Key Findings
XHEART has launched the X7 Ultra, a mass-production autonomous-driving chip providing 700 TOPS of compute. According to CEO Li Zonglin, this performance is measured against real mass-produced large models, ensuring the quantized deployed model maintains over 99.99% similarity with the trained FP32 floating-point model. The X7 series has been in mass production for eight months, with cumulative shipments exceeding 200,000 units across 10 equipped vehicle models and orders for more than 40 models.
Technical Details
The X7 Ultra utilizes a “large models define chips” approach, reorganizing compute, storage, and data flow based on real-world model requirements. Key technical specifications include:
- Super-core NPU: Aggregates data exchanges into TB/s-level high-speed interconnects with dedicated acceleration engines for Transformer operators (ReLU, SOFTMAX, and Transpose).
- Efficiency: NPU compute utilization reaches 98.7%, with energy efficiency 100% higher and memory bandwidth consumption 80% lower than industry flagship products.
- Hardware: Features the first mass-produced Armv9 automotive-grade CPU in China and an ASIL-D certified built-in MCU, meeting ISO 26262 ASIL-B functional safety requirements.
- Sensor Processing: Lidar processing bandwidth reaches 20 million points per second.
Background and Context
XHEART was spun out of the chip design department of Momenta. The company abandoned general-purpose chip routes in 2023 to focus on hardware that supports the shift toward end-to-end large models. This collaboration between Momenta and XHEART aims to support models with 10 billion to 100 billion parameters. Manufacturing is supported by TSMC, which provided process support for the X7 Ultra.
Significance and Outlook
The collaboration has resulted in the R7 Ultra, which expands model parameter scale by five to 10 times and improves safety performance by more than five to 10 times. Momenta CEO Cao Xudong expects L3 and L4 autonomous driving to reach an inflection point in 2028, moving toward scaled deployment by 2030.
Source: https://finance.biggo.com/news/29fd4583-d6e0-42d9-b711-717422408580
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
