digg

Join DiggAbout

Discover and share the best of the web!
Learn more about Digg by taking the tour.

See the original image at tgdaily.com

How a forgotten Intel invention could revolutionize the CPU

tgdaily.com — The next stage for Intel processors? - Integrated DRAM.

Bury
show profanity settings
expand all
  • dougvfr750, on 06/19/2008, -3/+15It doesn't seem like much, but this would be an amazing innovation in overall system speed
    • cawpin, on 06/19/2008, -10/+3No, if it was an innovation it would be new. It isn't.
      • Charlotte_Web, on 06/19/2008, -0/+12FTA:

        "A small research team inside Intel succeeded in reducing the size of DRAM cells to only two transistors and completely removing the capacitors. Conceivably, these two achievements could change the way how we will use DRAM in the future: For example, expensive and complex SRAM (static RAM) cells could be entirely removed from a CPU and replaced with DRAM."

        Sounds new to me.
    • sk11, on 06/19/2008, -1/+4Cache memory has been there for years now, this is just a step further. A real leap would be integrating a dedicated graphics core onto the system cpu. Weren't amd and ati discussing this very thing?
      • mattbatt77, on 06/19/2008, -0/+5I thought Nahalem was suppose to integrate some graphics on the CPU core and of course Larrabee, though starting out as Intel's attempt to create a dedicated GPU to outperform NVIDIA and ATI will eventually also have components on CPU's as well. This is what I have gathered so far at least . . .
  • MarkusX, on 06/19/2008, -6/+3810x faster than nowerdays CPUs? Sweet!
    • Shakermaker, on 06/19/2008, -6/+4Nowerdays?
    • cnosal, on 06/20/2008, -0/+3Note that's 10x improvement in cache performance, not core clock speed.
  • 44Bigs, on 06/19/2008, -3/+14Old news.. Bitboys Oy 'had' this in their 'GPU's' ten years ago!
    • strictnein, on 06/19/2008, -0/+8I heard Infinium Labs is going to use it in their upcoming console.
  • devin_mm, on 06/19/2008, -7/+68Dugg because it was something actually interesting, not the usual BS on digg.
  • Midnightrambler, on 06/19/2008, -10/+162Anyone else think that pic was hardwood flooring?
    • brownspank, on 06/19/2008, -1/+11More like glazed tiles, actually. And the reflections look like church windows.
    • synwolf, on 06/19/2008, -0/+10I'd buy it.
    • Wargasmic, on 06/19/2008, -1/+8The next innovation at Intel! It's the 1000 core cpu that you tile on your floor.
    • MadOgre, on 06/19/2008, -0/+4I though it looked just like my kitchen...
  • Lukesed, on 06/19/2008, -5/+125*Whew*

    I misread the description as "The next stage for Intel processors? - Integrated DRM."
    • eengineer, on 06/19/2008, -0/+12Shh.. Don't give them ideas. I can see the RIAA/MPAA approved media center pc featuring the Intel CDS* chip with integrated DRM now.


      *CDS = Can't Do *****.
  • dblespresso, on 06/19/2008, -0/+20I suspect they didnt promote it because it may not be manufacturable. There could be expensive additional mask steps, thermal disipation, or yield issues. Other companies include Dram through a stacked die which allows the dram die to be manufactured and tested in an optimal setting for memories.
    • Mohonri, on 06/19/2008, -0/+13I think you may have hit the nail on the head. The article notes that the DRAM concerned requires more power than the SRAM that is currently in common use. Considering the fact that power dissipation is one of the limiting factors in CPU development, adding a bunch of power-hungry DRAM may be impractical (though still very cool).
      • sexybobo, on 06/19/2008, -0/+4The article also said the 45nm will make it easier to do. It might have been a head of its time as in the idea was great but they didn't have the technology to do it but there have been a lot of advances in technology since it was first conceptualized
        • phoenix3200, on 06/19/2008, -0/+4SRAM is far less efficient as the size is reduced. It's likely that, when comparing the two, DRAM is beginning to look like a better idea - not that the process itself makes it any easier.

          The masking problem isn't present - the purpose of the booth was to show that DRAM could be integrated without additional masks.
    • HonoredMule, on 06/19/2008, -1/+1There's also the question of how RAM manufacturers would respond to the threat of having their market undercut if this received a lot of media attention before even being turned into a real product.
  • dafragsta, on 06/19/2008, -1/+12In 1999(ish), someone decided it would be an awesome to put full speed L2 cache on die instead of on a riser card or on a PCB attached directly to the CPU. It took ~10 years to think that putting the memory on the die might be an equally good idea. C'mon guys. Even TGI Fridays realized you could put Jack Daniels sauce on anything and make it better pretty early on. Getting as much processing and memory power on the same die would DEFINITELY remove a lot of bandwidth bottle necks. So what if the dies will be an inch square.

