LONDON – Jai Menon, an IBM Fellow and chief technology officer with IBM's systems and technology group, has confirmed that IBM believes that phase-change memory (PCM), a non-volatile technology, could scale beyond where flash memory can go and could be of use in server computers in three to five years. Menon said he expects IBM to continue to develop its own PCM intellectual property but the company would likely offer that to a memory chip maker that could supply PCM components to IBM for use in server systems.
In a telephone interview with EE Times Menon said IBM is "very bullish" about PCM in the next three to five years and that it could have a role to play in server memory systems replacing flash memory. Flash is already making in-roads there because of the need to save power in server farms. Further out IBM's magnetic race-track memory could replace PCM, Menon said.
However, there are still a number of challenges to be overcome as Menon acknowledged. Menon did not provide EE Times with specific evidence of breakthroughs to address heater scalability or thermal limitations except to refer to a PCM array copper ionic conduction access-device paper published by authors from IBM at the Symposium on VLSI Technology published this year .
However, that access device is aimed at use in a stacked structure where thermal cross-talk has the potential to be an even more severe problem.
PCM is based on changing the material phase and the electrical resistance of a chalcogenide layer in each memory cell through the use of electrical heating. It is an attractive technology because of its non-volatility, theoretical high density and bit-alterability and has been touted as a possible replacement for both flash memory and DRAM. But the technology has proved difficult to commercialize and even as devices have made it to market using 90-nm and 65-nm process technologies, questions have been asked about the ability to scale the technology beyond flash memory, which is already being made at close to 22-nm.
@Charles.Desassure: I agree, I did not find anything that we don't know already about PCM & PRAMs available in public domain.
The thermal cross talk is a problem in both 2D and 3D applications, the latter being the worse. The thermal coupling through the substrate between junctions. Thinning the substrates and insulating layers may mitigate this to some extent but that is a change for the worse in 3D stacking.
@iniewski: I hear you, I don't see how PCM can scale better than flash; heating chalcogenide will put space restrictions on scaling. The alternative is worse, if switching rates are needed less than 5ns, there is a real possibility of thermal avalanche mode with scaling.
If IBM is 3 to 5 years away from commercializing this technology, they sure have their work cut out for them. Samsung may beat them to the punch!
Dr. MP Divakar
It's an interesting question. The crosstalk can be considered just from two adjacent pillars/stacks (same plane), with one having current through it. But I agree that 3D stacking (is that what you mean?) potentially could be more severe.
Thanks for this article. I have read this article over and over again and I am still wondering why Peter Clarke interviewed Jai Menon? Jai Menon is not saying anything that we don’t already know. We know that IBM is working on different projects. We need to ask IBM hard-hitting questions so we can know what they are during and where they are going? When a company is talking about three to five years on a project, to me; that sounds like you have more research to do and that is not "very bullish". But I am happy that Jai Menon is working on this exciting project.
I remain being skeptical here, I just don't see PCM scaling better than flash...working in a semiconductor industry for several years I have seen hundreds of cases when the company publicly talks about one technology but quietly pursues something different internally...Kris
Interesting. So it took 10 years for Micron to publicly admit in 2004 that PCM does not scale, after having signed an exclusive, royalty-bearing license to the phase-change memory patents in 1994. Then it took 10 years for Intel/STM/Numomyx to admit that PCM does not scale (a tacit admission, by failure to deliver the promised 45-nm chip in the beginning of this year). Samsung, having failed to produce the promised cell phone with phase-change memory by June 30 this year, should be next.
So, when will IBM admit that PCM does not scale? Looks like, in 2015, 10 years after that ill-fated "joint research initiative to explore the potential of a new form of computer memory technology called phase-change memory."
You can't make this up. And somebody at IBM is DEFINITELY not doing his/her job.
Mainly because I was asking Jai Menon about phase-change memory in the light of his recent reported comments (see foot of article).
Menon did say that IBM thinks magnetic "race-track" memory could be successfully a bit further out.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.