Slashdot stories can be listened to in audio form via an RSS feed, as read by our own robotic overlord.

 



Forgot your password?
typodupeerror
Data Storage Media Hardware Science

Purdue Researchers Demonstrate Low-Power, Fast FeTRAM Memory 50

Posted by timothy
from the there-is-no-university-of-purdue dept.
eldavojohn writes "Researchers at Purdue University's Birck Nanotechnology Center have released news of a proof of concept new ferroelectric transistor random access memory or 'FeTRAM.' This new technology is nonvolatile and the researchers claim it could use up to 99% less energy than current flash memory. Unlike most FeRAM technology that uses a capacitor, FeTRAM provides nondestructive readout by storing information using a ferroelectric transistor instead. From the article: 'The new technology also is compatible with industry manufacturing processes for complementary metal oxide semiconductors, or CMOS, used to produce computer chips. It has the potential to replace conventional memory systems.' So if they get this into production, you might not have to worry about your laptop cooking your genitals. They've been published in ACS (paywalled) and the professor leading the research has many patents filed relating to transistor nanotechnology."
This discussion has been archived. No new comments can be posted.

Purdue Researchers Demonstrate Low-Power, Fast FeTRAM Memory

Comments Filter:
  • by mrflash818 (226638) on Tuesday September 27, 2011 @10:02AM (#37526660) Homepage Journal

    ...I am drawn to it ;)

  • by mrflash818 (226638) on Tuesday September 27, 2011 @10:03AM (#37526676) Homepage Journal

    ...that is an attractive solution ;)

  • by jandrese (485) <kensama@vt.edu> on Tuesday September 27, 2011 @10:09AM (#37526722) Homepage Journal
    The summary makes it sound like super miniaturized core memory. I'm sure it's more complex than that, but it's still pretty cool.
  • by iggymanz (596061) on Tuesday September 27, 2011 @10:09AM (#37526726)

    the processor is the main source of gonad grilling heat, RAM is about 103% of power consumption. the one in my laptop pulls 31W when it is busy

    • by iggymanz (596061) on Tuesday September 27, 2011 @10:10AM (#37526744)

      hah, fix that text edit box slashcode monkeys, wrote ram is one to three percent of power consumption but look what came out

      • by Anonymous Coward on Tuesday September 27, 2011 @10:18AM (#37526860)

        It's there. It says "Preview".

      • by Reverand Dave (1959652) on Tuesday September 27, 2011 @10:24AM (#37526924)
        This is looking to be more of a replacement for flash and other solid state memory so it would not be the RAM in your PC that's getting replaced, it would be the HDD which does consume a good amount of power. A more efficient system for accessing the data on your drive would in turn lower CPU power usage which would in turn reduce overall power consumption of a standard computer or laptop. This would also have implications to reduce the drain in things like MP3 players and cellphones thus prolonging the battery life in those devices as well.
        • by WaffleMonster (969671) on Tuesday September 27, 2011 @11:39AM (#37527862)

          This is looking to be more of a replacement for flash and other solid state memory so it would not be the RAM in your PC that's getting replaced, it would be the HDD which does consume a good amount of power.

          From TFA:
          "They might also be much faster than another form of computer memory called SRAM"

          SRAM is the same stuff used in CPU memory caches.

          • by Anne Thwacks (531696) on Tuesday September 27, 2011 @12:12PM (#37528270)
            might also be much faster

            Looks like they want some funding. You may recall the expression "pigs might fly".

          • by Reverand Dave (1959652) on Tuesday September 27, 2011 @12:32PM (#37528604)
            SRAM is used in such small densities that its energy consumption is negligible. SRAM is indeed hugely inefficient and would likely be replaced with this type of memory should it become feasible, but I think GP was and I know I was speaking to DRAM which is what the majority of people speak of when they mention the RAM in their system. When was the last time you updated your SRAM?
            • by WaffleMonster (969671) on Wednesday September 28, 2011 @01:41PM (#37543440)

              SRAM is used in such small densities that its energy consumption is negligible. SRAM is indeed hugely inefficient and would likely be replaced with this type of memory should it become feasible, but I think GP was and I know I was speaking to DRAM which is what the majority of people speak of when they mention the RAM in their system. When was the last time you updated your SRAM?

