Will RRAM save Moore’s law? There’s a US $100B market at stake. The 2012 International Memory Workshop (IMW) was the best IMW thus far. Not sure if it was the quality of the papers which was excellent, the well organized workshop, or Milano Italy being the location. Let’s just say it was the combination. The event was held at the Melia Melano Hotel in Milano, Italy. The hotel was close to a metro station which made getting there easy and also convenient for exploring the city before/after the conference. One cool extra that the conference threw in was a metro pass good for all the days of the conference. And Milano’s central location in beautiful Northern Italy made it convenient for those who wanted to extend their stay and explore some awesome locations such as Venice, Verona, and the Italian Alps. Now for some technical stuff. It’s clear that NAND and DRAM are hitting the wall while the markets appetite continues to grow at an astounding pace. I know this “end of scaling” comes up every year, but I heard a difference at the 2012 IMW. The challenges are fundamental and on multiple fronts. For example, NAND is facing fundamental litho issues, over-whelming cell parasitic, too few electrons on FG, and ever decreasing reliability. The story is no better for DRAMs who have run out of rabbits to pull out of the hat. The challenges have been tough in the past and the industry has responded, but now the challenges are VERY tough and are going to require a disruptive change in memory technology. In comes the Emerging Technologies, namely RRAM, PCM, and STRAM. It’s agreed there won’t be a direct replacement, but a fast paced evolution adopting the advantages and applying the industry know-how to work around the disadvantages. By far, the emerging technology creating the most excitement is RRAM where Adesto Technologies is considered to be the pioneer. Many believe 3D RRAM is the obvious direction as was discussed in the SanDisk paper, “Flash Memory at the Cross-Road: Challenges & Opportunities” and it’s place shown in figure below.
Between the invited papers, the two sessions on RRAM, and the poster board papers, there were 18 papers discussing RRAM. Most focus is on bipolar metal oxide systems and the papers reported innovations in retention modeling, understanding and reducing variability in the SET/RESET, proper forming, and ideal cell construction for reliable and scalable memories. Although physics of operation is drawing high attention and excellent progress being made, it’s still considered the largest challenge to making RRAM memories mainstream. Other challenges raised were the stochastic nature of the switching and the higher currents needed for reliability. Many papers addressed these specifically. STRAM, pioneered by Crocus, also captured a fair amount of attention and still has a card to play in the evolution of the memory landscape especially in the embedded space. However, the material, operation, and scaling challenges are significantly more challenging than RRAM.
Shane Hollmer is a co-founder and the VP of Engineering of Adesto Technologies.