Computers are tasked with processing ever-increasing amounts of data to perform vital functions such as AI training, climate predictions, drug discovery, etc. We need faster, longer-lasting, and more stable computer memory to meet this demand.

(Photo: Unsplash/ Florian Krumm)


Need for Faster Nonvolatile Memory

Conventional computers store and process data in separate locations. The processing is handled by volatile memory, which quickly disappears when the computer is turned off. Meanwhile, long-term data storage is dealt with by nonvolatile memory, which is not fast but can hold information without constant power input. As information shifts between these two locations, congestion may occur as the processor waits for large amounts of data to be retrieved.

Sending data back and forth takes a lot of energy, especially with the current computing workloads. With this type of memory, experts hope to bring the memory and processing closer into one device so less energy and time will be used. However, many technical challenges exist in achieving an effective, commercially viable universal memory that demonstrates long-term storage and fast, low-power processing without compromising other metrics.

READ ALSO: Revolutionizing Computing: Nano Engineers Created Thinnest Ferroelectric Semiconductor to Advance AI and Break Moore's Law



New Candidate for Universal Memory

At Stanford University, experts have demonstrated the potential of a new material in making phase-change memory. In the paper "Novel nanocomposite-superlattices for low energy and high stability nanoscale phase-change memory," the research team detailed their scalable technology,y which can be fabricated at temperatures compatible with commercial manufacturing. The researchers hope their innovation will inspire further development and adoption as a universal memory.

The proposed computing memory relies on GST467, an alloy made of seven parts tellurium, six parts antimony, and four parts germanium. The researchers, led by Stanford materials science and engineering professor Eric Pop, found ways to compress the alloy between several other nanomaterials in a superlattice previously used to achieve good nonvolatile memory.

The unique composition of GST467 makes it particularly fast at switching speed. When integrated within the superlattice structure in nanoscale devices, the alloy enables low switching energy, which provides excellent stability and good endurance, making it nonvolatile. It can even retain its state for ten years or even longer. The GST467 superlattice clears several important benchmarks. Although phase change memory can sometimes drift over time, testing the material reveals that the designed memory is extremely stable. Since it operates below one volt, it is significantly faster than a typical solid-state drive.

A few other types of nonvolatile memory can be a bit faster, but they operate either at a higher voltage or higher power. This results in tradeoffs between speed and energy. Allowing the experts to switch at a few tens of nanoseconds while operating below one volt is a big deal in computing technology.

Additionally, the superlattice packs many memory cells into a small space. The research team shrunk the memory cells to 40 nanometers in diameter, almost less than half the size of a coronavirus. They are still exploring ways to compensate by stacking memory in vertical layers, which can be made possible by the low fabrication temperature and the techniques used to create the superlattice.

RELATED ARTICLE: Towards Quantum Computing: Physicists Surpass Current Supercomputers With New Programmable Simulator

Check out more news and information on Computing in Science Times.