100%主流車企的共同選擇：一個AI「通用底座」正在汽車行業成型

🔑 增強重點摘要

• •The 'universal base' refers to the industry-wide convergence on NVIDIA DRIVE Thor as the centralized compute architecture for next-generation software-defined vehicles.
• •This standardization is driven by the need to support massive Transformer-based models for end-to-end autonomous driving, which require unified hardware-software stacks to manage latency and power efficiency.
• •The shift to proactive services is enabled by the integration of Large Language Models (LLMs) directly into the vehicle's cockpit domain, allowing for context-aware, intent-based user interactions rather than command-based inputs.

🛠️ 技術深入

• •Architecture: NVIDIA DRIVE Thor utilizes the Blackwell GPU architecture, enabling high-performance inference for generative AI models within the vehicle.
• •Compute Density: The platform integrates 2000 TFLOPS of performance, allowing for the consolidation of cockpit, infotainment, and autonomous driving functions onto a single SoC.
• •Software Stack: Utilizes NVIDIA DRIVE OS and DRIVE IX, which provide the middleware for real-time sensor fusion and LLM-based voice/vision processing.
• •Interconnect: Employs high-speed NVLink-C2C for low-latency communication between the AI compute engine and the vehicle's sensor suite.

🔮 前景展望基於引用來源的 AI 分析

⏳ 時間線