A team of South Korean researchers has developed a new type of storage device that can be used to replace existing memory or neuromorphic computing for the next generation of AI hardware, making headlines due to its low processing cost and ultra-low power consumption.
A research team led by Professor Shinhyun Choi from the School of Electrical Engineering at the Korea Advanced Institute of Science and Technology (KAIST) (President Kwang-Hyung Lee) has announced the development of a next-generation phase-change memory device with ultra-low power consumption that can replace DRAM and NAND flash memory. A so-called phase change memory is a memory device that stores and/or processes information by using heat to change the crystalline state of a material to an amorphous or crystalline state, thereby changing its resistance state.
Existing phase change memories suffer from issues such as expensive manufacturing processes for large-scale devices and the need for large amounts of power to operate. To address these issues, Prof. Choi's research team has developed an ultra-low-power phase change memory device that electrically forms very small nanometer (nm) phase change filaments without the need for an expensive manufacturing process. This new development has the groundbreaking advantage of not only having a very low processing cost, but also being able to operate at ultra-low power consumption.
DRAM is one of the most commonly used memories and is very fast, but it is volatile, and the data disappears when the power is off. NAND flash memory is a storage device that has relatively slow read and write speeds, but it is non-volatile and can save data even when the power is off.
PCM, on the other hand, combines the advantages of DRAM and NAND flash memory with high speed and non-volatile. Therefore, phase change memory is being valued as a next-generation memory that can replace existing memory, and is being actively studied as a storage technology that mimics the human brain or a neuromorphic computing technology.
However, traditional phase-change memory devices require a lot of power to operate, which makes it difficult to manufacture practical mass storage products or implement neuromorphic computing systems. In order to maximize the thermal efficiency of memory device operation, previous research efforts have focused on reducing power consumption by reducing the physical size of the device using state-of-the-art lithography techniques, but they have encountered limitations in terms of practicality, with little improvement in power consumption, while cost and manufacturing difficulty increase with each improvement.
In order to solve the problem of power consumption of phase change memory, Professor Shinhyun Choi's research team has created a method for the electroelectric formation of phase change materials in a very small area, and has succeeded in achieving an ultra-low-power phase change memory device with 15 times lower power consumption compared to traditional phase change memory devices made using expensive lithography tools.
Prof. Shinhyun Choi is confident in the future development of this research in new research areas, saying, "The phase change memory device we have developed is of great significance because it provides a novel approach to solve a long-standing problem in production memory. The unit significantly improves manufacturing costs and energy efficiency. We anticipate that our findings will be the basis for future electronic engineering, enabling a wide range of applications, including high-density 3D vertical memory and neuromorphic computing systems, as it opens up the possibility of choosing from a variety of materials. ”
Link to original article
https://www.semiconductor-digest.com/kaist-researchers-developed-a-novel-ultra-low-power-memory-for-neuromorphic-computing/
Source | Semiconductor Industry Watch (ID: icbank) compiled from semiconductor-digest
☞ Business Cooperation: ☏ Please call 010-82306118 / ✐ or to [email protected]