Summary: Researchers introduced the world’s first chip capable of matching the operational speed of the human brain.
Fabricated using a standard 40-nanometer process, the sub-10-millisecond neural dynamical system utilizes phase-change memristors to execute key mathematical calculations directly within memory. Occupying just 0.28 square millimeters, the chip achieves up to a 478-fold speedup over enterprise GPUs in real-time cortical surface reconstruction while drastically lowering energy demands.
This hardware milestone paves the way for real-time brain-computer interfaces, intraoperative surgical navigation, and full-scale digital brain twins.
Key Facts
- Sub-10-Millisecond Brain-Speed Simulation: The chip processes continuous neural dynamics at operational speeds matching the native millisecond temporal scale of the human brain.
- Massive GPU Acceleration: In 3D cortical surface reconstruction tasks (mapping brain white and gray matter folds), the phase-change memristor chip achieved up to a 478.18× speedup compared to an enterprise-grade NVIDIA A100 GPU.
- Superior Energy and Latency Metrics: Compared to state-of-the-art Application-Specific Integrated Circuits (ASICs), the neuromorphic architecture operates 3.82× to 36.27× faster while consuming 11.75× to 24.73× less energy.
- Overcoming the Memory Wall: By utilizing in-memory computing with 9 pipeline stages running at 50 MHz, the system eliminates traditional data-shuttling overheads between memory and CPU/GPU processors.
- High-Fidelity Anatomical Reconstruction: The system generated smooth, closed, and topologically accurate 3D manifold cortical meshes, scoring exceptionally high on Average Symmetric Surface Distance (ASSD) and Hausdorff Distance metrics for neuroimaging accuracy.
Source: Peking University
A research team led by Professor Yang Yuchao of Peking University, together with researchers from the Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, has developed the world’s first chip that can match the functioning speed of the human brain.
The study, titled “A sub–10-millisecond neural dynamical system based on phase-change memristors,” was published in Science.
Background
Neural dynamical systems combine neural networks with mathematical equations that describe how complex systems change over time. They are useful for physical modeling, medical imaging, and three-dimensional brain reconstruction. However, these systems require repeated calculations, error checks, and adjustments to the size of each calculation step. In conventional computers, data must also move frequently between the memory and processor, increasing processing time and energy use.
Why it matters
Fast and accurate brain modeling is important for technologies that must respond in real time, including brain–computer interfaces, surgical navigation, and medical imaging. Existing hardware often requires too much time and power for these demanding calculations. By performing key operations directly in memory, the new chip reduces data movement and brings high-quality brain modeling closer to real-time use.
Key findings
Fabricated using a 40-nm process, the chip’s in-memory computing and conductance-drift arrays occupy only 0.28 square millimeters. It operates at 50 MHz and uses nine pipeline stages for each integration step. In neural dynamics calculations, the system is 3.82× to 36.27× faster and consumes 11.75× to 24.73× less power than state-of-the-art ASICs, or application-specific integrated circuits. In cortical surface reconstruction tasks, it achieves up to a 478.18× speedup compared with an NVIDIA A100 GPU.
The researchers used the chip to reconstruct the surfaces of the brain’s white and gray matter and generate 3D manifold-based surface meshes in real time. The system produced smooth, closed, and topologically consistent cortical surfaces while accurately capturing complex brain folds. It also performed well in average symmetric surface distance and Hausdorff distance measurements, demonstrating its ability to support high-fidelity brain modeling.
Future Implications
The chip could help move complex neural modeling from slow, offline processing toward millisecond-scale operation. In the future, the technology may support brain–computer interfaces, digital brain twins, real-time surgical navigation, brain-surface reconstruction, and tools for studying neurodegenerative diseases such as Alzheimer’s and Parkinson’s.
Key Questions Answered:
A: A phase-change memristor is a memory element that changes its electrical resistance when exposed to physical heat states, allowing it to store continuous analog values rather than just binary 0s and 1s. This allows the chip to calculate complex mathematical equations directly inside the memory array, skipping time-consuming data transfers.
A: Conventional GPUs process heavy physical and mathematical models by constantly sending data back and forth between computational units and high-bandwidth memory. The phase-change chip performs these calculations in-place, achieving up to a 478-fold speedup in 3D cortical reconstruction over an A100 GPU while drawing a fraction of the power.
A: Millisecond-scale processing enables real-time brain-computer interfaces (BCIs) that interpret neural signals without lag, intraoperative surgical navigation systems that update 3D brain scans during procedures, and real-time “digital brain twins” to model neurodegenerative disorders like Parkinson’s and Alzheimer’s.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this neurotech research news
Author: Jiang Zhang
Source: Peking University
Contact: Jiang Zhang – Peking University
Image: The image is credited to Neuroscience News
Original Research: Open access.
“A sub–10-millisecond neural dynamical system based on phase-change memristors” by Lei Cai, Yaoyu Tao, Chenchen Xie, Longhao Yan, Shiqian Li, Ruihong Shen, Zelun Pan, Xile Wang, Bowen Wang, Daijing Shi, Yihang Zhu, Teng Zhang, Yixin Zhu, Xi Li, Zhitang Song, Ru Huang, Yuchao Yang. Science
DOI:10.1126/science.aee6277
Abstract
A sub–10-millisecond neural dynamical system based on phase-change memristors
High-fidelity geometry for physical-world modeling demands real-time, dense, and differentiable deformation fields on manifolds.
Neural dynamical systems (NDSs) using adaptive stepsize integration with embedded neural networks excel at these tasks but still suffer latency on the order of hundreds of milliseconds. In this work, we report a sub–10-millisecond NDS hardware leveraging the precisely controlled conductance drift of phase-change memristors and their multilevel compute-in-memory capabilities.
We fabricated a 40-nanometer NDS chip for the challenging surface reconstruction tasks. Compared with state-of-the-art NDS hardware, our NDS design achieves a latency of 2.12 milliseconds (below 10 milliseconds) for single-iteration NDS computations with an error tolerance of 10−7 and delivers 3.82× to 36.27× faster speed while consuming 11.75× to 24.73× less power.
The end-to-end NDS latency through hardware measurements and simulations outperformed graphics processing unit A100 by 50.38× to 478.18×.
