Lingxi Launches Neuromorphic Super-Node Cluster
智东西 · wechat · 2026-07-19
During WAIC, Lingxi Technology unveiled its neuromorphic super-node cluster product, **LynAInfra**, designed for high-speed inference scenarios. The article notes that current inference workloads suffer from high energy consumption and latency. Neuromorphic chips, featuring compute-in-memory, event-driven processing, and many-core parallelism, are ideal for low-latency inference and can serve as an alternative to high-speed architectures like Groq. Hardware-wise, the LynAInfra128 consists of 128 HP640 boards, cascading thousands of neuromorphic chips in a single rack. It delivers nearly 100P@FP16 computing power while keeping total power consumption under 30kW. It supports both Scale-up and Scale-out, utilizing the LynxLink high-speed interconnect protocol and unified memory address compilation to scale single- and multi-rack deployments. The article also highlights its air-cooled, modular design, which is compatible with existing data center racks and power grids, thereby reducing deployment and retrofit costs. The core value proposition is summarized as "trading watts for bits": compared to traditional GPUs, inference energy efficiency is improved 3-10x, and time-to-first-token (TTFT) latency is compressed to the tens of milliseconds range. Beyond commercial applications, the article also touches on its significance for green data centers, domestic compute supply chains, and dual-carbon goals.
More from Embodied
- Generative Bionics turns its Gene.01 humanoid from CES concept into a walking platform — CyberRobooo · 2026-07-21
- Auki pitches physical AI infrastructure for stores, warehouses and supply chains — broodsugar · 2026-07-21
- Safe RL paper on directional constraints accepted to IROS 2026 — Jan_R_Peters · 2026-07-21
- Sunday Robotics says one example can teach its home robot a new behavior — saranormous · 2026-07-21
- San Francisco event seeks sponsors for the first 6-foot humanoid robot fight — cixliv · 2026-07-21
- Tiny memristor chip brings brain-model inference under 10 ms — striketheviol · 2026-07-21