It seems that barely a day goes by without an announcement related to ReRAM or CBRAM product related developments. In this entry, I’m going to restrict myself to news coming out of the larger companies prior to the start of this Blog. For these companies, most announcements reveal a significant level of integration and working silicon with at least multi-Mb functionality. In future entries we will update and also look at the news form smaller and emerging companies.
There were a couple of relevant presentations at the recent 2012 ISSCC conference. For example Panasonic described a cross point array ReRAM with impressive write throughput. One of the problems with scaling crosspoint arrays is that the leakage paths increase. Panasonic appear to have introduced a back to back diode structure associated with each cell to mitigate leakage. In the same session the National University/TSMC collaboration described their 65nm ReRAM with a low voltage current sensing scheme. Clearly another important development for moving ReRAM closer to volume production. As a reality check, it is worth noting that NAND Flash development continues apace with, for example, a 128 Gb chip being described by Toshiba/SanDisk. It is difficult to make an ‘apples to apples’ comparison of performance but it is clear that ReRAM’s write performance is far superior to Flash. (A subject for a future Blog entry?). Certainly, some of the highest throughput numbers were presented at the 2011 ISSCC in Sony’s CBRAM paper (213 MB/s write). Also at ISCC 2011, MLC (multi-level per cell) operation in ReRAM was described by authors from ITRI with 160ns write/verify. In between the two most recent ISSCC conferences, Elpida announced a 50nm ReRAM chip with DRAM like write speeds and laid an ambitious schedule for productization in partnership with Sharp. Given Elpida’s current problems, one should perhaps take this with a grain of salt but nonetheless it is impressive. Micron is also working in CBRAM as revealed in their winter 2012 investor briefing. In fact, Micron appears to be actively pursuing several alternative memory technologies including the Unity semiconductor complex oxide ReRAM although this may be impacted by the recent sale of Unity to Rambus. Sandisk’s analyst briefing from February 2012 describes ‘3D ReRAM’ following on from 3D NAND as a technology for the future. However, they claim that EUV lithography will be needed (I’m not sure why) which could be a problem! The Hynix/HP collaboration also has ambitious plans (and schedule) based on HP’s memristor cell as evidenced by various recent announcements and comments mainly from the HP team. Samsung have been relatively quiet recently especially when discussing functional ReRAM memory with a high level of integration. Some interesting tidbits have appeared at the recent IEDM and in the materials literature especially with respect to 3D integration. Macronix is another company who discussed a ReRAM cell structure with MLC operation at last year’s IEDM.
So, there is clearly a lot of interest at the main memory companies. Ambitious product schedules have been quoted but I suspect it may take a little longer for large capacity ReRAM/CBRAM to appear in the market place.
Christie Marrian, www.ReRAM-Forum.com Moderator