Unigroup Zhanrui releases end-side AI platform solution that can be applied to car cockpits – Passionategeekz –

[CNMO Technology News]At the Shanghai World Mobile Communications Conference which opened on June 18, Unigroup Zoning officially launched UNISOC end-side AI platform solution, providing full-link technical support for the intelligent transformation of terminal devices. Faced with the industry trend of artificial intelligence technology accelerating penetration into various terminals, this solution breaks through the computing power bottleneck through chip-level innovation and is committed to building an AI infrastructure that adapts to diverse scenarios.

As a world-leading platform chip design company, Unigroup Zhanrui has launched two major solutions, single chip and heterogeneous distributed, based on the accumulation of mature 5G SoC technology. Among them, the single-chip AI solution achieves a breakthrough in increasing the speed of large-model generation by 20% and reducing the power consumption of inference by 60% by deeply integrating AI capabilities into the T9100 chip. This solution adopts a multi-core heterogeneous architecture and dynamic voltage frequency adjustment technology, and has been mass-produced in smartphones, learning tablets and other devices. For example, the BBK AI learning machine has been launched on the market.

For complex scenarios such as industrial control and smart cars, heterogeneous distributed solutions support flexible computing power configurations from 1T to 100T, and can run more than 30 AI models of different parameter scales at the same time. Its distributed architecture design allows multi-chips to work together and memory bandwidth is expanded on demand, especially adapted to robots, edge computing terminals and other devices that require multimodal interaction.

This platform solution has three major technical advantages: first, the full-scene coverage capability, and by supporting cutting-edge technologies such as big models and AI Agents, it can realize more than 20 functions such as voice interaction, intelligent image search, and real-time translation; second, the development tool chain innovation, and the self-developed AISDK provides a visual development environment and plug-in deployment tools, which improves the efficiency of AI application development by 40%; third, the ecological openness, the original MCP interface standard supports end-cloud collaboration, and can quickly access third-party algorithms and services.

In Zhanrui’s technical demonstration, smart terminals equipped with UNISOC platform demonstrated multi-modal processing capabilities: the car cockpit can run voice assistants, environmental perception and decision-making systems at the same time, and industrial equipment can realize real-time collaboration of visual detection and predictive maintenance. It is worth noting that through the coordinated optimization of software and hardware, this solution reduces the energy consumption of AI computing by 55% compared with traditional solutions and reduces the storage usage by 30%.

All rights reserved, no reproduction without permission