Researchers at the University of Cambridge have developed a high-performance memristor using a specialized hafnium oxide.
Researchers at the University of Cambridge have developed a high-performance memristor using a specialized hafnium oxide . A memristor is a low-energy component that mimics how our brain cells connect, helping conserve energy.
This nanoelectronic device could potentially reduce AI energy consumption by up to 70%. “Energy consumption is one of the key challenges in current AI hardware,” said Dr Babak Bakhit, lead author from Cambridge’s Department of Materials Science and Metallurgy. “To address that, you need devices with extremely low currents, excellent stability, outstanding uniformity across switching cycles and devices, and the ability to switch between many distinct states,” Bakhit added. Energy-efficient chipsStandard computer chips are inefficient. These chips spend most of their energy shuffling data between a memory unit and a processor. This digital commute creates heat and wastes power.Our brains don’t work that way. We process and store information in the same place: the synapse.The Cambridge team’s technique uses a specialized form of hafnium oxide to do exactly that.Researchers have developed a neuromorphic chip using a stable, low-energy memristor that handles both tasks in a single chip. Most memristors rely on “conductive filaments” that are unpredictable and unstable. The Cambridge team developed a more stable alternative using a hafnium-based thin film. It swaps out erratic, high-voltage filaments — which physically grow and snap inside older devices — for a smooth, reliable switching interface.The new device incorporates strontium and titanium to create internal p-n junctions that act as smooth electronic gates. Instead of relying on chaotic structural changes, this new device regulates electricity by simply adjusting an energy barrier at the material’s interface. It offers the precision and reliability needed for large-scale AI systems.“Filamentary devices suffer from random behaviour,” said Bakhit. “But because our devices switch at the interface, they show outstanding uniformity from cycle to cycle and from device to device.”700°C hurdleThis hafnium-based breakthrough reduced power consumption by using switching currents a million times smaller than those of older technology.Furthermore, this development also supports hundreds of stable, distinct levels of electrical flow — the precise “multi-tasking” ability required for advanced analogue in-memory computing.Laboratory tests confirmed these devices are both durable and brain-like, reliably enduring tens of thousands of cycles while holding their data for about a day. Most importantly, they mimic biological learning by recreating “spike-timing dependent plasticity.” It is the same process our own neurons use to strengthen or weaken connections based on the timing of incoming signals.“These are the properties you need if you want hardware that can learn and adapt, rather than just store bits,” said Bakhit.Despite the excitement, the finish line is still a few miles off. Currently, the fabrication process requires temperatures of 700°C. That is far too hot for standard semiconductor manufacturing, which prefers much cooler conditions to avoid melting delicate components.Dr. Bakhit, who spent three years on “a huge number of failures” before the breakthrough last November, is now focused on bringing that temperature down to make it compatible with modern factory lines. If they can successfully lower this temperature, the technology is poised to become a game-changing solution for ultra-low-energy AI hardware.This device could slash energy consumption by 70% while providing the stability and adaptability needed for large-scale, brain-like computing.The findings were reported in the journal Science Advances.
Chip Computer Chip Energy &Amp Environment Energy Consumption Hafnium Oxide Inventions And Machines
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Nearly 70% of US Voters Oppose Iran Ground Invasion as Trump Weighs Troop SurgeBrett Wilkins is a staff writer for Common Dreams.
Read more »
70 House GOP members urge Trump admin on Chinese-vape crackdownFox News Channel offers its audiences in-depth news reporting, along with opinion and analysis encompassing the principles of free people, free markets and diversity of thought, as an alternative to the left-of-center offerings of the news marketplace.
Read more »
Bitcoin prediction markets see 70% chance BTC price crashes to $55K in 2026The most recent news about crypto industry at Cointelegraph. Latest news about bitcoin, ethereum, blockchain, mining, cryptocurrency prices and more
Read more »
70 Years Ago, DC Revolutionized Comics and Set Up Marvel’s Success70 years ago, DC Comics kicked off the Silver Age of Comics with Showcase 4 and laid the groundwork for Marvel's success in the years to come.
Read more »
Fallbrook man, 70, charged in deadly shooting pleads not guiltyMichael Anthony Burke is accused of fatally shooting Martin Lucas and seriously injuring Julio Leon on Monday.
Read more »
3 Observations From Illinois' 105-70 Blowout of Penn in the NCAA TournamentAnd a freshman – two, in fact – would lead the Illini in an offensive avalanche that overwhelmed the undermanned Quakers on Thursday
Read more »