              I was assuming the reader would understand my point.

              If it lives up to the hype:
              Faster than SRAM.
              Uses the same area of silicon as flash
              More power effecient
              Practically infinite cycle life
              Non volitile

              What on earth prevents it from replacing everything? SRAM, DRAM, disk drives?

              • by Reverand Dave (1959652) on Thursday September 29, 2011 @09:19AM (#37553508)

                What on earth prevents it from replacing everything? SRAM, DRAM, disk drives?

                Having worked in the semiconductor industry for over 10 years I can't dispute that this would be an acceptable replacement for all forms of memory inside of your system. However, I've seen several technologies (MRAM, FeRAM, RAMFlash) come and go that were touted as "the ultimate memory and storage technology" (URAM) that although were feasible, completely fizzled due to market and manufacturing concerns. DRAM is a mostly low profit industry where SRAM is a very high profit industry. Flash is a pretty stable market now with good margins, where DRAM is a stable market with horrible margins. The most likely place to target replacements is the low and medium profit margin industries and leave the high profit ones where they are. Even most NOR Flash has been dumped in favor of NAND Flash due to more stable profit margins in most markets even though some applications would work fine with NOR as opposed to NAND.

                That is what on earth prevents if from replacing everything. Not that I disagree that it shouldn't, I'm just explaining.

        • by Mr Z (6791) on Tuesday September 27, 2011 @12:39PM (#37528708) Homepage Journal

          Don't confuse "power" with "energy". Power is energy consumed divided by the amount of time required to consume it. Power is in watts (joules/second), whereas energy is in straight joules. (Or other units that cancel out the time factor, such as kWh.)

          For example, I've noticed since I installed an SSD in my laptop that my automated backups run way more efficiently. Previously, with an HDD, the whole machine would slow to a crawl, and the computer spent most of its time waiting for disk seeks. Now, the computer keeps chugging along and the backup completes quickly and quietly in the background.

          This probably consumes less overall energy, thereby prolonging battery life. However, because the backups run so fast, the CPU now gets very, very warm during backups (70+ Celsius), because it's actually consuming more power during the backup. Before the SSD, the CPU would barely heat at all, since it spent all that time idling waiting for the HDD.

          The net result? My laptop is more likely to burn my lap now than before, at least during backups, even though it is probably using less overall energy and providing a greater overall battery life. So, if this FeTRAM is even faster than the flash in my SSD, it could make the CPU's peak power consumption even worse, despite the whole system being more energy efficient overall.

          • by WorBlux (1751716) on Tuesday September 27, 2011 @03:01PM (#37530890)
            Running the CPU at full speed and then dropping into idle is more efficient than running at half speed. Especially since intel processors are getting to the point you can just turn off the cores you don't need.
            • by Mr Z (6791) on Tuesday September 27, 2011 @03:17PM (#37531186) Homepage Journal

              It is more energy efficient, especially if you can fully power off in idle and avoid wasting energy on transistor leakage.

              I'm just saying that the short term power consumption (measured in joules/second, aka. Watts), will be higher even if the total energy usage (total joules) is lower. If that "short term" is measured in 10s of seconds, then things still get hot enough to roast your chestnuts, so to speak.

              Some math: Suppose the CPU has to do a total of N joules of work(say, computing checksums and compressing files) to complete the backup. It'll either do it in a bunch of short bursts as data arrives, or it could do it in one solid chunk if the data arrives fast enough. With a slow hard drive, suppose that backup takes 1000 seconds, and that with my shiny new SSD, it takes only 100 seconds. The CPU power consumption for the backup is 10x for the SSD (N/100 versus N/1000), but for 1/10th as long.

              The extra CPU efficiency comes from is from saving leakage. Suppose the CPU leaks M joules per second when on, and nothing when off. You'll save M*900 joules CPU leakage using the SSD. Still, during the backup, the CPU will dissipate a nut-baking (N/100 + M) Watts for the SSD backup, versus the comparatively cooler (N/1000 + M) Watts for the HDD backup.

              (Note, I'm not including the energy used by the SSD or the HDD. I'm only considering the CPU here.)

