Stories
Slash Boxes
Comments
Search

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Intel Doubles Capacity of Likely Flash Successor

Posted by ScuttleMonkey on Mon Feb 04, 2008 01:11 PM
from the double-or-nothing dept.
Technology Hardware
Intel has announced a new technique that allows them to effectively double the storage capacity of a single phase-change memory cell without adding cost to the current fabrication process. "Phase-change memory differs from other solid-state memory technologies such as flash and random-access memory because it doesn't use electrons to store data. Instead, it relies on the material's own arrangement of atoms, known as its physical state. Previously, phase-change memory was designed to take advantage of only two states: one in which atoms are loosely organized (amorphous), and another where they are rigidly structured (crystalline). But in a paper presented at the International Solid State Circuits Conference in San Francisco, researchers illustrated that there are two more distinct states that fall between amorphous and crystalline, and that these states can be used to store data."
+ -
story
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

Intel Doubles Capacity of Likely Flash Successor 50 Comments More /

 Full
 Abbreviated
 Hidden
More
Loading... please wait.

  • Salt shaker please (Score:3, Insightful)

    by techpawn (969834) on Monday February 04 2008, @01:22PM (#22294472) Journal
    "Intel Doubles Capacity of Likely Flash Successor" this from a site that had a huge Intel logo on it for how many months?

    It's neat tech, but as long as flash keeps getting bigger and cheaper we won't see it's 'Successor' for a while.

    • Long time in the lab (Score:4, Insightful)

      by EmbeddedJanitor (597831) on Monday February 04 2008, @01:35PM (#22294734)
      Even when a technology becomes shippable it tends to take quite a while for it to catch on. It is easy to make small lab batches, but reliable low-cost high-volume production takes a lot longer. NAND flash was invented in 1988 but only really got going in around 2003 - 15 years later.

      • Re: (Score:3, Insightful)

        by networkBoy (774728)

        It is easy to make small lab batches
        I see you don't work in an R&D lab doing PCM...

        Before I left my former job we were working on PCM.
        It was anything but easy to make in small batches in the lab. Our average yield of 100% good die was under 1 die/wafer.
        We had plenty of 50% dice, but very little fully functional ones.
        -nB
        • He choose a poor phrasing.
          He should have said something like:
          "Relative to each other, it's easier to make one functional unit in a lab than to make 1000 functional units a day"

      • Except... (Score:3, Interesting)

        by Khyber (864651)
        ...the reliable low-cost high-production facilities already exist, as the process deviates very little from the CMOS manufacturing process. It's the same material that is used in rewritable optical media, and on top of that, it's basically just glass. Where you once needed stable unchanging silicon for memory/data storage, now we're just using different states of glass. Most of your concerns are addressed in this technology, and this is why I'm watching it very closely. Go read up a bit here. (PDF WARNING) [ovonic.com]

    • Re:Salt shaker please (Score:4, Informative)

      by Microlith (54737) on Monday February 04 2008, @01:49PM (#22294938)

      as long as flash keeps getting bigger and cheaper we won't see it's 'Successor' for a while.

      As I understand it, flash (nand) capacity grows with the shrinking of the trace size. It's also cheap because it's produced in mass quantities.

      Everything that has made flash high capacity and cheap can be applied to PCM, only PCM has a number of advantages:
      - more durable, since it doesn't force high voltages over blocks to erase them
      - smaller cells, allowing more to be packed in the same space
      - rewriteability. You don't have to erase a block to change a single byte. It's more like RAM or hard disks in that respect.

      So what will likely happen is a slow change from FLASH to PCM as the major flash manufacturers transition their products to this technology. It'll still have the same form factor, and most people won't notice aside from an increase in capacity.

      IANAPCMEBIWNS (I am not a pcm expert but I work near some...)
      • I'm not sure about the density of PCM vs flash but one of the biggest advantages of flash right now is that it uses the same fabrication process as other semiconductors. It's still CMOS.

        Consequently, you have economies of scale that translate from the other microelectronics markets. More importantly, you have one monolithic chip with control, interface, encryption logic, etc. all on one chip with one fab run. Many of our chip designs use small pockets of flash memory here and there (specially available f

      • rewriteability. You don't have to erase a block to change a single byte. It's more like RAM or hard disks in that respect.

