The Challenge Of In-Memory Computing

In the face of increasingly serious storage wall problems, memory access power consumption problems, and the drive brought by artificial intelligence applications, the integration of computational storage/in-memory computing/storage computing provides a promising method. Judging from the current implementation methods, the integration of computational storage/in-memory computing/storage computing is divided into two routes: based on mature volatile storage and immature non-volatile storage. Regardless of the route, there are certain challenges:

1) Based on mature volatile storage: The integration of computational storage/in-memory computing/storage computing in this way requires the fusion of processor technology and memory technology. However, due to the different manufacturing processes of the current processor and the memory, if the function of the memory is to be realized on the processor, the storage density of the memory may be reduced; if the function of the processor is to be realized on the memory, the performance of the processor may be affected. run speed. It's hard to get a good compromise between performance and capacity.

2) Based on immature non-volatile storage: The natural fusion of storage and computing of non-volatile storage is the best device for building computational storage/in-memory computing/storage-computing integration. However, due to the current manufacturers and processes are not mature. Economically, apart from more investment in the existing memory manufacturing facilities used to produce these new technologies, it will be very difficult to get users to switch to new technologies as long as they can still use DRAM or Flash.

Sourse:https://en.witmem.com/news/news/the_challenge_of_in_memory_computing.html