Why is the computing in memory market favored?

The integration of storage and calculation technology is regarded as the core of artificial intelligence innovation. The organic combination of storage and computing, directly using the storage unit for computing, greatly eliminates the overhead caused by data migration, and solves the "storage wall" and "power consumption wall" problems of traditional chips when running artificial intelligence algorithms. Improve the efficiency of artificial intelligence computing by ten times or even a hundred times, and reduce costs.The reasons why I have always been optimistic about the integration of storage and computing technology: First, the computing power and the amount of computing data are increasing exponentially every year. However, Moore's Law is close to the limit, and each generation of chips only improves performance by 10-20%. The second is that the computing power of the von Neumann architecture has been limited by the memory wall, and only by solving the problem of the memory wall can the computing power be further improved. Among various solutions, computing in memory is the most direct and efficient.Low-power sub-threshold computing optimizes the power consumption of existing logic computing. Generally, the energy efficiency can be increased by 2-4 times, but the computing power is correspondingly reduced, so only targeted optimization can be carried out. Computing in memory is a new type of computing architecture, which performs two-dimensional and three-dimensional matrix operations, and its energy efficiency and computing power can be increased by 100-1000 times. Neuromodal computing is a chip designed for brain-inspired algorithms. There are different implementation methods, such as analog computing, digital computing, clockless computing, or integrated storage and computing. In the post-Moore era, it is impossible to optimize the overall computing power through technological improvement, and heterogeneous computing and new architectures become more important.