Memristors hold promise as a way to potentially cut energy consumption of smartphones and efficient supercomputers by up to a factor of 100.

Memristors — named as a portmanteau of memory and resistor — can be programmed to have different resistance states so they store information as resistance levels. These types of circuits enable memory and processing in the same device, cutting out the data transfer bottleneck typically found in conventional computers that separate the memory and the processor.

"Historically, the semiconductor industry has improved performance by making devices faster. But although the processors and memories are very fast, they can't be efficient because they have to wait for data to come in and out," said Wei Lu, a University of Michigan professor of electrical and computer engineering and co-founder of memristor startup Crossbar Inc.

Unlike ordinary bits, which are 1 or 0, memristors have resistances that are on a continuum. Some applications such as neuromorphic computing take advantage of the analog nature of memristors, but so far in ordinary computing trying to differentiate among small variations in the current passing through a memristor device is not precise enough for numerical calculations.

Researchers at the University of Michigan found a way around this problem by digitizing the current outputs, or defining current ranges as specific bit values. The team mapped large mathematical problems into smaller blocks within the array to improve efficiency and flexibility in the system.

These blocks, called memory-processing units, could be useful in implementing machine learning and artificially intelligent algorithms. These blocks are also useful in tasks that are based on matrix operations such as simulations used for weather prediction. Simple mathematical matrices, such as tables with rows and columns of numbers, can map directly onto the grid of memristors, researchers said.

Once the memristors are set to represent the numbers, operations that multiply and sum the rows and columns can be taken care of simultaneously, with a set of voltage pulses along the rows. The current measured at the end of each column contains the answers. This is in contrast to a processor that typically reads the value from each cell of the matrix, performs multiplication and then sums up each column in a series.

The full research can be found in the journal Nature Electronics.

To contact the author of this article, email