Phase change memory breakthrough could make it 1,000x faster than DRAM

By Shawn Knight · 9 replies
Aug 16, 2016
Post New Reply
  1. Phase change memory (PCM) is an area of interest that researchers around the globe have been investing time and money into for years. Exactly how viable it’ll be remains to be seen although new research out of Stanford University looks incredibly promising.

    PCM works by shifting between two physical states: crystallized (with low resistance) and amorphous (with high resistance). It already has multiple advantages over the storage technology in use today but it’s not perfect.

    As Extreme Tech notes, one industry-wide shortcoming that hasn’t yet been addressed has to do with latency. Memory clock speeds have increased greatly over the last 30 years or so but latency has improved much more slowly.

    Aaron Lindenberg, an associate professor of materials science and engineering at Stanford that led the research, looked into how quickly PCM begins to change from one state to another and more importantly, if they could somehow harness the speed. Their findings bode well for the future of PCM.

    Applying a 0.5THz electric pulse to PCM for picoseconds at a time creates crystallized filaments that can be measured and, in theory, be used to store data. This happens while the bulk of the material is still in its amorphous state but more importantly, the filaments are created on the picosecond timescale. Considering that conventional DRAM operates on the nanosecond timescale, PCM could theoretically be thousands of times faster than traditional RAM in use today.

    While promising, there are still plenty of hurdles that must be overcome. As the publication notes, THz pulses aren’t yet a thing in motherboards (at least as far as they (or we) know) as their signal doesn’t work well with the thin copper wires used in today’s hardware.

    What’s more, the memory industry can be quite volatile at times as anyone that has kept tabs on pricing over the years can attest to. It should come as little surprise, then, that chipmakers want to be certain (or at least, as certain as possible) that a new technology will be of benefit in the long run before shifting away from proven techniques (and by proxy, steering the market in a different direction).

    That said, the research is incredibly promising and further bolsters the potential of PCM. Speaking of, Intel’s upcoming Optane line of solid state drives (powered by its new 3D XPoint technology) is thought by some to at least partially be based on the principles of PCM.

    A leaked Intel roadmap from a couple of months ago indicates that Optane drives are slated to arrive late in the fourth quarter or perhaps early next year alongside the launch of Kaby Lake processors. Perhaps then, we’ll get some more information on exactly how XPoint technology really works.

    Permalink to story.

  2. MoeJoe

    MoeJoe TS Guru Posts: 711   +381

    Build it (inexpensively) and they will come ...
    p51d007 likes this.
  3. fps4ever

    fps4ever TS Addict Posts: 102   +74

    Seems like they will need to build a new type of tracing bridge from the memory to the cpu if copper isn't sufficient enough? And we all know how motherboard makers love to cut costs vs increasing them. :-|
    p51d007 likes this.
  4. Moneyd623

    Moneyd623 TS Member Posts: 19

    "What’s more, the memory industry can be quite volatile at times"

    Har Har, I see what you did there.
    p51d007 likes this.
  5. I was almost convinced that Intel and Micron(?) have indeed commercialized phase change memory. This article makes me believe that perhaps I was getting excited for nothing and that pcm is still not ready?
  6. Uncle Al

    Uncle Al TS Evangelist Posts: 3,350   +1,999

    Sounds like an inevitable development in the methodology, but I would simply ask, other than money, is there a downside on the technical side? More heat, electrical draw, etc?
  7. namesrejected

    namesrejected TS Guru Posts: 398   +299

    What peaks my interest is the possibility of storing more than 1's and 0's. This would be very useful for quantum processors. With a ratio of 1:1000 you could have ranges that have multiple values. Very interesting.
  8. miqui

    miqui TS Rookie

    Silver would be more ideal but the chemical coating for corrosion is expensive as all get out. Corrosion would negate the benefits, effectively shrinking the conductive cross section. And I think (I am probably wrong) the anti corrosion chemical reacts with one of the VOC's used in board manufacturing. The compound is used in pharmaceuticals., theadizol (sp?). So they would need to lay it down on the traces after the board gassed out. I dont know if that would be feasible.
  9. wiyosaya

    wiyosaya TS Evangelist Posts: 1,935   +764

    I have heard of this from time-to-time. When it is available for me to buy, I'll believe it.
  10. commanderasus

    commanderasus TS Booster Posts: 195   +72

    In research and development the hardest thing to do is to take your working model you built in the lab into the real world. This is a longs ways from being available in your local computer store.

Similar Topics

Add your comment to this article

You need to be a member to leave a comment. Join thousands of tech enthusiasts and participate.
TechSpot Account You may also...