Chinese researchers from Peking University and the Chinese Academy of Sciences have developed a brain-modelling chip that reconstructs complex brain structures in about 0.5 seconds.
The chip delivers 50x to 478x the performance of Nvidia’s A100 GPU for this specialised task, but it is designed specifically for brain modelling rather than general AI computing.
Using in-memory computing, the chip performs storage and computation in the same memory array, reducing data movement, lowering power consumption and boosting speed.
Researchers also repurposed phase-change memristors to speed up brain simulations, turning a known hardware limitation into a computing advantage.
The breakthrough could accelerate brain-computer interfaces, digital brain twins, neurological disease research and surgical planning.
Author: Saikat Bhattacharya
