Overviews and forecasts of the semiconductor markets can provide a reality check on the more ‘controversial’ statements made about product roadmaps. Usually these are based on an optimistic view of technology readiness rather than the cold reality of market opportunity. I know, because I have made them myself! Recently a couple of substantial reports have appeared describing the semiconductor memory market with a strong emphasis on emerging non-volatile memory technologies (although I should add ‘emerging’ is interpreted differently in each case). Unfortunately these reports are costly (many thousands of dollars per license) so my comments are based on selected highlights that have appeared across the web. The report from Yole Development defines emerging memory as FeRAM, MRAM/STTRAM, PCM and ReRAM. It sees the major growth for PCM (NOR Flash replacement) and STTRAM (SRAM replacement) with both technologies being used to replace embedded Flash in applications such microcontrollers (MCU) for secure and industrial applications, although presumably Panasonic have other plans for the latter (see the recent Video post). The combined PCM and STTRAM business is forecaste to increase to $1.6B out of a total $2B for emerging memory in 2018.
The high volume opportunity for ReRAM is seen as NAND Flash replacement for the mass storage market. Here, the SSD will be the main driver as the technology moves from enterprise (server) based system to the desktop and laptop. This will start to occur when the (incumbent) NAND technology finally runs out of steam. Predicting this is fraught with uncertainty but SanDisk have been quoted as saying that there are two nodes left for conventional planar NAND (1Ynm later this year and 1Z next year). At that point, 3D NAND is seen as the successor, in SanDisk’s case, based on Toshiba’s BICS architecture. This basically involves building the NAND strings perpendicular to the wafer rather than in the plane of the wafer. 3D NAND has scaling issues of its own and when that limit is reached, the big opportunity for ReRAM is foreseen provided it is ready with cost-competitive chips with comparable density to 3D NAND. The big question, of course, is when will this occur? Yole at least are fairly unequivocal in stating that use of ReRAM (or 3D ReRAM) could start in the 2017-2018 time frame with ReRAM revenues surging from 2018. Again the initial opportunity will be in enterprise systems where the projected superior performance of ReRAM over highly scaled NAND Flash will be the driver as costs are driven down towards 3D NAND levels.
Another point alluded to in the Yole report is that ReRAM adoption is critically dependent on the major existing memory suppliers. At present 5 companies* (see if you can name them!) account for 90% of combined DRAM and NAND sales. It seems inconceivable that a company outside this group else could be in a position to supply ReRAM chips that would satisfy the mass storage market by 2018. Of course that it is not to say another company could not provide/license ReRAM technology or IP to one of the big 5 by 2018 as ARM have shown in the microprocessor market (apologies for the double negative….).
Christie Marrian, www-ReRAM-Forum Moderator
*Samsung, Micron, Sk Hynix, Toshiba and Sandisk