Researchers from Peking University, in collaboration with the Shanghai Institute of Microsystem and Information Technology under the Chinese Academy of Sciences, have developed a groundbreaking memristor chip capable of modeling the complex, folded surface of the human brain with high precision in under 10 milliseconds.
According to the Operative Information Center-OMM, citing international media reports, the findings of this research have been published in the prestigious journal Science.
Experts note that artificial intelligence algorithms used to model complex, time-varying systems are widely applied in fields such as physical simulation, medical imaging, and three-dimensional brain reconstruction. However, traditional computing architectures often suffer from latency and high energy consumption due to the constant data transfer between memory and processors.
The new memristor chip addresses this bottleneck by performing computations directly within the memory where data is stored. This architecture significantly enhances processing speed while drastically reducing energy consumption. Manufactured using 40-nanometer technology and occupying just 0.28 square millimeters, the chip operates at a frequency of 50 MHz.
During performance testing, the chip demonstrated remarkable efficiency. In calculating neuronal dynamics, it proved to be 3.82 to 36.27 times faster than modern ASIC processors while consuming 11.75 to 24.73 times less energy. Furthermore, in the three-dimensional reconstruction of the cerebral cortex, the chip exhibited a speed advantage of up to 478 times compared to the NVIDIA A100 graphics processor.
The tests confirmed that the chip could model the boundaries between white and gray matter with high precision, preserving complex brain folds and creating topographically accurate three-dimensional surfaces. Researchers believe this technology could play a pivotal role in the future development of brain-computer interfaces, digital brain twins, real-time surgical navigation systems, and the study and treatment of neurodegenerative diseases such as Alzheimer's and Parkinson's.
Memristor technology represents a significant shift in hardware design, moving away from the traditional Von Neumann architecture toward neuromorphic computing, which mimics the neural structure of the human brain to achieve greater efficiency in AI tasks.
