Memristor Chip Achieves Millisecond Neural Computing
新智元 · wechat · 2026-07-13
A joint team from Peking University and the Chinese Academy of Sciences published a new study in Science, proposing a "controllable in-memory computing" paradigm based on phase-change memristors for the real-time computation of neural dynamical systems.
Key Breakthroughs
- Developed the world's first millisecond-level neural dynamics system chip based on phase-change memristors.
- Reduced single-step computation latency to 2.12 milliseconds.
- Achieved a 3.8236.27x speedup and an 11.7524.73x reduction in power consumption compared to advanced ASICs.
- Delivered a 50.38478.18x speedup over NVIDIA A100 GPUs on tasks like brain cortex surface reconstruction.
Significance
- This chip pushes complex neural dynamics modeling into millisecond-level online real-time processing.
- The paper demonstrates its effectiveness in brain white/gray matter cortex reconstruction and 3D manifold grid generation.
- The author believes this will provide a new hardware foundation for brain-computer interfaces, brain digital twins, neural navigation, and neurodegenerative disease treatments.
More from Infra
- Nebius says SlimSpec speeds speculative decoding 8–9% without shrinking the vocabulary — Arindam_1729 · 2026-07-21
- NVIDIA brings its Cosmos 3 Edge world model to Jetson for on-device robot control — liu_mingyu · 2026-07-21
- A silicon photonic reservoir chip compensates fiber distortion in real time at 28 Gbps — bravo_abad · 2026-07-21
- Chamath says open-sourcing Grok would push AI margins from models to infra and apps — Dan_Jeffries1 · 2026-07-21
- EU AI competitiveness is under pressure as firms double down on chips, ethics, and talent — nordicinst · 2026-07-21
- AI bottlenecks are shifting to memory, optics, yield control and power — thedealdirector · 2026-07-21