Emerging types of non-volatile memory chips will gain considerable ground in mainstream applications by 2020, according to the latest forecast from Yole Developpement. However, DRAM and NAND will remain to be the dominant memory technologies for years.

The global market for today's emerging non-volatile memory (NVM) technologies will expand from $65 million in 2014 to $7 billion by 2020, an impressive rise of more than 118 per cent per year. However, this business will represent less than 10 per cent of the total 2020 memory market, so there's still a long way to go before the newcomers replace today's NAND flash and DRAM chips.

Memory chips are key building blocks in everything from cameras to cars. Thus standalone chips represent one of the largest semiconductor market segments, totalling $78 billion last year, according to Yole. The need for memory will surge in the next few years especially thanks to the continued explosion in Internet traffic driven by the rising number and use of connected devices.

NAND and DRAM are well established, but their scalability is becoming increasingly complex and problematic. Thus new non-volatile memories striving to become substitutes have been in the works for more than a decade. The list of contenders includes RRAM, MRAM, STTRAM, PCM and FRAM. In the last two years, RRAM and STTMRAM have taken a leading position whereas PCRAM has fallen from grace.

So far, sales of MRAM/STTMRAM and RRAM are still limited mainly to niche industrial and wearable markets due to their limited density. MRAM/STTMRAM new product availability has been delayed by Everspin, Crocus Technologies and Avalanche Technologies. Micron's PCM sales have declined sharply due to the collapse of the entry-level phone market.

On the bright side, several alliances targeting STTMRAM development have helped build an ecosystem. Panasonic and Adesto introduced additional RRAM products, and Micron has announced plans to introduce two memory products in 2015 and 2018.

Improving scalability and therefore price per gigabit is the key to winning mainstream NAND and DRAM applications. Presently, limited product densities increase the price per gigabit to several $100/Gb, a limiting factor for mass market adoption since NAND and DRAM chips cost less than $1/Gb.

In the next five years, scalability and chip density of the new memories will be greatly improved, and this will open up many new applications.

Enterprise storage will be the killer market by far for emerging NVM chips in 2020. The novel technologies will greatly improve data centre storage performance, where requirements are intensifying as traffic rises. To reduce system latency, customers will use STTMRAM and RRAM at first as cache in new types of storage-class memories. Then DRAM will start to be substituted by STTMRAM in 2018, if STTMRAM's costs decline sufficiently.

MCUs for wearables, smart cards and other markets will increasingly adopt the emerging memories as the scaling of embedded MCU flash runs out of steam, especially after 2018 at the 28nm node. In the short term, the emerging memories will be adopted rapidly in the wearable and smart card markets thanks to their lower power consumption and increasing density.

Mass storage is being served by NAND memory, but 3D NAND scaling is expected to slow down by 2020. In the meantime, 3D RRAM scalability is expected to surge thanks to big memory makers' efforts and a new entrant, Crossbar, that is targeting an aggressive time to market for a terabit chip, expected in 2018. This market will start getting traction prior to the end of scaling of 3D NAND thanks to RRAM's benefits of higher speed and lower power consumption.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved