First tensor processor chip based on carbon nanotubes could lead to energy-efficient AI processing

https://techxplore.com/news/2024-08-tensor-processor-chip-based-carbon.html

"systolic array processor network rhythmically performing 2-bit integer convolution/ matrix multiplication in parallel... allowing free passage using 3K field effect transistors with carbon nanotube (instead of conventional semiconductor) channels, organized as 3*3, 9 total processing units, forming systolic array ... reduces read/ write of static random-access memory, saving energy... 5-layer convolutional neural network, image accuracy up to 88%, power consumption only 295μW... potential reduced latency. more bandwidth"