Researchers in China have developed a memristor-based chip capable of modeling the brain’s complex, folded cortical structure in under 10 milliseconds—roughly matching the speed of neural processing in the human brain. Led by Yuchao Yang of Peking University in collaboration with the Shanghai Institute of Microsystem and Information Technology, the team demonstrated a phase-change memristor system that delivers real-time brain-like computation. Their findings were published in Science.

Neural dynamical systems combine neural networks with mathematical models to simulate how complex systems change over time, enabling applications such as physical simulations, medical imaging, and 3D brain reconstruction. However, these systems require continuous calculations and frequent adjustments, making them computationally demanding. Traditional computers further limit performance by constantly transferring data between memory and processors, increasing both processing time and energy consumption—an obstacle for real-time applications.

Figure 1. Near-Biological Brain Modeling

Memory-Based Computing Speeds Brain Simulations

Real-time brain modeling has the potential to advance technologies such as brain–computer interfaces, surgical guidance, and medical imaging, but conventional hardware is often too slow and energy-intensive for these demanding tasks. The new chip overcomes this limitation by performing key computations directly within memory, minimizing data transfers between memory and the processor. This approach enables faster, more energy-efficient, and high-precision brain modeling. Figure 1 shows Near-Biological Brain Modeling.

New Chip Surpasses Specialized Hardware

Fabricated using a 40-nanometer process, the memristor chip integrates in-memory computing within a compact 0.28 mm² design operating at 50 MHz. In neural dynamics tasks, it delivered speeds 3.82 to 36.27 times faster than leading application-specific integrated circuits (ASICs) while using 11.75 to 24.73 times less power. For cortical surface reconstruction, it achieved up to a 478-fold speed increase over an NVIDIA A100 GPU. The researchers also demonstrated real-time reconstruction of the brain’s white and gray matter boundaries, producing accurate 3D cortical surface models that preserved the brain’s complex folded structure.

The chip achieved high accuracy on standard surface-quality metrics, including average symmetric surface distance and Hausdorff distance, demonstrating that it can deliver both rapid processing and precise brain modelling [1]. By reducing complex neural simulations from slow offline computations to millisecond-scale performance, the technology could enable future applications such as brain–computer interfaces, digital brain twins, real-time surgical navigation, brain-surface reconstruction, and advanced research into neurodegenerative diseases like Alzheimer's and Parkinson's.

Reference:

1. https://scitechdaily.com/scientists-develop-a-groundbreaking-chip-that-operates-at-brain-like-speed/

Cite this article:

Janani R (2026), Scientists Create a Chip That Matches the Brain’s Speed, AnaTechMaz, pp. 1026