Slashdot Mirror

← Back to Stories (view on slashdot.org)

Samsung Develops Power-Sipping DDR4 Memory

Posted by timothy on from the continue-to-enjoy-the-treadmill dept.

Alex writes with this excerpt from TechSpot: "Samsung Electronics has announced that it completed development of the industry's first DDR4 DRAM module last month, using 30nm class process technology, and provided 1.2V 2GB DDR4 unbuffered dual in-line memory modules (UDIMM) to a controller maker for testing. The new DDR4 DRAM module can achieve data transfer rates of 2.133Gbps at 1.2V, compared to 1.35V and 1.5V DDR3 DRAM at an equivalent 30nm-class process technology, with speeds of up to 1.6Gbps. In a notebook, the DDR4 module reduces power consumption by 40 percent compared to a 1.5V DDR3 module. The module makes use of Pseudo Open Drain (POD) technology, which allows DDR4 DRAM to consume just half the electric current of DDR3 when reading and writing data."

36 of 152 comments (clear)

  1. How much power comparatively? by HockeyPuck · · Score: 2

    In a typical notebook, how much power does memory actually consume compared to other components (CPU, HD, screen, wireless transmitter etc..)?

    1. Re:How much power comparatively? by NoSig · · Score: 3, Informative

      Sleep mode leaves the ram powered but powers down most other things, is what I think he is saying. So ram may be the most significant power consumer for sleep mode.

    2. Re:How much power comparatively? by PatPending · · Score: 5, Funny

      Didn't you read the title of the summary? It doesn't "chew," it "sips."

      --
      What one fool can do, another can. (Ancient Simian Proverb)
    3. Re:How much power comparatively? by galvanash · · Score: 3, Insightful

      Hardly any. I remember skimming through a study of component power consumption and IIRC memory topped out at something like 5% total draw. So memory with half the power draw will buy you about 10 minutes. Whoopdeefuckingdoo.

      That is with the display turned on... Most portable devices spend a considerable amount of time with the display turned off to conserve power. To put this into perspective, on an HTC Desire android device with an AMOLED display the screen uses about 50%-60% of total power, memory is probably like you said around 5% (I have never seen hard numbers for the power draw only for memory, but 5% is probably close). If it is 5% with the display on, it would be around 15% or so when it is off, which is quite a bit more significant. Also, memory always uses power - even when it is not storing anything useful... Hence the more memory the device has the more power the RAM draws. Just saying, cutting RAM power use in half can be quite significant. It might be 10 minutes if you are using the device constantly, but it could well be an hour or more of extra standby time depending on how heavily you use the device.

      --
      - sigs are stupid
    4. Re:How much power comparatively? by phantomfive · · Score: 2

      It was completely helpful. When in sleep mode, the RAM is powered up, but most everything else is turned off. Thus you can develop an intuition of how much power RAM takes compared to the other components.

      I can't say for sure the power draw from all the components, especially since they vary, but this processor draws 600 watts. Now that's what's known as an unhelpful response.

      --
      "First they came for the slanderers and i said nothing."
    5. Re:How much power comparatively? by wizardforce · · Score: 2

      power consumption = heat that needs to be removed. Heat becomes a bigger problem the smaller the components are. Reduce the amount of heat produced and you've just made it easier to produce even smaller components.

      --
      Sigs are too short to say anything truly profound so read the above post instead.
    6. Re:How much power comparatively? by hairyfeet · · Score: 2

      Which is why I think the next big breakthrough (which will make someone Bill Gates rich) will be the ability to selectively turn memory cells off just as we can turns parts of the CPU/GPU off in AMD and Intel chips (from what I understand Nvidia is pretty much "all or nothing" except on Tegra). what one would have to develop is a "smart controller" most likely on the RAM module itself, one that knew which cells were in use and when given the "we are in low power mode" signal by the OS would have the ability to electrically isolate the running cells and power down the non working set.

      IMHO it is that which will make the next big leap, not all this DDR slight decreases which IMHO just serve to keep the price of RAM raised. I mean we are just now getting to where DDR 3 is affordable! Besides unfortunately all the mobile devices try to rip off Apple with their iSliver batteries so the public have been pretty well trained to keep a charger handy. Even if you are talking a 20% gain with these micro ultra thin batteries that really won't be much. But the ones that figure out how to selectively turn off cells, they will be the ones to make incredible amounts of money especially if they patent the hell out of it. After all it will be able to have an assload of RAM, yet use almost nothing when sleeping. Who wouldn't want that?

      --
      ACs don't waste your time replying, your posts are never seen by me.
    7. Re:How much power comparatively? by drinkypoo · · Score: 2

      How about this: you can double the amount of RAM for the same power budget. Batteries are not getting better as fast as we would like to use more RAM.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    8. Re:How much power comparatively? by TheRaven64 · · Score: 2

      First you'd have to move everything to the DIMMs that you're keeping on. This means that all the pointers would change, so you'd have to have a way to keep track of that.