        While I don't disagree with your point overall, isn't that exactly unlike a hard disk? On a hard disk, you must rewrite an entire sector to change a byte, and not only that, you must wait until the platter spins around to the right spot again in order to do it.

      • The story you're referencing is about advances in Single-Level Cell NAND NAND flash memory, which is the more expensive longer-wear-cycle cousin of the Multi-Level Cell flash memory that's been taking over the USB Flash market. MLC is more dense, and the price/gig has dropped by about 75% since last summer. Unfortunately, MLC technology only supports about 10,000 write cycles per cell (vs. about 100,000 for SLC), so you need to use wear-leveling drivers to keep it from wearing out, but that's still good e
  • If they found only one middle state, they could implement { true, false, file_not_found [thedailywtf.com] } enum
  • N/T
  • Oh look, another innovation that we will probably never see. I love new ideas - but instead of reporting on possible uses of technology science, how about actual science- stuff you will see this in 2 months reports. The "LOOK - SUPERBATTERIES WITH CARBON NANOTUBES - CHARGE IN SECONDS - 5 BAZILLION HOURS USE" kinds of articles are interesting, but not are they slashdotworthy. In Seinfeld's world, it definitely would not be sponge worthy.
  • Shall we call them 'amorlline' and 'crystphous'?

    • Re: (Score:3, Informative)

      by networkBoy (774728)
      PCM litho tech in not compatible with CPU litho tech.
      So I doubt this will be happening any time in the near future.
      -nB

    • Re:Two Billion Transistors on Their Latest Chip (Score:4, Informative)

      by SirLoadALot (991302) on Monday February 04 2008, @01:23PM (#22294492)
      No, processor caches are made out of SRAM, which is very fast but takes a lot of silicon space. Flash and phase-change RAM are totally different technologies. They aren't intended to be as fast as SRAM, but they are non-volatile, storing their contents without power. Phase-change or a similar technology may end up giving DRAM a run for its money if it gets fast enough, but I've never heard of an alternative to SRAM for internal processor caches.
    • The new memory is described as being as fast in reads as DRAM, which is an order or two of magnitude slower than register memory, which is what you need in a processor. Intel's adding of so much memory into the CPU interests me far more. Add enough, and the main memory is IN the processor (a-la the Transputer). From there, a little reorganizing reverses the arrangement - from memory in a CPU to a process in memory (PIM) architecture. Special-purpose PIM has been done before - Cray embedded a communications
      • It's not register memory that's taking up half the die space; it's SRAM cache. There's nothing all that interesting about it. It's old technology that's just gotten tiny enough to fit a ton of it on a CPU die.
        • Depends on who you are. Intel is notorious in being able to pack tons of SRAM into a small die space and this isn't just due to process shrinks. There are all sorts of flavors of SRAM, the standard taking 6 transistors but some taking as little as 1 depending on density/speed trade-offs.

    • Re:Only Double? (Score:5, Informative)

      by nonsequitor (893813) on Monday February 04 2008, @01:26PM (#22294556)
      It's double the number of bits. If you look at the largest value a 16 bit number or a 32 bit number can store, its not "double" in size. When it comes down to it, they're just bits, how you use them is up to you.

      • Re: (Score:2, Insightful)

        by deanlandolt (1004507)

        It's double the number of bits. If you look at the largest value a 16 bit number or a 32 bit number can store, its not "double" in size. When it comes down to it, they're just bits, how you use them is up to you.
        I suppose that's a lot like saying IPv6 "quadruples" the number of addresses of IPv4 -- you know, 128 bits versus 32 bits. I mean, they're just bits, right?
        • That's not what he said: he specifically said that it's "not 'double' in size". 32 bits *is* double the *number of bits* of 16 bits.

      • Re: (Score:2, Informative)

        by Dan Posluns (794424)
        Another way of looking at it:

        Say we use a bit to store the result of a coin toss. True for heads, false for tails.

        With two bits, we can store the results of two coin tosses. There are four possible outcomes when two coins are tossed, ranging from neither of them being heads, to only the first or the second being heads, or both of them being heads.