  • Confusing muddle (Score:4, Insightful)

    by jmorris42 (1458) * <jmorris@b[ ].org ['eau' in gap]> on Tuesday September 27, 2011 @10:39AM (#37527076)

    Just the summary has me confused. Is this a RAM replacement or a FLASH replacement tech? Or both?

    I guess between green getting grant money and industry looking to lower their power bills pushing energy savings is mandatory, but I never though flash/ssd was a power hog anyway, so if it is a flash/hdd replacement that wouldn't be all that important. Now ram modules, especially high performance 'gamer' memory has heat sinks and gets plenty hot enough to matter. Especially in heavily loaded servers hosting a lot of virtuals, those puppies get loaded up on ram so I suspect would account for a fair chunk of the total power budget in a rack full.

    But I wouldn't throw venture capital at em just yet. Every few months it seems we see a story about a new memory tech. Some of them, MRAM for example, do eventually surface but they can't scale up enough to compete with conventional memory so have to settle for a niche where their special properties make them viable. Again, look at MRAM. You can buy the stuff and it really works. It is sold as a drop in replacement for old EEPROM and SRAM chips in the old DIP packages. Not only low power operation, it retains memory with the power off and no need for a backup battery. But a few Mbits per chip seems to be the current limit so it isn't a threat to either flash or dram unless it can scale up a thousand fold. Kinda like the old magnetic bubble memory that was always a few years away from making hard drives obsolete back in the '80s.... until hard drive capacity per dollar grew so much faster than bubble memory could hope to catch up to and R&D died out on it, leaving it but a footnote in tech history.

    • by Anonymous Coward on Tuesday September 27, 2011 @11:41AM (#37527876)

      If you ask the Prof., it's the all singing, all dancing, (future) replacement for everything. And it'll do the dishes, and get you a date. However, reality: quoting from TFA: "However, our present device consumes more power because it is still not properly scaled," Das said. "For future generations of FeTRAM technologies one of the main objectives will be to reduce the power dissipation." So even the headline grabber "far less power" is "future devices". Yawn.

    • by Anonymous Coward on Tuesday September 27, 2011 @12:05PM (#37528204)

      RAM does not use a lot of power[1]. RAM chips usually get hot because they receive no cooling except for the already heated air used by the CPU's heatsink and fan (if in a PC).

      Don't get me wrong, using less power for a task is always a nice improvement, but it won't change your laptop from scorching your nuts. The CPU would be the prime candidate for improvement there.

      [1]This datasheet for a Kingston product (http://www.valueram.com/datasheets/KHX9200D2_1G.pdf) say it uses a smidgen over 2W.

      • by jmorris42 (1458) * <jmorris@b[ ].org ['eau' in gap]> on Tuesday September 27, 2011 @01:24PM (#37529406)

        Ok, I hadn't bothered to dig up real numbers. Thanks. So take a server loaded to the gills with RAM and it could easily draw more power than the CPU, especially if they are low power ones like the AMD HE parts the ram on each cpu could outdraw the processor. For that mater, if you only put 64GB in at 2W per G that is 128W, more than any CPU that has any business going into a 1U or even most 2U rack mount cases. The newest ram modules might draw a little less per GB from die shrinking but since DDR3 clocks faster it probably ends up in the same ballpark. All the attention has went into giving processors all sorts of power saving features, looks like attention needs to turn to memory now.

        • by Anonymous Coward on Tuesday September 27, 2011 @03:58PM (#37531810)

          So take a server loaded to the gills with RAM and it could easily draw more power than the CPU, especially if they are low power ones like the AMD HE parts the ram on each cpu could outdraw the processor. For that mater, if you only put 64GB in at 2W per G that is 128W, more than any CPU that has any business going into a 1U or even most 2U rack mount cases. The newest ram modules might draw a little less per GB from die shrinking but since DDR3 clocks faster it probably ends up in the same ballpark. All the attention has went into giving processors all sorts of power saving features, looks like attention needs to turn to memory now.

          Attention has been paid to memory power use for a very long time now, you just weren't aware of it. This is something I see on slashdot a lot; if it's not in their face people assume nothing's being done.

