HP Labs and Hynix Semiconductor, Inc. announced today that they have entered into a joint development agreement with the intent of bringing Memristor technology to market as a viable commercial memory product.  The new memory type called, ReRAM or Resistive Random Access Memory, has been the subject of much speculation and rumor since it was announced by HP in 2006. Although the deal is non-exclusive it comes as somewhat of a surprise that HP chose Hynix. The non-exclusive nature of the HP-Hynix deal enables future participation by other manufacturers required for commodity pricing and supply.

According to R. Stanley Williams, director of HP Labs? Information and Quantum Systems Laboratory, availability of the end-user products based on ReRAMs are due out by the end of 2013. ”This is a darkhorse technology,” Williams said. ”We think this will break out of the pack.’
Williams made it clear that HP does not want to be in ReRAM business. No financial details were disclosed.

Resistive Memory
Memristor Milestone Roadmap by HP LabsInvented in 1971 by Leon Chua at the University of California at Berkeley and disclosed in his September, 1971 paper "Memristor – The Missing Circuit Element", the idea lay dormant for 35 years before being resurrected and moved to practice by researchers at HP Labs.

Resistive memories are based on materials whose resistance can be electrically altered between high and low conductive states.  The structure of the non-volatile memory device consists of a titanium dioxide memory element sandwiched between an upper and lower cathode/anode metalized grid array. Because the memory recording technique involves resistance the effect of process voltage scaling is greatly reduced.  The simple patterning structure of the device allows vertical stacking of the arrays. Combined, they enable the all important multi-generational product roadmap. Resistive memory is considered a strong candidate for replacing Flash memory commencing at the 32nm process node

ReRAM is inherently resistant to cosmic radiation disturbs that affect DRAM and to a lesser degree Flash. The technology is capable of storing twice as many bits as Flash at much lower power. Williams said that target applications are any that are currently using Flash or DRAM – he further added that the technology has the potential to serve as a universal storage medium replacing DRAM, flash and disk drives. The implication being that ReRAM is the legendary universal memory.

HP seems to be following Apple?s "own and control" philosophy. They believe that the effort will allow them to produce superior products for their customers. Williams alluded to the fact that products supplied to the company from Hynix might contain special circuitry to help differentiate HP products from that of the competition.

Hynix has other non-volatile projects in progress including one in which they?re partnered with Samsung in work on MRAM technology as well as work with Grandis Inc. spin-transfer torque RAM. Other companies are working on resistive memory including 4DS, Adesto, Unity and several startups still under stealth. 

Earlier this year IMEC announced a program to investigate resistive RAM memory cells as part of their ongoing research effort.  Five of the leading memory manufacturers including Hynix are involved in this program.

Paradigm Shift in the Wind?
HP announced, earlier this year, that memristor technology can also perform a logic function.  The impact of reconfigurable logic where the difference between memory and logic is no longer separate could distinctly reorder the world of computing and semiconductor memory production.

Reconfigurable computing has been under investigation since the 1960?s.  Though much remains to be learned, the basic concept is that the data, the logic to process it and the code which stipulates the configuration of the logic all occupy and execute within the same device. Inter-device dynamic data flow controls pathways between arrays of such devices.  A dispatch mechanism launches and receives the results of spawned data applets. Leading FPGA companies have been deeply involved in this research which makes them additional candidates for HP?s resistive memory technology downstream.

The notion that very large arrays of reconfigurable logic using ReRAM is not lost on HP. In fact it is strikingly similar to the production process used for memory devices and this is not lost on Hynix. So what about the microprocessor companies?  I believe they?ll take a wait and see attitude.  By the way several of the under stealth companies are rumored to be working on "stuff"…