    Ok, that might be oversimplifying things a BIT, but this had to be obvious.
    • migshark, on 06/19/2008, -0/+3Scaling played its part too, the size of the cache necessary is dependant upon the processors other capabilities.
    • mizarone, on 06/19/2008, -0/+13Actually by 1999 the Coppermine version of the Pentium III had full speed L2 cache. However, it did take about 10 years to put that cache on the die is because to took ten years to make the transistors small enough to fir everything on the die. It looks like that DRAM on the die will be for replacing some of the SRAM used to the cache (L3, L2 levels), not shunting some of the system memory onto the die. I wouldn't want that much memory on the die anyways, that will produce tremendous amounts of heat.
  • migshark, on 06/19/2008, -6/+2Depending on the additional costs necessary, this must really annoy those who currently have supercomputers under construction.
  • Cheeseburgers, on 06/19/2008, -0/+33where are the quantum computers?
    • RealmDown, on 06/19/2008, -0/+42They are here sometimes.
    • Asrrin29, on 06/19/2008, -0/+13I'm uncertain.
    • burnedtubes, on 06/19/2008, -0/+5No one has seen them yet to force them to make up their minds.....
    • RussellDovey, on 06/19/2008, -1/+6They're over there... No wait, they're over here too... Hmmm.
    • Suricou, on 06/19/2008, -1/+6Comfortable in their pools of liquid nitrogen.
  • jackal42, on 06/19/2008, -14/+4Can i get dram in my G-spot?
  • vector66, on 06/19/2008, -0/+54Thanks, it's good to see this type of articles again on Digg!
  • tripledjr, on 06/19/2008, -1/+9*HEAR
    • dgendreau, on 06/19/2008, -0/+3Yeah...
      "we rarely here about the way how a processor actually communicates"
      Was this written by a 9 year old?
      • secrity, on 06/19/2008, -0/+9People rely on spell checkers to do there proofreading four them.
    • Jimbo51, on 06/19/2008, -3/+1Blatant typo
    • linuxpenguin, on 06/19/2008, -0/+2You would think websites like this could afford a spelling and grammar checker. It looks like they're using Joomla! which has modules that can check your spelling, but you'd think they'd check their grammar too - if not with a computer program, at least with a quick manual look-through before publishing.
  • pinoylife, on 06/19/2008, -10/+5Does this mean I can run Crysis on High settings?
  • supermanly, on 06/19/2008, -0/+43C'mon AMD, stay competitive, I don't want a monopoly here.
    • Arkz, on 06/19/2008, -0/+3Intel STILL haven't integrated the memory controller, something AMD did years ago! if the Phenom wasn't weak next to the high end C2Q's we would still all be in love with them
      • JnOrris, on 06/19/2008, -1/+3When you compare price to performance the new Phantom processors compete very favorably with Intel, hell I just bought a X4 9600+ for $120 on ebay, the equivalent intel quad cpu is $100+ more. Sure the quad intel processors are a bit better, but for the price I'll take the cheaper AMD processor (and motherboards).
        • ArgonTime, on 06/19/2008, -0/+1It isn't equivalent if it's better...
        • Rizmaster, on 06/20/2008, -1/+2Clearly Orris doesn't know the first thing about the 9600+. It's weaker than a ***** Core2Duo.
    • Enron, on 06/19/2008, -7/+2I will never buy another AMD product. They are horrible, made in China garbage. AMD = Asians Making Dollars.
      • CHAD3814, on 06/20/2008, -0/+1Wow, that's about the most racist thing I've read all day.... And all I do is read digg.
    • Arkz, on 06/20/2008, -0/+3You mad? AMD are great! both companies are! i still have my K6-2 and Pentium MMX and they both work fine at their blazing 400MHz and 233MHz!
  • stefano1234, on 06/19/2008, -1/+0Definitely it'll make our lives more easier, more happier. Nowadays, real invention counts!
  • NRay7882, on 06/19/2008, -1/+1New Intel chipset: woodblock. Seriously though, all I needed to see was that DRAM is slower and uses more power than SRAM. I'll stick with more expensive, but better.
  • l815, on 06/19/2008, -2/+2I was an AMD fan, until that is I bought a laptop with Core Duo. Not to say that is much, but to my experience, I must say, I'm impressed Intel. They seem to be growing more and more, for the better!
  • LebenOjanen, on 06/19/2008, -4/+8tripledjrtripledjr : Thanks, it's good to see this type of articles again on Digg!

