Chinese scientists’ brain-mimicking chip ‘up to 478 times faster than Nvidia A100 GPU’
New chip performs data storage and computation in a single memory array, enabling real-time modelling of complex brain structures, team says
According to its developers, this chip could not only transform diagnostics and treatment for conditions such as Alzheimer’s disease, but also boost the performance of brain-machine interfaces and assist surgeons.
Researchers from Peking University and the Chinese Academy of Sciences reported the breakthrough in a peer-reviewed study published in Science on Thursday, detailing a 40-nanometre memory chip with an integrated artificial neural network.
The device overcomes long-standing computational limits, enabling it to reconstruct complex brain surfaces in less than half a second – making it 50 to 478 times faster than state-of-the-art Nvidia A100 graphics processing unit (GPU) systems, according to the team.
Lead author Yang Yuchao, a professor at Peking University’s school of integrated circuits and deputy dean for its school of electronic and computer engineering, told state-run Guangming Daily that the chip could accurately render the brain’s folds for medical applications.
“This breakthrough opens up new possibilities for brain-computer interfaces and the diagnosis and treatment of brain diseases,” he said. “In the future, personalised and dynamic digital brain twins will become possible.”
“It also provides a hardware foundation that can operate in real time for intraoperative neuronavigation [a navigation system for surgery], early screening for Alzheimer’s disease and personalised interventions,” Yang added.