New hardware device to make artificial intelligence applications more energy efficient
ECE researchers have demonstrated a new device where data never leaves memory, called computational random-access memory (CRAM). This research comes from Wang and his collaborators’ patented work on Magnetic Tunnel Junctions, key to spintronic devices and used to improve hard drives, sensors, and other systems including Magnetic Random Access Memory (MRAM). The CRAM architecture enables true computation in and by memory, and it breaks down the wall between computation and memory which has been a bottleneck in traditional von Neumann architecture. CRAM performs computations directly within memory cells, utilizing the array structure efficiently, which eliminates the need for slow and energy-intensive data transfers. The state-of-the-art hardware device could reduce energy consumption for artificial intelligence (AI) computing applications by a factor of at least 1,000. The team plans to work with semiconductor industry leaders for large scale demonstrations and production of hardware to advance AI functionality. The team were supported by researchers from the University of Arizona.
The University of Minnesota Department of Electrical and Computer Engineering team included Yang Lv, Distinguished McKnight Professor and Robert F. Hartmann Chair Jian-Ping Wang, Professor Ulya Karpuzcu, researchers Robert Bloom and Hüsrev Cılasun, Distinguished McKnight Professor and Robert and Marjorie Henle Chair Sachin Sapatnekar, and former postdoctoral researchers Brandon Zink, Zamshed Chowdhury, and Salonik Resch. The researchers from University of Arizona included Pravin Khanal, Ali Habiboglu, and Professor Weigang Wang.
The research is published in npj Unconventional Computing, a Nature journal under the title, "Experimental demonstration of magnetic tunnel junction-based computational random-access memory." To read the details of the research, visit the npj Unconventional Computing website.
The research was supported by grants from the U.S. Defense Advanced Research Projects Agency (DARPA), National Institute of Standards and Technology (NIST), the National Science Foundation (NSF), and Cisco Inc. Research including nanodevice patterning was conducted in collaboration with the Minnesota Nano Center and simulation/calculation work was done with the Minnesota Supercomputing Institute at the University of Minnesota.
The College of Science and Engineering first posted this news.