During HiPhi Wing Day in 2023, the latest achievements in intelligent cockpit technology developed by HiPhi Automotive drew great attention. The company unveiled a self-developed high-computing intelligent cockpit platform. This platform was the first in the automotive industry to feature the Qualcomm QCS8550 chip, supported by FPGA, MCU, and gateways – three kinds of infrastructures that comply with aviation and automotive grade standards. By developing in parallel with chip and automotive-grade large systems, we aim to find the optimum balance between high reliability and high computing power, hoping to provide users with a convenient, expandable, and continuously evolving intelligent cockpit experience.
This new cockpit platform is HiPhi Auto’s innovative solution to emerging industry issues and new customer demands, crafted based on a modular high-reliability safety architecture. By harnessing the power of chip parallelism and automotive-grade large system development, it ensures vehicle-grade reliability, safety, and stability, while fulfilling users’ demands for a high-performance, ecologically diverse, and iteratively developed intelligent cockpit.
The deep collaboration between HiPhi Auto and Qualcomm allows the new intelligent cockpit platform to be the first to feature the Qualcomm QCS8550 chip, further enhancing the computational power of the intelligent cockpit and making the car’s performance comparable to top-tier phones.
In terms of AI, the Qualcomm QCS8550 chip offers the highest AI computing power up to 96 TOPS. It is the first to support local operation of the bulky Transformer model on vehicle terminals, delivering a superior AI experience, faster response, and user privacy protection. The chip is also the first mobile chip to support hardware ray tracing, enabling advanced gaming graphics for the vehicle interface.
Based on this intelligent cockpit platform, HiPhi Auto in collaboration with Microsoft, has released a local voice large model based on GPT. Combined with the latest chip acceleration technology, it deploys cloud-based recognition synthesis capabilities to the edge, uses multi-language support and massive data to enable multi-language mixed recognition, breaking dialect barriers, and thereby making voice assistant conversations more natural and intelligent.
HiPhi Auto plans to conduct small-scale internal testing on the current HiPhi X model by the end of this year, with expectations being set to assemble it in bulk to cars by the first quarter of 2024.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.