          The main problem at the moment is one of those fundamental physics things: it takes lots of energy to move bits at a high speed across a significant physical distance. At the moment, I believe offchip IO uses significantly more energy in DDR3 DRAM than what's needed to access and refresh DRAM contents. Moving to a different storage technology would not help very much, since DRAM bit cells are reasonably low power as it is.

          This is one reason why the IO voltage for SDRAM keeps going down. Single data rate SDRAM used 3.3V IO, DDR 2.5V, DDR2 1.8V, and DDR3 1.5V, with a low-power variant of DDR3 at 1.35V on the way, and DDR4 likely somewhere below 1.3V. Power (including IO power) scales with the square of the voltage, so there have been substantial reductions in power from each SDRAM generation just based on supply voltage. And that's not the only thing, there have been many other more esoteric power-oriented features introduced over time.

  • by Anonymous Coward on Tuesday September 27, 2011 @10:52AM (#37527252)

    Who has the patents and was the research publicly funded?

    If it is publicly funded with government grants then how proper is it for the research to be patented?

  • by Anonymous Coward on Tuesday September 27, 2011 @11:11AM (#37527518)

    Is Perdue paying licensing fees to MIT's PR department for the syntax of the subject line?

  • by eyenot (102141) <eyenot@hotmail.com> on Tuesday September 27, 2011 @11:44AM (#37527914) Homepage

    This http://en.wikipedia.org/wiki/Memristor/ [wikipedia.org] is more interesting. But, which one are we going to see, first? I'd prefer to see HP's memristor dominate the memory, storage, and processor markets. But are we going to have to wait through some stage of planned obsolescence, first, while all these minor variations on existing components arrive and all companies seek to maximize profits on those before moving onto something that makes them all obsolete?

  • by hackertourist (2202674) <hackertouristNO@SPAMxmsnet.nl> on Tuesday September 27, 2011 @12:36PM (#37528674)

    I looked into putting an SSD into my laptop, but the stories of short life (and getting shorter with each reduction in process size) are putting me off. Would this FeTRAM be more resilient?

  • by Anonymous Coward on Tuesday September 27, 2011 @12:40PM (#37528724)

    There have been so many attempts at the perfect replacement for flash memory. I remember when SiGe and GaAs (for high speed) was going to be the next big thing and bubble memories for non volitaile, FRAM's (where are they now?), and the list is on and on and on. I've not RTFA but I'll believe it when I see it. If it needs any kind of special process, special equipment, chemistry outside the realm of standard CMOS or comes even close to breaking the traditional installed CMOS infrastructure you won't see it happen. No major semi is going to take the risk. Big Fabs these days are running at 5B a pop or more. No one is going to risk that.

    • by blair1q (305137) on Tuesday September 27, 2011 @01:27PM (#37529446) Journal

      Flash takes those diversions from ordinary CMOS.

      The only thing that matters is profit margin.

      If this produces more on the bottom line, Flash will be the next Floppy.

    • by Anonymous Coward on Tuesday September 27, 2011 @04:09PM (#37531956)

      If it needs any kind of special process, special equipment, chemistry outside the realm of standard CMOS or comes even close to breaking the traditional installed CMOS infrastructure you won't see it happen. No major semi is going to take the risk. Big Fabs these days are running at 5B a pop or more. No one is going to risk that.

      Flash isn't really standard CMOS any more. It's like DRAM - the major flash mfrs all have fabs and process nodes dedicated to flash and nothing but flash.

      Which doesn't negate your argument, of course. Strengthens it, just from a slightly different direction. However, the known likely limits of flash scaling (it's going to run out of steam in 1 or 2 more process nodes) have led to the big flash mfrs investing in many of these possible flash alternatives, because whoever picks the winning horse (assuming at least one of them can win) is going to have a major advantage sometime in the not-so-distant future.

  • by Anonymous Coward on Tuesday September 27, 2011 @05:09PM (#37532820)

    my ssd uses 4w peak
    this will apparently use 40mw peak
    somehow seems a little off

    also no details about speed either
    it'll be useless if it's as slow as a floppy disk

  • by Anonymous Coward on Wednesday September 28, 2011 @02:49PM (#37544630)

    Ship it.

Mediocrity finds safety in standardization. -- Frederick Crane

Working...