    THIS!!
  • meed, on 06/19/2008, -2/+1I can see the malware writers chomping at the bit to put their code into static ram on a cpu...
    • xenuxenuts, on 06/19/2008, -0/+2I think you might be mistaking static ram for non-volatile ram.
  • Virgule, on 06/19/2008, -2/+3So, when will computers be finally FAST ENOUGH?

    ?:D
    • luersuve, on 06/19/2008, -3/+3They will be fast enough when the run Crysis
    • chrisduser, on 06/19/2008, -1/+8When they can simulate the universe in real time and real time ^ 10, and so accurate that they can predict the future. At this point, the universe becomes conscious and the laws of physic become of her control.
    • CSharpSauce, on 06/19/2008, -1/+3I believe the answer is when we reach 42.... exoflops
    • reaperhatch, on 06/19/2008, -0/+2When they can render reality in real time.
      but then again I could see people overclocking that.
  • stephensdh, on 06/19/2008, -7/+6I've always thought CPUs should have a variable clock, this way when you have a 3% cpu usage you can slow the clock down 20 times and dramatically reduce cpu power consumption. It would be great on the battery. Then when you did need the cpu, if it was cold you could run it at 150% until the heat builds up. Just a thought.
    • VSKBadCRC, on 06/19/2008, -0/+19Sorry to burst your bubble, but a lot of them already do these things, especially laptops.
    • PueSi, on 06/19/2008, -0/+12You mean like AMD's Cool n' Quiet and Intel's Speedstep?
    • nominalgeek, on 06/19/2008, -3/+1What does this have anything to do with this article?
    • johndisko, on 06/19/2008, -2/+5Oh come on, don't bash the guy, it's a legitimate question/ponder even if it doesn't have to do much with the subject..

      peace
    • theblacknight, on 06/20/2008, -0/+1Did it occur to anyone else that he's being sarcastic by pointing out something that's been obvious for years?
  • jjssbb, on 06/19/2008, -1/+0Absolutely fantastic innovation. Imagine the reduction in ability in a human brain with seperate processing and storage areas - I think this is one of the most remarkable processor design innovations in years.
    • iduridur07, on 06/19/2008, -0/+0This is great and all, but in all the real world performance of server/desktops/laptops i have seen it is the
      hard drive that is the bottleneck consistently. I want FULL sata 2 speeds from my disks.

      Take for instance people building vmware boxes, They max the RAM and get a fast CPU. They usually forget about their 7200 rpm disks being slow.
  • DifferentAngle, on 06/19/2008, -0/+1It's unlikely that the DRAM would be significantly more dense. The two transistors need to be big enough to form a cap, and the 6 SRAM transistors are the smallest transistors on the chip - they violate the spacing rules to make SRAM as dense as possible. The DRAM in your computer memory is dense because they make the caps deep using a special process designed specifically for that purpose.

    The reason why it wasnt mentioned in whatever brochure is because the idea has been around for a long time, hasnt worked, and will never work (at least not better than SRAM). We will see 3D die stacking with DRAM chips long before we ever see integrated DRAM on an intel chip.
  • vttom, on 06/19/2008, -1/+0Been there, done that:

    http://domino.research.ibm.com/comm/research_proje ...
  • tinkafoo, on 06/19/2008, -0/+32Yep, I'm a nerd. I clicked on this before I clicked on the g-spot story.
    • Azuroth, on 06/19/2008, -0/+3Whereas, I'm at work, so I dugg the g-spot story, clicked on this, and will click on the other story when I get home :)
  • robbiemuffin, on 06/19/2008, -0/+2quote: However, we rarely here about the way how a processor actually communicates between these components and such technologies usually do not make it into marketing brochures.

    Looks like President Bush was right. Rarely is the question asked; 'is our children learning?'
  • phoenix3200, on 06/19/2008, -0/+2I'm sorry, but TG has no effing idea about what they're talking about. First, the capacitor hasn't really been removed - a PMOS transistor merely replaced it so that it's possible to integrate it in the die. Next, DRAM is a huge power waster. If you disregard that you have to refresh it periodically and that a read operation will definitely destroy the cell's contents, you still need to worry about separate circuitry to convert the small stored signal in the cell into a larger, usable signal. On top of this, process variations would destroy the reliability of the cells themselves.

