HP & Hynix join forces for memristor fab
3-year joint development
HP is partnering with Hynix to bring Memristor technology from lab to fab.
The Memristor, posited as a fourth basic electrical circuit element, is said by HP to have the potential to be a form of uniform memory that could replace flash (being non-volatile), DRAM and even hard disk drives. It could possibly perform logic functions. Memristors were brought from theoretical to actual existence in 2006 by researchers in HP Labs’ Information and Quantum Systems Laboratory (IQSL).
Hynix is a memory manufacturer and has been involved in research and development into what could follow flash memory as the limits of that technology are reached. A couple of years ago it was interested in STT-RAM (Spin Transfer Torque Random Access Memory) which is now being championed by Grandis.
The other main flash follow-on technology is Phase Change Memory (PCM) and this is being worked on separately and independently by Samsung and Micron, via its Numonyx acquisition.
The agreement between HP and Hynix will see them jointly developing memristor technology in the form of Resistive Random Access Memory (ReRAM). ReRAM uses materials that change resistance when a voltage is applied across them.
HP senior fellow and IQSL director Dr Stan Williams described what HP is sharing with Hynix: “It’s not just the memristor, there’s also architecture, circuit design, error correction coding – we’re bringing the complete package … Personnel from my lab will be going over to Korea and living there to help make this work. We’re committed to working together very closely to get a working product ready for the marketplace in as short a time frame as possible.” The two companies are aiming for a 2013 product announcement.
HP reckons memristor products could run 10 times faster than flash memory and draw 10 times less power. Williams says that they shouldn't have the limited write endurance of flash chips. They provide twice the capacity of flash for the same cost, but he didn't say whether this was a comparison with expensive single level cell flash or its cheaper multi-level cell variant.
HP and Hynix plan to begin jointly developing memristor technology in a range of areas and then choose the first products in which to deploy it later in the development cycle. They have set themselves a goal of doing this in three years so the product areas must be pretty well understood already.
The two intend to make memristor products that could be easily swapped for flash memory and use existing semiconductor manufacturing methods. HP said it isn’t planning to become a memory chip maker itself, or to restrict the licensing of its memristor technology.
It thinks it will get a time-to-market advantage. HP Labs director Prith Bannerjee said: "Practically everything that HP sells has some sort of memory in it: printers, PCs and enterprise systems – so there’s very little that HP makes that won’t in some way be touched by this.” ®
Memristors versus bubble memory
Regrading memristors going to the wayside like bubble memory, maybe. But the problem bubble memory had was it was trying to compete mainly with hard drives, and they became faster and cheaper than bubble memory. Bubble memory stayed on in applications that required ruggedness. Later on, when flash came out, it was so much faster than bubble memory that it displace it in the remaining markets.
Memristors? The big problem was to discover and develop them to begin with, which has been done. The manufacturing techniques appear to be quite conventional, they've been ramped up already to 1ns speeds, and should use less space per bit than flash (so it should be cheaper than flash), and no write life limit. I would not have guessed flash would have taken off as much as it has, given how much more costly it is than conventional hard disks, but it has -- and I think that is going to be memristor memory's competitor rather than competing with hard disks.
I agree, we see this egregious usage much too frequently, especially in TV commercials.
Anyone remember "magnetic bubble memory" from the 1980s, that was always going to be the next big thing?
Memristors will go the same way.