New pc storage expertise might energy the AI ​​of the longer term

A research team led by the University of Cambridge has developed a novel computer memory design that promises a significant increase in performance while reducing the energy requirements of internet and communications technologies.

According to the university, AI, algorithms, internet use and other data-driven technologies are estimated to account for over 30% of our global electricity use within the next decade.

“This explosion in energy demand is due in large part to the shortcomings of current computer storage technologies,” said first author Dr. Markus Hellenbrand from the Department of Materials Science and Metallurgy in Cambridge. “With traditional computing, there’s storage on one side and processing on the other, and data is moved back and forth between the two, which takes both energy and time.”

Researchers experimented with a new type of technology called resistive switching memory. Unlike traditional memory devices that can encode data in two states (one or zero), this new type of memory can allow for a continuous range of states.

It does this by applying an electrical current to certain materials, increasing or decreasing their electrical resistance. The different changes in electrical resistance create different possible states for storing data.

“A typical USB stick with a continuous range, for example, could store ten to a hundred times more information,” explains Hellenbrand.

The team prototyped a hafnium oxide-based device that had previously proved challenging for resistive switching memory applications. This is because the material has no structure at the atomic level. However, Hellenbrand and his co-scientists found a solution: They mixed barium into the mixture.

“These materials can function like a synapse in the brain.

When barium was added, it formed highly structured barium “bridges” between thick hafnium oxide films. An energy barrier was created where these bridges meet the device contacts, allowing the electrons to pass. The energy barrier can be raised or lowered, changing the resistivity of the hafnia composite and in turn allowing multiple states to exist in the material.

“The really exciting thing about these materials is that they can function like a synapse in the brain: they can store and process information in the same place as our brain can,” Hellebrand said.

Researchers believe this could lead to the development of computing storage devices with far greater density and performance but lower power consumption, making the technology particularly promising in the field of AI and machine learning.

A patent for the technology has been filed by Cambridge Enterprise, the university’s commercialization arm, and the scientists are now working with industry to conduct larger feasibility studies. They claim that integrating hafnium oxide into existing manufacturing processes will not pose a challenge as the material is already used in semiconductor production.

Comments are closed.