        If we double the number of bits, we can store the result of four coin tosses. There are now sixteen possible outcomes, but we're still only storing the result of f

    • Re:Only Double? (Score:4, Informative)

      by SirLoadALot (991302) on Monday February 04 2008, @01:28PM (#22294594)
      Currently the phase change RAM can only store 1 bit per element, in two possible states -- 0 or 1. They are changing this four possible states, which corresponds to two bits -- 00, 01, 10, 11. Hence, the amount of data that can be stored is doubled. The number of bits held per element increases by one every time the number of possible states is doubled.

    • Re:Only Double? (Score:5, Informative)

      by wizardforce (1005805) on Monday February 04 2008, @01:31PM (#22294638) Journal

      Now, if you are using binary storage, and each bit stores 1 of 2 values, then you have the possibilities of: 00,01,10,11. Which is 4 different values. Now if you have 4 states for each bit (which I guess wouldn't, by definition, be a bit anymore), then you have 00,01,02,03,10,11,12,13,20,21,22,23,30,31,32,33. So, you have have squared the amount of information you can store.
      simply put, no. 1 bit can either be 0 or 1 *not* 00, 01, 10, 11, that would be TWO bits. doubling the number of combos for one "bit" would likely be stored as 00,01,10,11 although what you're actually doing here is storing 0001 and 1011 together hence two "bits" of info can be stored per unit in a 4 state system.
      • I think I recall reading that a three-state system is best because it's the closest possible number to e (2.718), which was mathematically proven to be the most efficient state to process.

        Since there is no known way to represent e physically due to its fraction, "3" is as good as it gets.
      • I think what the GP meant was that you have more bits per symbol (or memory cell), then you would double the capacity.
        A similar thing is why the Baud rate [wikipedia.org] isn't equal to the bit rate on 56.7K modems.
        • No, it's twice the amount of data. If you can store 1 bit, you have:

          0, 1

          Two bits:

          00, 01, 10, 11

          In order to double it again, we simply add another bit:

          000, 001, 010, 011, 100, 101, 110, 111

          Each binary digit doubles the data capacity. Just like each decimal digit results in 10x the capacity.

          For any given number of bits per cell, n, we have 2^n combinations. This technology added one more bit, so it increased the amount of data storage by:

          2^(n+1) / 2^n = 2

          It only "doubled" because we started with n = 1. Th
        • But a bit is a contracion of binary digit, and binary means two. In communications, you have the same distinction between bits per second and baudrate. Bits per seconds give you the amount of ones and zeros transmitted, while the baudrate gives you the amount of symbols per digit. If a symbol can be one of four possible tones (or phases), you get 2 bits per baud, and that's what happening here again.

          • the baudrate gives you the amount of symbols per digit.
            Sorry for the mistake: that should be symbols per second, not per digit.

    • You are counting the number of VALUES that could be stored, not the equivalent number of bits needed to store those values. With your logic, starting with 8 slots that store 2 states, going to 8 slots that store 4 states, we'd be going from 256 to 65536. But that's not 256 times the capacity of bit storage; it's still just a doubling of the capacity. For every slot that can store 4 states, that's just the equivalent of 2 bits. That's certainly twice as much as before, and it means this technology of mem

    • Each item that could store two bits can now store twice as many bits of data. Before you had one item that had two states: (0, 1) Now you have one item that has four states (0, 1, 2, 3). With two states, you would need two items to make four possibilities: (00, 01, 10, 11). So you get the same amount of information with 1/2 the number of storage items, hence double the capacity. Using your example, you would need four bits to store the same possiblities as two items with four states: (0000, 0001, 0010

    • Re: (Score:3, Informative)

      by disassembled (977342)

      Will we now have computers that do base 4 arithemetic rather than base 2?
      A base-4 digit is the same as two base-2 digits, information-wise, so it doesn't really matter how the information is stored. If you want to store a byte in some piece of hardware, you can store it as 4 base-4 values or 8 base-2 values, and it'll all look the same to the CPU.
      • Not really. Logic circuits use binary voltage levels so the single voltage output from the pcm representing two bits has to be converted into two separate signals sent to the CPU. So it does matter how its stored, because you have to convert the base 4 to base 2 before sending it to the CPU. This isn't a challenge, but it does need to be done.
    • Those are good questions.

