3D is the memory chip buzzword du jour, because reducing feature sizes is expensive, while stacking memory cells up is much less so. Fabs can use older, stable processes to build larger capacity chips with less technical risk.
Samsung has announced products using their 3D flash. But NAND flash is not the only memory technology that can benefit from 3D economics.
Where speed and reliability are critical various forms of Resistance RAM (RRAM or ReRAM) are starting to make inroads. Crossbar, a venture-backed startup in Silicon Valley, is pushing the 3D bandwagon even further.
Crossbar's RRAM technology is 20 times faster than NAND flash in reads and writes, while using 1/20th the energy. They've also shown they have 3D scalability with their "1TnR" (1 Transistor driving n Resistive memory cells) technology, where a single transistor can drive hundreds of memory cells.
You think that's air you're breathing?
But reliably reading cells in a high-density 3D matrix isn't easy due to "sneaky" current leakage between the closely-packed cells. How do you know the current being read is the right one?
This is a major problem. As an abstract for another recent RRAM paper says:
Although several reported resistive switching structures show excellent memory characteristics, implementation of high-density Resistive RAM arrays lags behind, mainly due to the sneak currents issues.
That's the problem that Crossbar is saying it solved in today's paper at the International Electron Devices Meeting (IEDM). IEDM focuses on reporting breakthroughs in nanometer-scale CMOS transistor technology.
The Field Assisted Superlinear Threshold (FAST) Selector suppresses sneak currents and gives a clean read. Better yet, it is implemented as a passive crossbar array, a relatively simple structure to fabricate.
The Storage Bits take
While NAND flash will continue to hold a huge cost advantage over RRAM, there are significant applications where a fast, reliable and low-power non-volatile memory is economic despite extra device costs. Crossbar believes they can eventually build 1TB RRAM devices using 3D techniques.
RRAM's lower energy requirement - NAND flash requires ≈20v for a write - and 3D density will make it a natural for phones, phablets and notebooks, as well as power-constrained sensors. It looks like the technical problems are being solved and that commercial RRAM will be available in 2016.
Our digital civilization requires reliable, long-lasting digital storage. RRAM is an important step in that direction.
Comments welcome, as always.