      Fortunately, if you come from some time after the 1980s, this is done already. Nothing except the kernel sees physical memory addresses, they see virtual memory addresses. These are mapped to the physical address by the MMU / page tables, and often do change over the apps lifetime (e.g. when a page is swapped out then in, it is not always returned to the same physical page).

      --
      I am TheRaven on Soylent News
    9. Re:How much power comparatively? by Wain13001 · · Score: 2

      It tells you nothing about the ONE UNCHANGED FUCKING VARIABLE...

      Not very good at the pre-algebra are we?

      Here's a hint, x+y

  2. Good news by del_diablo · · Score: 5, Interesting

    Now, lets pair this with a ARM core and hope we get a reasonable hack that allows a wireless that does not eat power like the current ones.....
    Then lets enjoy our ARM-puter: Portable, powerful, and battery for more than a day of use.

    1. Re:Good news by Khyber · · Score: 2

      Not happening. Higher-frequency transmissions need more power to go further. Lower frequencies don't carry as much data, so there's a huge trade-off in play.

      --
      Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
    2. Re:Good news by LordLucless · · Score: 2

      Or just connect smarter - my n900 connects to wireless networks on demand and auto disconnects when not in use. I can get 1.5 - 2 days of light usage off its battery.

      --
      Just because you're paranoid doesn't mean there isn't an invisible demon about to eat your face
    3. Re:Good news by Khyber · · Score: 2

      Lower frequencies don't carry as much data, as I stated before.

      Also, bear in mind that for each extra device connected to the wireless access point, that's less available bandwidth overall. That 54mpbs rating for 802.11g is a pooled rating, meaning every time you add a new client onto the AP you lose some bandwidth - not every single person gets 54mbps throughput unless you're the only person using that access point/that channel.

      Also, most the lower frequencies are already allocated for certain services by the FCC, and thus you can't put internet on those wavelengths without risk of interrupting those services. Going into any of the areas unlicensed would likely be a big headache as there's very little bandwidth in those ranges anyways.

      --
      Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
  3. "Power Sipping" by Aboroth · · Score: 5, Insightful

    Does anyone else besides me hate that term?

    1. Re:"Power Sipping" by nyctopterus · · Score: 2

      In what context do you object to "ecosystem"? What word would you prefer we use for the system of biological interdependency?

  4. Pseudo Open Drain (POD) technology by PatPending · · Score: 3, Funny


    PatPending (talking to friend on phone during a bash help session): It's called Pseudo Open Drain (POD) technology
    Friend: Okay, I'll try that...
    Friend(typing): sudo open drain
    Friend: Argh! I hate this command line bullshit!

    --
    What one fool can do, another can. (Ancient Simian Proverb)
    1. Re:Pseudo Open Drain (POD) technology by sumdumass · · Score: 2

      It didn't work because it's Pseudo science./

  5. Meh by lennier1 · · Score: 4, Insightful

    I'd rather have them finally mass-produce 8 and 16 GB modules for the desktop market.

    1. Re:Meh by war4peace · · Score: 2

      Sir, you have my vote.
      Actually, less swap means less HDD churning, so the power consumption might be the same even with addition of more RAM.

      --
      ...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
    2. Re:Meh by Kjella · · Score: 3, Insightful

      what "Desktop" user needs that much, or even runs an operating system that can address it?

      Need we can discuss, but the price difference between Win7 32 and 64 bit versions is ~0 and I've not heard anyone complain about 64 bit drivers anymore. Mac I think is the same and Linux has of course supported 64 bit forever. Unless you're talking about an Atom that's not 64 bit capable, there's no particular reason not to get an OS capable. That is unless you still want to wipe a new box and install XP...

      --
      Live today, because you never know what tomorrow brings
    3. Re:Meh by chichilalescu · · Score: 2

      kjella was kind enough to discuss the operating system.
      now, regarding the use: games, desktop effects, working with extremely large files (highschoolers editing movies), ridiculously large images, and so on. I won't mention actual professions.
      Anyway, the first hint is "Desktop" user. someone who only wants instant messaging and facebook will be satisfied with a laptop.

      --
      new sig
    4. Re:Meh by thegarbz · · Score: 2

      As someone who regularly edits images larger than 100 megapixels with multiple layers in 16bbp let me just say, why the heck would you want that much on your desktop?

    5. Re:Meh by lennier1 · · Score: 2

      what "Desktop" user needs that much, or even runs an operating system that can address it?

      Are you serious? The specs for Windows 7 list 192 GB as within its capabilities and even RHEL v6 had a theoretical limit of 1 TB. But it will be a while until desktops will have to deal with the 256 TB an amd64 CPU's 48-bit address space would allow in theory.

