Skip to content
AI NEWS JOURNAL
AIBase AI NewsAI NEWS / JOURNAL

Specifically Designed for End-to-End Large Models: XHEART Unveils the X7 Ultra Intelligent Driving Chip, Featuring China’s First Mass-Produced Armv9 Automotive‑Grade CPU

On October 29, XHEART unveiled its next-generation intelligent driving large‑model chip, the X7 Ultra. In response to the industry trend of shifting from multiple small models to “end-to-end large models,” the company adheres to the philosophy of “large models define the chip.” Starting from the fundamental requirements of real‑world mass‑production models, it reverse‑engineered and redesigned the computing architecture, storage solutions, and data flow mechanisms, while also leaving ample room for future model evolution. The article further discusses core computing power and architectural design.

On October 29, XHEART officially launched its new generation intelligent driving large‑model chip—the X7 Ultra. Facing the industry trend of smart driving evolving from traditional multiple small models toward “end-to-end large models,” XHEART has embraced the core principle of “large models define the chip.” From the very beginning, the company decided to start with the underlying needs of actual mass‑production models, working backward to redesign the chip’s computational architecture, storage scheme, and data‑flow methods, while simultaneously reserving sufficient capacity for continuous future model upgrades.

In terms of core computing power and architectural design, the X7 Ultra delivers a large‑model compute capability of 700 TOPS, with measured alignment to production‑grade models reaching an impressive 99.99%. Specifically engineered for large models, this chip adopts an ultra‑large‑core NPU architecture, achieving TB/s‑level improvements in intra‑core data exchange compared to conventional inter‑core communication. Additionally, its multi‑tiered cache architecture combines high capacity with strong concurrency, effectively reducing latency and power consumption during inference.

To ensure both model accuracy and operational efficiency, the X7 Ultra integrates a high‑precision operator acceleration engine tailored for large models. It features extreme hardware‑level optimizations for key Transformer operations such as Layernorm, Softmax, and Transpose, and natively supports FP16/FP32 high‑precision vector computations. With these innovative designs, the NPU’s hardware parallelism is fully unlocked, achieving a computation utilization rate of up to 98.7%; meanwhile, memory bandwidth consumption drops by 80% compared to other industry flagship products, resulting in a 100% improvement in overall energy efficiency.

Beyond its robust NPU performance, the X7 Ultra also marks China’s first mass‑produced Armv9 automotive‑grade CPU, delivering a 20% performance boost and a 10% increase in energy efficiency. Furthermore, through an upgraded large‑cluster architecture, the chip significantly enhances multi‑core collaboration and L3 memory access performance, while integrating a brand‑new SVE2 vector processing engine that offers stronger DSP (digital signal processing) capabilities than the previous NEON implementation.

Sources

Read original ↗

Related topics

← Continue to all updates