An innovative new memory device that combines the features of DRAM and NAND flash memory is being developed by a group of researchers at the Korea Advanced Institute of Science and Technology (KAIST).
The team's breakthrough, led by Professor Shinhyun Choi from the School of Electrical Engineering, is a cheaper and potentially used method to replace existing memory solutions or implement neuromorphic computing in next-generation AI hardware. We promise power efficient solutions.
According to KAIST, the new device utilizes next-generation phase-change memory with ultra-low power consumption and can replace both DRAM and NAND flash memory.
Adopt a novel approach
DRAM typically provides fast performance, but is volatile and loses data when power is removed. NAND flash memory provides a solution by retaining data even when the power is turned off, but it cannot match the speed of DRAM. This new phase-change memory offers a non-volatile, high-speed solution that combines the best of both worlds.
Previous versions of phase change memory suffered from high power consumption. Despite attempts to reduce consumption by shrinking the physical size of devices using cutting-edge lithography techniques, the savings have been minimal and costs have skyrocketed.
To overcome this, Professor Choi's team established a method to electrically form phase change materials in an extremely small area, and succeeded in developing an ultra-low power phase change memory device. Notably, it consumes 15x less power than previous phase-change memory models that used expensive lithography tools, a major advance in the quest for cost- and energy-efficient memory development.
“The phase-change memory device we developed is important because it provides a new approach to solving remaining problems in manufacturing memory devices with significantly improved manufacturing costs and energy efficiency,” Professor Choi said. , further expressed his expectations. This new research lays the foundation for future electronics, paving the way for high-density three-dimensional vertical memory and neuromorphic computing systems.
This isn't the only neuromorphic computing solution we're working on at KAIST. Last month, local scientists unveiled an AI chip that matches the speed of Nvidia's A100 GPU, but is smaller and uses significantly less power.
