Non-silicon memory technologies have been studied for about as long as have silicon-based technologies, but the silicon technologies have always been preferred. Why is that, and why should anything change?
This is a question that The Memory Guy is often asked. The answer is relatively simple.
Silicon memory technologies benefit from the fact that they have always been manufactured on process technologies that are nearly identical to those used to produce CMOS logic, and can therefore take advantage of the advancements that are jointly developed for both memory and logic processes. In fact, before the middle 1980s, logic and memory processes were identical. It wasn’t until then that the memory market grew large enough (over $5 billion/year) that it could support any additional process development on its own.
Even so, memory processes and logic processes are more similar than different. This synergy between memory and logic continues to reduce the process development cost for both memories and logic.
Emerging memories depart from this path. Nearly all emerging memory technologies use new materials that aren’t used in logic processing, so they don’t benefit from this synergy. When compared to silicon, these new materials are not as well understood, and that lack of understanding can lead to yield issues.
So what’s different about today? Why are emerging memories suddenly falling into favor?
Consider that the reason NAND flash is migrating to 3D is because a planar floating gate, the basic bit cell for NAND and NOR flash, can’t shrink beyond about 15nm. This was the point at which all NAND flash manufacturers converted (at great difficulty and expense) from planar technology to 3D.
Consider, too, that the semiconductor industry’s most advanced logic processes have already shrunk to 10nm and are now starting to ramp to 7nm. A number of the SoCs (Systems on a Chip) that are built on these logic processes would benefit by including nonvolatile memory for firmware, but the memory would need to be produced on a 15nm process. A chip with 7nm logic and 15nm flash might be less economical than the same chip would be using 15nm processes for both logic and memory. From this perspective it seems inevitable that some new technology will be required in order to allow nonvolatile memory to continue to scale with logic.
If a new memory technology is developed for logic chips, then the process know-how from that development effort can be applied to stand-alone memory chips for a very modest additional cost. This implies that a large market for discrete emerging memory chips could also develop. There are other factors at play in this case, though, and they are explained in great depth in the report that this blog post is drawn from.
Considerable effort has already been spent to develop emerging memories to replace NAND once it reached its scaling limit, which indeed occurred at 15nm. NAND flash developers, though, managed to find a way around the issue by moving to 3D, and that has put the NAND market out of reach for the next several years. Emerging memory companies have shifted their focus and now hope that an impending DRAM scaling limit, perhaps as the technology shrinks below 10nm, will open up a significant market for their technologies. This may indeed occur, but DRAM developers still have a number of routes to further scaling.
From today’s perspective it appears that emerging memory technologies are likely to first reach high volume production as the embedded memory part of logic SOCs, and subsequently migrate to become an important part of the discrete memory market.
This post is the first of a series on emerging memory technologies, looking at it from several angles, and predicting how these technologies will change both the chip market and the market for the capital equipment used to produce these chips. It consists of excerpts from a recently-released report from Objective Analysis and Coughlin Associates: Emerging Memories Poised to Explode.
There are six sections:
- Why Emerging Memories are Necessary
- Understanding Bit Selectors
- The Technologies: MRAM, ReRAM, PCM/XPoint, FRAM, etc.
- Process Equipment Requirements
- Emerging Memory Companies
- Forecasting Emerging Memories
The Memory Guy has provided these to help readers understand the emerging memory technologies and markets. Questions and comments are appreciated.