An advance in a speedy type of microchip could help engineers integrate computers’ short-term and long-term memory.
For all the recent advances in the speed of computers, their command centers remain relatively inefficient. A central processor does all the thinking and quickly stores a bunch of 1s and 0s on a chip called dynamic random access memory, or DRAM. But DRAM only works when the computer is on, so it can serve only as short-term memory. Data needed for the long haul has to be stored on separate magnetic disk drives or on flash drives such as a camera’s memory card.
For decades, researchers have vied to create universal memory: a chip that combines the speed and reliability of DRAM with the archival abilities of flash. The advance, published June 11 in Nature Communications, fixes a weakness of a leading universal memory contender called ferroelectric RAM.
Although it’s fast and energy efficient, FRAM has had problems with long-term reliability. To determine whether a bit is a 1 or a 0, the chip has to apply a voltage that compromises the data. Then it must rewrite the data to preserve it. Those steps gradually degrade the storage capacity.
Ramamoorthy Ramesh, a materials scientist at the University of California, Berkeley, worked with a team of engineers at Nanyang Technological University in Singapore to develop a method for reading data without having to destroy it and then rewrite it. Their solution was to shine a very dim light at each bit-containing cell and measure the current that came out of it. The amount of current indicated whether the bit was a 1 or a 0. Most importantly, the light-shining process preserved the data, with no rewrite step necessary.
The researchers read and wrote data hundreds of millions of times with their prototype FRAM chip with no signs of degradation. In contrast, flash memory has a limit of several hundred thousand read/write cycles. “The innovative idea is the readout,” says Kang Wang, an electric engineer at UCLA. “This innovation may improve the chance of FRAM to be implemented in industry.”
Ramesh acknowledges that engineering issues and economic ones remain before FRAM transforms computing. Several other RAM technologies could serve as universal memory, including some backed by tech giants like Intel and Samsung.