While talking to David Eggleston, Principal, Intuitive Cognition Consulting* about the future of ReRAM/CBRAM, he pointed me to an interesting session at the latest Flash Memory Summit on ‘Flash below 20nm’. In the session, Applied Materials, SK hynix, Micron and SanDisk were all represented. The 4th major player in the field, Samsung, spoke in a different session highlighting their 3D NAND technology. The basic message of all the presentations was that 2D (planar) Flash is hitting a brick wall. Various physics based issues were described related to inter-cell coupling along with patterning challenges. But perhaps the biggest factor is simply one of economics, i.e. continued downscaling will no longer result in a decrease in the price per bit. I hope to have more on this once I connect with one of the session authors.
Back to the manufacturers. They all see a transition to 3D (vertical) NAND after the ‘final’ 1Znm generation of planar NAND with Z approximately 6. (However I wouldn’t be at all surprised if Z ends up being a bit smaller.) As of today, 19nm is approaching full scale production with various reports of 1Znm being in final development. Where the presentations differ is the roadmap and reasons for moving to 3D. The most aggressive is Samsung have already introduced 3D NAND technology and are selling 3D NAND based products. See last month’s Blogs for more details. Samsung reiterated that the transition gave performance rather than economic benefits over their latest generation of planar technology.
sk Hynix were quite clear that their next Flash memory product will be planar and based on a 16nm technology. They pointed to a number of technical issues of with the decreasing number of critical electrons in a Flash memory cell (approaching 10 as dimensions shrink to 10nm) described as most critical. They also highlighted a number of technical issues with 3D, noticeably the need yield improvement and the complex multi stack patterning and etching.
Micron/Intel transitioned away from the ‘traditional’ high aspect ratio floating gate NAND cell at 20nm (above). This was replaced with more planar high-K NAND cell. Essentially the floating gate, where the electrons are stored, became flat. This is a most impressive piece of technology development and must have been in development for many years. To my knowledge it is unique amongst the NAND memory manufacturers. Micron claim this will ease the transition to 16nm at which point they see a transition to 3D (Vertical) NAND technology with the important proviso that it ’allows Micron to introduce 3D NAND when it makes financial sense’ (Micron’s italics). Given the investment they have made in their 20nm high-K technology, I interpret this as meaning that they will do their utmost to squeeze another shrink from their planar technology.
The most conservative with respect 3D NAND turns out to be SanDisk who see their 1Ynm (a true 19nm) technology entering ‘meaningful’ production at the end of this year with 1Znm, a year later. Their 3D NAND (called BiCS) is shown as entering pilot production at the end of 2015, i.e. in over two years time. Furthermore BiCS is ‘targeted to provide a meaningful cost reduction versus 1Z’ Indeed BiCS is reported to have a smaller horizontal footprint per cell that their competitors.
It is fair to say that the transition will not be abrupt and that planar and 3D will co-exist for several years. See the figure below from SanDisk where they compare their approach to introducing 3D NAND (dark green) to a competitor (grey) along with SanDisk’s planar 1Z (red) technology. Note the x axis is not time but rather ‘Breadth of Customer Applications’. But given that generations of Flash memory typically occur at 12 month intervals, one can make a reasonable guess at the time scales involved.
Christie Marrian, ReRAM-Forum moderator.
*Dave can be contacted at [email protected]