenient development access and debugging environments, and use more Friendly and smarter.

In terms of 5G, it supports 5GNRTDD+FDD dual-carrier aggregation, 130MHz spectrum bandwidth aggregation, 4G/5G dynamic spectrum sharing, and full-scenario coverage enhancement technology. The uplink rate of near-point users in the cell is increased by 60%, and the network coverage is increased by more than 100%. It also supports 2G to 5G multi-mode full Netcom, Sub-6GHz full frequency band, SA/NSA, dual card dual 5G, EPSFallBack, VoNR HD voice and video calls and other communication features.