Study: Analog Computing Solves Complex Equations Using Lower Energy
Researchers, including University of Massachusetts Amherst engineers, demonstrated that their analog computing device, called a memristor, completes complex scientific computing tasks that bypass the limitations of digital computing. Yes, you read that right.
Important scientific questions can be explored using complex equations, but digital computing systems are approaching their limit in terms of speed, energy consumption, and infrastructure for performing computations. The work, published in Science, explains that current computing methods require moving data between memory and computing units. It becomes a traffic jam with complex tasks moving larger amounts of data.
To solve the challenge, traditional computing increases bandwidth; however, the team at UMass Amherst, the University of Southern California, and TetraMem Inc. implemented in-memory computing with analog memristor technology as an alternative.
The memristor is a combination of memory and resistor. The memristor controls the flow of electrical current in a circuit and “remembers” the prior state, even when the power is turned off. The device can be programmed into multiple resistance levels, increasing the information density in one cell.
A memristive circuit performs analog computing using physical laws in a massively parallel fashion, accelerating matrix operation, a frequently used but power-hungry computation in neural networks. The computing is performed at the device rather than moving the data between memory and processing.
The team’s memristor can complete low-precision computing tasks, like machine learning, and applications such as analog signal processing, radio frequency sensing, and hardware security. The memristor solved static and time-evolving partial differential equations, Navier-Stokes equations, and magnetohydrodynamics problems.
It took over a decade for the UMass Amherst team and collaborators to design the memristor device and build sizeable circuits and computer chips for analog in-memory computing.
