3D Solid State Memory, Part One

http://www.forbes.com/sites/tomcoughlin/2015/07/30/3d-solid-state-memory-part-one/

This is the first in a two-part article on three dimensional non-volatile solid-state memory/storage devices. The first part will talk about 3D NAND flash technology, but will also include some discussion of the recent Micron/Intel 3D XPoint non-volatile memory technology announcement.
3D NAND flash memory products were introduced by Samsung Semiconductor at the 2013 Flash Memory Summit (Tom Coughlin is the general chairman of the Flash Memory Summit) and began shipping in 2014. The basic idea of 3D NAND is to stack the flash storage cells vertically on the silicon substrate. The shipping Samsung products have 32 of these cells stacked in these holes to create a storage density much higher than with planar flash memory with the storage cells side by side on the chip.
As a consequence of the 3D structure, the lithographic features on the NAND chip can be larger, resulting in products with higher write endurance. This makes increasing the number of bits per cell easier and so Samsung V-NAND products are often three-level cells (TLC) rather than the two-level cells (MLC) that have been common in client and enterprise SSDs. More bits per cell and more cells per chip give a higher storage capacity with a reduction in the cost per bit.
Samsung is now shipping several client SSD products using its version of 3D NAND, called V-NAND with capacities up to 1 TB. Since the Samsung introduction all the major flash memory manufacturers have announced plans to bring 3D NAND flash to market by 2016. SanDisk and Toshiba announced their plans to move to 3D NAND by 2016, believing that they would be able to support the industry with their planar NAND before then. Micron and Intel recently announced their own 3D NAND products that could enable 3.5 TB M.2 SSDs as shown below.
This week Micron and Intel announced their 3D XPoint non-volatile memory technology and Intel said that they would be sampling 128 Gbit XPoint chips shortly.  The companies said that their XPoint technology combines the performance, density, power, non-volatility and cost advantages of all available memory technologies on the market today. The technology is up to 1,000 times faster and has up to 1,000 times greater endurance than NAND, and is 10 times denser than conventional memory.  They didn’t give many details on the technology behind XPoint except to say that it is a cross-point storage technology and doesn’t require CMOS transistors associated with the memory cells.  Instead the company appears to be using diode switches with the cells rather than transistors.  This helps with increasing memory density since diodes take less space on the chip than transistors.

There are several candidate technologies that Micron and Intel may be using. These include several types of resistive RAM (ReRAM), spin based MRAM (ST-MRAM) and Phase Change Memory (PRAM). Many of these technologies use a cross-bar architecture and in principal the cross-bar architecture can be scaled to the third dimension, allowing higher density storage per chip area. Intel has reportedly said that this technology doesn’t use ReRAM and that has led some to speculate that it may be a Phase Change memory (see Daily Tech article by Jason Mick,
Micron and Intel have worked on Phase Change memory for many years and hold or license several patents on this technology. Several years ago Micron began shipping some phase change memory for evaluation and at the 2014 flash memory summit HGST was showing HDDs that included some 3 year-old Micron phase change memory chips as a cache in the HDD. Micron and Intel have invested heavily in phase change memory in the past and own many patents on this technology. It is likely that the XPoint technology is a 3-D phase change technology using diodes and that Intel and Micron are planning to use this technology to support Intel’s up-coming generation of processors.
After Chinese Tsinghua Unigroup made a $23B bid for Micron, Micron and Intel have announced a new fast and cheap non-volatile memory technology that can be scaled to 3D architectures without transistors, that they indicate is ready to roll out. This will increase the perceived value of Micron but also could change the face of non-volatile memory applications in the next few years and further validate the role of 3D memory.