    The only thing Intel presented here is that it is feasible to integrate DRAM on to the processor. They didn't announce it advance because they would have been ridiculed if they put faith in this design.
  • jfsimard79, on 06/19/2008, -0/+3Let's not forget using atoms as bit storage, quantum computing, and also I read that there may be a time travel cpu that calculates a result before the instuction set even asks it.
  • bundwallah, on 06/19/2008, -0/+0Seems like a natural progression. I wonder what kind of densities they can achieve. I assume it cant be that much unless they want to end up with massive CPU bricks like the Itaniums. Those things are huge! I'm sure a GB of RAM on the CPU won't hurt though. :)

    CPU's are plenty fast these days though, its the motherboard buses that need to evolve. The"PC" standard is quite old. Sure PCI-X, PCIE are nice addons but fundamentally, we've got the same old PC from 10 years ago. I remember reading about a company experimenting with fibre-optics on the motherboard. That's pretty cool.
  • Wakkyweed, on 06/19/2008, -1/+4To me the article read like this:

    Intel's new chip (gobble-de-gook) processor generation 65 (absurd tech spec) technology that holds the potential to (jabberwocky) reduce the size of DRAM cells (more absurd tech specs) in contrast to (something to do with 2 transistors as opposed to 6 transistors) and in conclusion (technical specs about bandwith) chips will be faster.
  • Biohazard6601, on 06/19/2008, -3/+1is it sad that I completely understood that?
  • hollyminkowski, on 06/19/2008, -0/+3Getting everything closer together is so important now that cpu speeds are getting so high.
    Got to feed the beast as quickly as possible and external bus wiring is so last century.

    Heat is the one big problem...how do you get the heat away from the silicon fast enough.
    Replacing silicon wafers with the more robust diamond wafers would help since they are much more heat tolerant. Cultured diamond manufacturers are improving wafer creation, but the quality/quantity/price is just not there yet.

    A 3d block with lots of cpu cores and integrated dram and many other hardware parts that are now spread out on a motherboard is a logical step. Something like a 256 core, terabyte dram processor block with active cooling and optical data I/O ports is in our future. A super-computer building block.
    The block's only physical connection would be 2 wires to feed it dc power....as low voltage as possible to help with the heat problem. Someday when the power requirement for such a processor assembly gets low enough it might not have any physical connections at all and be powered optically or inductively.
  • LeRenard, on 06/19/2008, -3/+1I don't see how it can be efficient to keep building on the x86 architecture. At some point, won't they have to scrap that and start from scratch?
    • Suricou, on 06/19/2008, -0/+3Nice idea, but there is far too much software base. It's an option for gameing platforms - hence the PS3 and it's Cell. Less so for PCs and servers, unless you include an x86 backwards compatability that really defeats the object. Look at the transition to x64 for example.
  • stutimandal, on 06/19/2008, -0/+1DRAMs suffer from intense leakage currents and high refresh rates in scaled (45nm and below) technologies. I don't see why this will solve the SRAM leakage problem.
  • LoudMusic, on 06/19/2008, -0/+1Yay, faster CPUs. I'm still more interested in faster disk times. Or better yet, a system where there is no difference between RAM and disk.
    • Suricou, on 06/19/2008, -0/+1You can have that now. DRAM-using ramdisks are available. They are just really, really, really expensive.
  • franl, on 06/19/2008, -0/+5The article says "This approach would completely change the programming model since there are no longer any concerns over cache misses."

    Since 99.99% of all userspace code is written without regard to cache misses, how exactly does this "completely change the programming model"?
    • Suricou, on 06/19/2008, -0/+3The programmer might not have to worry too much about misses, usually. But the optimising compiler certinly does.
  • humperdeath, on 06/19/2008, -1/+1I used-to-could do that.
  • Drchester, on 06/19/2008, -0/+1I dugg this only because the author can't spell or write. This is the worst sentence I have read in a long time: "However, we rarely here about the way how a processor actually communicates between these components and such technologies usually do not make it into marketing brochures." W. T. F. does that mean? I HERE this guy needs some help in his prose. Grab some English classes buddy. I may take you seriously next time.
  •  

    Login or register to add a comment

    Create a new account or login to join in the conversation on Digg. You'll also be able to Digg stories to help promote things you like.


© Digg Inc. 2008 — Content posted by Digg users is dedicated to the public domain.
DIGG, DIGG IT, DUGG, DIGG THIS, Digg graphics, logos, designs, page headers, button icons, scripts, and other service names are the trademarks of Digg Inc.