It was less than 24 hours ago since the last post echoing an article on the NYT on memristors, and two more articles have appeared on both the NYT and CNN with the reports of the announcement, made on Tuesday by HP, that it would commercialize a new computer memory technology with Hynix, the South Korean chip maker.
From the articles, we learn that HP and Hynix will build memristors-based computer memories, with the goal of making the new technology commercially available in about three years. As Neurdon readers know, memristor main application is for dense nonvolatile memories. This technology can be a likely candidate to substitute current flash memory cards for products like cameras, cell phones, and PCs, and possibly beyond. What makes us even more excited is the fact that this substrate will be used as well for advancing neural simulations, and open up new, uncharted territories in the scale, miniaturization, and portability of neural models. It is nice to learn that a commercialization roadmap has been put in place that will facilitate the development of such technology. More info on this can be found in the CNN and NYT posts.
Below is the video with Stan Williams, ays Stan Williams, HP Senior Fellow and founding Director of the Information and Quantum Systems Lab (IQSL), commenting on the news, available here.
Another related article found at EETimes (link below) correctly notes the tendency of Hynix to hedge it’s bets in next generation memory. They have been developing phase change RAM since 2003 and have licenced Ovonyx’s phase change patents in 2007. Hynix has also collaborated with Samsung and Grandis Inc. on MRAM. Stan Williams and HP are obviously very good at getting publicity for what they persist in incorrectly calling a “memristor.” But as the article states “HP does not want to be in the ReRAM business” and my bet is that as phase change gains traction Hynix will drop the memristor.
http://www.eetimes.com/electronics-news/4207222/HP–Hynix-move-to-commercialize-the-memristor-semiconductor
Hi Blaise,
thanks for the comment. I am not sure the whole agreement is in place for a very expensive publicity stunt, but time will tell.
We will have this post as “the Blaise prophecy”….
Max
Don’t get me wrong. I do believe Hynix will legitimately attempt to develop a metal oxide version of ReRAM in collaboration with HP by 2013. My point is that they have also worked with other companies in developing Phase Change RAM and MRAM and are likely to invest in any other next-gen non-volatile memory that looks promising. It is a smart business move especially considering that HP is one of the world’s biggest memory customers. However, metal oxide RRAM has a lot of catching up to do relative to MRAM and PCRAM.
I also don’t think non-volatile memory is actually a good application for memristive devices because:
1) It is a multi-billion dollar market. This may sound good on the surface but it is usually very difficult to create disruptive changes in large markets and incremental changes to well established technology is most often preferable.
2) It is very likely that Flash transistors will be scaled using vertical nanowire FETs within the next ten years (Samsung, Infineon, and Micron have all demonstrated this in the lab). There is also concurrent development in multibit cell versions of Flash. If memristive memory is to compete it will have to outperform these technologies which are more compatible with existing Flash.
3) Intel and Samsung have demonstrated support for Phase Change memory and there is a much longer history of development going back several decades to Ovshinsky’s work on the ovonic switch in the 1960′s (Incidently Ovshinsky also developed numerous neuromorphic designs based on phase change memory). In contrast, although there was some work done in metal oxide RRAM in the 1960′s, most of the recent work was only initiated a few years ago and there are many more unknowns.
4) The memristive systems approach of Chua and Kang is actually more useful to applications involving analog processing or to neuromorphics than to non-volatile memory. Memristive systems is after all a subclass of dynamic systems analysis which is typically used in analyzing signal processing or control systems rather than memory arrays. If one manufactures ReRAM arrays having binary rather than analog resistance states is there really any practical usefulness in applying memristive systems modeling at all (other than publishing academic papers and getting attention in the press)?
However, even given all of the above points I think it is good in the long run that HP is promoting memristive ReRAM and that Hynix is working toward production. They will fail but their failure may pave the road for the development of more useful application of memristive technology.
Blaise,
Resistive memory research is a valid and perhaps one of the most advanced memory developments at present. But, for about 7 years, the literature does not show any major resolution on how to make it a manufacturable technology.In the case of HP memristor, they disclosed it themselves that the purpose was to fight the Samsung patent on the “resistor switching action”. The problem is that since the thirties, when NiO violated band theory, the Metal/Insulator transition is well studied and produce at least 3 Nobel Prices. In fact there are many types of resistive Rams – MRAM is one of them. A memcapacitor is an FeRAM. In the case of what the “memristor” came to represent, based on an elementary circuit synthesis that ignored the physical mechanism that causes the two resistive states, it is truly a sham. For example, if you look at the IV , pick a voltage and on the y-axis pick two current levels. That will define the two resistive levels: Behold, the Model, I1R1=I2R2, the problem is, what causes I1 and I2 be histerectic – that where device and materials physics comes in. The reason is that in nature there are only 3 types of insulators: Anderson, Mott and band. If you do not know this, you think that the extended band states are real in Transition metal, and you do not account for electron pairing. It is a bit like explaining Superconductors with V =IR. Thus, shame on you HP. What the heck!!! In the case of their data in the characteristics of the Titanium Oxide Memristor, we in the field can easily see the lack of stability and symmetric hysteresis. Why? because all d-block transition metal oxide will behave as a memristor, but not as the fictional; device that HP disclosed. With only 30 nm thickness, charge injection will precipitate a Metal / insulator transition, however in order to get a stable device, they need to double the thickness and have a massive reduction in oxide vacancies. Instead, they ignored all the known mechanisms of such a transition and decided to model “Vacancy champaign bubbles” as they described it. The electrons as they interact with the d-orbitals create the phase transition. But, vacancies make every switching a random event – thus, there is no technology. But, HP can walk away after Hynix fail and slowly hide away this stunt. The ability to retain even at 400C has been demonstrated in true ReRAMs. It is a combination of Metal/insulator transition caused by electron density variation. The switching is controlled by ballistic transport and the switch in the density of states caused by electron-electron correlation. Unfortunately, if you never study anything beyond semiconductors, you just cannot explain and perfect something that behaves without the kinetic energy term of a schrodinger’s equation. Thus, we have the surprising display of a major US company telling the world a bit of pseudo science and now involving a gun for hire with Hynix.