Chipmakers are beginning to incorporate multiple types and flavors of DRAM in the same advanced package, setting the stage for increasingly distributed memory but significantly more complex designs. Despite years of predictions that DRAM would be replaced by other types of memory, it remains an essential component in nearly all computing. Rather than fading away, its footprint is increasing, and so are the number of options. There are several factors driving this expansion. Among them: o The number and density of compute elements in a system are increasing to process more data faster, particularly for AI/ML and data center applications. More cores require more memory, because there needs to be a place to store data as well as process it. o Normally, much of this would be dealt with in cache, and SRAM has been the memory of choice for L1 through L3 cache. But SRAM no longer is scaling at the same rate as digital logic. As a result, it takes up a larger percentage of real estat
In the past, analysts, consultants and many other experts have tried to estimate the cost of new chips using the latest process technology. They concluded that by the 3-nanometer node, only a few companies could afford it, and by the E-level node, no one might be able to afford it. Much has changed in the last few process nodes. More and more startups are succeeding in making advanced node chips that cost far less than those highly quoted numbers. Behind those numbers are some broad changes in chip design and manufacturing. Among them: Many advanced node chips are either highly replicated arrays of multiple cumulative processing elements for AI/ML. These are relatively simple compared to integrating different elements on a single chip, but require characterization of their thermal issues, noise, and various use cases and applications. Advanced packaging techniques, which have become mainstream since the creation of these early estimates, allow chipmakers to bundle together chips