Crossbar architectures have a notorious "sneak path" problem requiring non-linear elements such as diodes which creates manufacturing and power issues.
The first ReRAM products will more likely use 1T1R memory cells to avoid sneak paths. In fact Panasonic has just released the first ReRAM evaluation kit based on a 1T1R memory cell using tantalum oxide as the memory resistor.
Thank you for citing examples of RRAM architectures that use transistors. The RRAM designers mentioned in my article are using pure crossbars without transistors in order to attain the highest possible densities.
This is not generally true. There are several examples in the patent literature of using selection transistors in RRAM memory cells. The "memristor" is good for propaganda but does not explain anything about the physics of ReRAM despite the BS of Stan Williams and Leon Chua.
The common element between memristors and RRAM is that they require no transistors in their bit cells--just the resistive material--thus allowing much smaller bits by virtue of using ultra-high density passive crossbars.
IMEC is a minor player in ReRAM. They have only one relevant US patent covering a form of ReRAM based on NiO(US7960775). Even this patent has questionable validity in view of earlier patents from Sharp and Samsung which teach graded oxide forms of ReRAM.
Calling ReRAM a memristor is good for getting press attention but as far as I know provides no useful information to help manufacture ReRAM devices.
I compiled some comparative patent data related to memory resistor patents at the following link:
@resistion, that would appear to be an important distinction. How can anyone make an optimized device if we don't have a good understanding of the underlying physics? Are HPs interfacial memristor and IMECs filamentary RRAM really two sides of the same coin?
Are the RRAM and memristor the same device? I don't think there is consensus yet. Just because Chua and HP say so, only makes them a vocal minority at this point.
RRAM performance metrics generally do not necesarily follow memristor behavior, so that's why we haven't see much overlap in the literature.