      I'm terribly out of date on this, but - in the old days with multiple chips for a memory bank, there was address decoding circuitry that would point you to the right chips/pins. If - and I'm clueless here - the same thing is still being done but it's all been reduced to fewer chips, then this _might_ imply that you still do the address decoding as before, but you have fewer address wires to route. In other words, we used to need two cells for four - now we have four in one cell -
      • Each cell, being just a variable resistor, will output a certain amount of voltage. Let's assume a 1.2V supply:

        00 = 0.0V
        01 = 0.4V
        10 = 0.8V
        11 = 1.2V

        You'd then do a crude A-D conversion (it's crude and small/lower power because it only needs to handle 4 distinct voltage levels). Off the top of my head, the smallest would be two pairs of N-FET and P-FET. One will have the threshold voltage biased at 0.2V and the other biased at 1.0V using a voltage divider from VDD to GND.

        The output of the first pair of FET
        • Actually, scratch that. You'd probably want to bias the voltages coming out of the cell. Two paths, one to bias it by Vt-0.2V and the other Vt-0.4V and have them go to separate FET pairs.

    • You can label the now total of 4 states however you like, such as 00/01/10/11 or 0/1/2/3 or A/B/C/D or T/A/C/O. But whatever they are, Intel would need to, at some point, convert this all back to 2 bits with states 0/1 when interfacing with external binary circuits. If they don't know how to do that they are welcome to "Ask Slashdot".

    • An interesting piece I ran across many years ago about ternary (and other bases -- try base-e!) systems, and how they _can_ be better at some things than binary.
      http://www.americanscientist.org/template/AssetDetail/assetid/14405?&print=yes [americanscientist.org]
      • Mathematically speaking, e is the most efficient base. Imagine a decode being a search tree and imagine the base being the number of children per node. A decode function is simply traversing the tree. The first level of the tree being the first bit, second level being the second bit, etc.

        To find any number N, one would need a tree of L levels and B children per node at each level. B^L must then be >= N in order to guarantee that the number N is represented in the tree. We can then take this to mean

        • T = S*B^(ln(N)/ln(B))

          To find the base that would result in the lowest search time, take the double derivative of T with respect to B and find the roots (peak and valley where search time either maximizes or minimizes).
          First, I believe you mean roots of the _first_ derivative (not second) to find extrema. Second, I don't believe the heuristic model, because S*B^(ln(N)/ln(B)) has no dependence on B!

    • Re: (Score:3, Interesting)

      by imgod2u (812837)
      You'd be surprised how much of your computer isn't "binary" per se. If you have a modem, I think the standard is a base-16 transmission code. Flash memory currently contains 2-bits-per-cell cells. Hell, the quad-pumped signal going from memory to processor (if you have a Core 2 or P4) isn't "binary" per se.
      • At one time they had analog computers, where the values were represented by voltages. I built a very, very primitive one in the 7th grade (actually it was more like an electric slide rule), but grownups made big complex ones.

        The advantage of analog computers was that they had no rounding errors. The disadvantage was noise.

    • And the alternative would be...?

    • Re: (Score:3, Informative)

      by imgod2u (812837)
      Not really. If you think about it, fundamentally there just as many states a bunch of atoms can arrange itself in as counting electrons. You're really only bound by Plank's constant in how resistive or conductive a collection of atoms are. Assuming you had a device sensitive enough to detect the variation.

      The trick is, of course, in how fast you can change those states. I would imagine electrons are much easier to move than whole atoms. I understand how read speed for PCM is faster than a transistor bu

      • Unless there are some sort of quarternary logic gates that I dont know about.

        Binary logic - yes, no.

        Quarternary logic - Yes, probably, possibly, no.

        Hmmm. 50 percent of the choices there are indeterminate; better stick with the trinary 'yes, maybe, no' model.

All trademarks and copyrights on this page are owned by their respective owners. Comments are owned by the Poster. The Rest © 1997-2009 Geeknet, Inc.