STEM与日常科技·英语30篇（5）

8 / 30

忆阻器：让计算发生在存储地

Traditional computers shuttle data constantly between separate memory chips and processors—a bottleneck called the von Neumann wall.
Memristors are tiny electronic components that both store information and perform logic operations in the same physical spot.
They change resistance based on how much electric charge has passed through them—like biological synapses.
Stacking thousands of memristors in 3D grids creates ‘compute-in-memory’ hardware that skips data transfer delays.
Early AI chips using memristors run neural network tasks up to seven times faster using half the energy.
They also mimic how the brain learns—adjusting connection strength gradually instead of flipping digital bits.
Manufacturers now integrate them into edge devices like security cameras and hearing aids.
Because they retain state without power, they cut boot-up time and enable instant-on functionality.
This architecture doesn’t replace CPUs—it reshapes where and how computation happens most efficiently.
It’s not magic; it’s physics, nanofabrication, and a fresh idea about what memory can do.