      I'm not talking about the usual game kiddies who only want to impress others with system stats which they can't fully utilize anyway. Even if some people don't like to see them grouped with the little toy in a teenager's bedroom, regular professional workstations are still using desktop components.
      Areas like professional 3D work eat RAM and CPU cycles for breakfast. And there also are professions like developers and administrators who need to run several virtual machines in the background throughout the day. There certainly is demand

    6. Re:Meh by Fri13 · · Score: 2

      Yes, demand is there, even it is small when compared to typical office/game use.
      Photographers, artists, 3D modeleres (hobbiest) etc. And of course gamers would already want 8-16Gb of RAM as it just makes everything so much nicer when you do not need to care about RAM use.

      Try opening a 50 12Mpix RAW photos open at once and edit them in photo manipulation program...

      So where are our cheap 8-16Gb blocks?

    7. Re:Meh by hcdejong · · Score: 2

      3D CAD. My colleagues regularly run into RAM limits with 4 GB.

    8. Re:Meh by tygerstripes · · Score: 2

      I'm not judging here, just wondering: have you considered having more than one box? If money is no object then fair enough, but it sounds like you're shelling out a lot for top-notch hardware to do lots of mid-level tasks, when you distribute the work on a KVM setup. You'd save a bundle in hardware, reduce your VM overheads, and introduce some healthy redundancy for when that very expensive rig does something smokey and difficult to diagnose.

      --
      Meta will eat itself
    9. Re:Meh by Corporate+Troll · · Score: 2

      Unless you're talking about an Atom that's not 64 bit capable

      So, I'm just imagining running a 330 and D510 with a 64-bit operating system? Now, of course, you can't find Atom motherboards that support more than 4GB RAM and the two I run have only 4GB RAM and some of it is "stolen" by the graphics card.

      From what I understood, Atoms can't address more than 4GB though, but running 64-bit instructions is no problem.

      Apart from that: yes, modern Atom CPUs do run 64-bit operating systems.

    10. Re:Meh by Corporate+Troll · · Score: 2

      ...and I'm wrong... The Exxx series seem to be non-64 bit. haven't seen those anywhere yet, but since they have been released in Fall 2010, they qualify as "new".

    11. Re:Meh by petermgreen · · Score: 2

      On 32-bit linux these days each process can have up to 3GB of mappings (with the top 4GB of virtual address space being used for the kernel, there were some patches to allow 4GB for each process but they slowed down context switches and afaict were never widely adopted)

      Physical memory wise 32-bit linux supports PAE and afaict all intel and amd chips that support x64 also support x86 with PAE. Most 32-bit linux distros offer a PAE kernel though not all select it by default (debian calls it linux-image-2.6-686-bigmem ).

      64-bit kernels can also run 32-bit apps and allow them to have a full 4GB of virtual address space.

      --
      note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
    12. Re:Meh by petermgreen · · Score: 2

      with the top 4GB of virtual address space
      That should have said the top 1GB of virtual address space.

      --
      note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
    13. Re:Meh by level_headed_midwest · · Score: 2

      I don't think ECC is a mandatory part of the DDR4 spec, but the module shown in the picture in TFA is an ECC module.

      --
      Just "gittin-r-done," day after day.
    14. Re:Meh by rednip · · Score: 2

      As I'm hungry, I'd rather have a ham sandwich rather than mass produced 8 and 16 GB modules. -- is that insightful too?

      --
      The force that blew the Big Bang continues to accelerate.
  6. yeah, yeah, thats all fine, but by Blackout+for+Hungary · · Score: 2

    What about latency?

  7. POD explained by overshoot · · Score: 5, Informative
    In a classical open-drain connection, the active device pulls down and the bus termination pulls up. For a pure transmission line, this works just fine -- the current wave from the turn-off of the driver is effectively identical to the current wave from the turn on. In practice, open-drain uses more static current than a push-pull driver against a center termination and since the line isn't a pure transmission line (lumped capacitances, stubs) the rising edge is slower than the falling edge.

    POD addresses this by actively pulling up at the beginning of a rising edge, then releasing the pullup to avoid a bus contention later. This reduces the termination current (at some cost in impedance mismatch, but it's already a sloppy line) and improved switching symmetry.

    --
    Lacking <sarcasm> tags, /. substitutes moderation as "Troll."
  8. Nope by overshoot · · Score: 3, Informative

    One thing not mentioned in the article or summary is whether or not this technology reduces standby power consumption in DRAM.

    POD by itself doesn't reduce power consumption in standby, since both POD and SSTL turn off the bus drivers then. The older POD technologies from the GDDR families use Thevenin termination, though, so the terminators draw a lot of unnecessary current when they're enabled (as distinct from the result with a dedicated termination supply.)

    If you really want to know how this all works, JEDEC has the DDR4 standard available for free download. Follow the "free standards" link.

    --
    Lacking <sarcasm> tags, /. substitutes moderation as "Troll."