With the development of AI technology, besides the different requirements for processors, memory faces different challenges from before, according to Jinman Han, president of Samsung Electronics Memory Product Planning Senior Unit.
Jinman Han, speaking at the GSA Memory + Summit Forum, said that in the field of artificial intelligence, the types of storage devices are more than ever before, and have developed from a single storage category in the PC era to a variety of products suitable for different products in different fields to meet the needs of different devices for storage.
This makes the future development of the whole industry show a different feature from the past, that is, more and more manufacturers realize that the technological structure of equipment is changing, the CPU-centric design thinking in PC era is failing, and memory is becoming the center of design.
Jinman Han believes that changes in design architecture have led more and more vendors to think about how to squeeze existing resources to improve memory performance, and even Intel is moving towards memory-centric direction.
Under the influence of this idea, memory will face more and more challenges.
Firstly, the fundamental reason for the change of design architecture is that the performance of processors has been exponentially increasing, while the performance of corresponding memory only keeps linear growth, which makes the performance gap between processors and memory wider and wider. "Processors can improve performance by stacking numbers, while memory cannot simply use this method."
Secondly, Jinman Han emphasizes that with the change of memory cell and the increase of storage density, although the performance of memory has improved, the power consumption of memory has also increased. From the current development situation, power consumption has become the biggest problem to improve the overall performance of memory.
Finally, the process and stack layer limits. In Jinman Han's view, the number of layers of memory can not be increased indefinitely. To solve this problem, one solution is to increase the size of storage blocks to avoid more layers of stacking. Jinman Han emphasizes that stacking of layers has become increasingly uneconomical, and also poses a great challenge to the security of storage. The problem of a single block will lead to a greater problem of data storage.