Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
AMD Hardware

On-CPU Peltiers From AMD? 226

Posted by timothy
from the speculatoryism dept.
Hack Jandy writes "Remember those people who lived on the edge and put peltiers between their CPU and heatsink (or your favorite beverage)? A peltier is a devices that gets cold on one side and warm on the other when an electrical current passes through it. It looks like there is talk that AMD will actually incorporate some of these devices on the CPU according to Xbitlabs. AMD already incorporates some degree of the peltier effect with it's Silicon on Insulator."
This discussion has been archived. No new comments can be posted.

On-CPU Peltiers From AMD?

Comments Filter:
  • pelltier? (Score:3, Funny)

    by micronix1 (590179) on Sunday October 24, 2004 @01:13PM (#10614851)
    dont put it on the wrong way.


    • According to an old Slashdot article [slashdot.org] a British company called Cool Chip Plc [coolchips.com] has something that uses the "Quantum Mechanical Electron Tunneling" to achive "unbelievable cooling efficiencies".

      According to the Press Release [coolchips.gi] it is claimed that the device is so good that "a panel about two inches square will have the capacity to provide the air conditioning for a living room" !

      In comparison, according to Cool Chips's press release, most existing cooling systems use compressors and environment-damag
  • by Qzukk (229616) on Sunday October 24, 2004 @01:15PM (#10614865) Journal
    The problem with peltier coolers is that if it breaks down, the once cooling surface becomes an insulator. Plus, if the hot side gets too hot, the cooling process breaks down, so anyone using this would have to use a cooler that can draw the heat away as fast as the CPU-side peltier can kick it out, which would probably be another, larger peltier.

    I'd rather stick to external cooling systems that I can monitor and replace if necessary.
    • by gl4ss (559668) on Sunday October 24, 2004 @01:20PM (#10614898) Homepage Journal
      i don't think the realiability is the real issue.

      it seems xbitlabs is just.. well.. doing what journalist wannabees like - take some facts and twist them.

      for one: YOU DON'T GET MORE -EFFECTIVE- cooling with peltiers. you end up using more power than you would with normal cooling. the total heat output gets _increased_.

      unless they(chips) can't work in normal room temps there's not really much point in using peltier cooling in cooling them, except if you somehow manage to get the effect 'for free' or something.
      • by Pyromage (19360) on Sunday October 24, 2004 @01:35PM (#10614978) Homepage
        The key thing with peltiers is that they just *move* heat. This can be more effective cooling because you can move heat from the CPU core (normally a very small area) to a much larger area. Yes, your net heat is a bit more, but you have a reasonably sized area to cool, which is a much easier problem.
        • by gl4ss (559668) on Sunday October 24, 2004 @01:55PM (#10615075) Homepage Journal
          peltiers not really at 'effective' at that.

          and let's stop here for a minute, simplifiedly.. a 100watt cpu, put some what, 250 watts(? or so) into the peltiering.. then you got 350 watts to get rid of 3 millimeters away from the original source!(you still need water & whatever to get rid of the heat)

          with current efficiency it's only useful in extreme situations where you wouldn't mind such waste. it's only useful if you need such low temps for the cpu that you can't attain them otherwise!

          it's not just a "bit more". and as for to getting it to a more reasonable area.. that's what heatsinks are for, that's what you would use anyways with the pelt setup to get rid of the heat(or watercooling or whatever, the point is you don't really spread the heat to a larger surface with peltiers).
          • A 250W peltier is massive. 50W is more reasonable for high end cooling. Remember, the chip does not turn all the power it recieves into heat, so 250W would cool it to beyond where you'd realistically need.
            • Actually, a chip does turn all the power it uses into heat. The "useful work" that's done by a CPU is just charges moving around, but eventually it all ends up as heat anyway.
              • Well, ultimately everything gets turned into heat, but to say that 100% of the energy going into the CPU is released as heat is massively inaccurate, actually it violates several laws of physics. If the CPU was capable of turning 100 W of electricity into 100 W of heat while also doing any work at all (much less the very significant work a CPU does) using that initial energy would obviously be rather impossible.
                • If a CPU didn't turn 100 % of electrical power into heat, what should the remaining energy be transformed to? The CPU neither lights nor drives the PC around the room. Kernel compilations don't count as physical "work".
                • Most but not all of the electric power going into the CPU is turned into heat in the CPU. The CPU drives external wires, some of which may be terminated. Energy is lost in the termination, and some (hopefully small) amount is turned into elctromagetic radiation.
            • haha 50watt peltier for modern high end cooling??

              hahahahahaaha.. if you could get away with such it would be quite a bit more popular.

              the peltier needs to be able to move all the heat coming from the cpu(and yes it churns out everything as heat pretty much).
          • It's really geting to the point where it's impractical even with water cooling, 300W and more is a hell of a lot of power to shift, more and more people are dabbling with refrigeration as the next step up from a plain water cooled solution, the next step up from that is cascaded refrigeration... geeks need toys to live :)

            Sorry for my spelling, just to stop the abusive ****'s who moan about it, write a spellchecker for slashcode if youre that upset.
        • by leathered (780018) on Sunday October 24, 2004 @02:14PM (#10615178)
          The key thing with peltiers is that they just *move* heat.

          Which is exactly what your domestic refrigerator does, merely moves heat from the inside via the evaporator to the outside to the condenser. In fact heat cannot be destroyed at all (think conversation of energy), merely moved elsewhere.
        • It's not a bit more heat. It's a lot more heat.

          The only time they're worth it is when you're trying to achieve a temperature below the ambient temperature. Otherwise, it's easier to put a heat spreader on the chip (as AMD and Intel already do) so the contact area with the heat sink is bigger.

          These peltiers aren't going to go in any general purpose CPUs as we know them.
      • Wrong.

        Peltier increases temperature difference. Difference helps dissipation.

        What is easier: In environment of 30C, to get a plate of metal down from 50C to 40C or from 120C to 100C?

        If you watch heat dissipation curves you'll notice heat dissipation drops rapidly when temperature is getting close to ambient. So in your basic case keeping the CPU at 70C using plain CPU fan is way harder than keeping the CPU at 40C using Peltier module while keeping the other side of the module at 150C using a plain fan. N
        • i almost put that in there.

          the diff is there but in real life computing it's not practical to have some parts that are over 150c(and with the effiency % of peltiers available it would be just insane)..

          what the tech could be what they're really meaning this could be something that's built into the chip to move the heat away from the lower layers to the top, or something, even then it would be incredibly wasteful

          • Note 150C on surface of the radiator turns into reasonable 50C maybe 1cm away from the CPU.

            I wonder why won't they start producing cases and motherboards with built-in water cooling, safe CPU socket including all the water plumbing, then some water channels THROUGH the core. Don't transfer the heat to surface of the CPU, just receive it inside. Most of the "on-chip plumbing circuitry" is already there - in inkjet printer heads, which are in fact quite sophisticated ICs with ink channels driven through them
      • YOU DON'T GET MORE -EFFECTIVE- cooling with peltiers. you end up using more power than you would with normal cooling. the total heat output gets _increased_.

        The same is true of ANYTHING; even the cooling fans in your computer result in more total heat output. But try building a computer without them!

        Peltier coolers could serve a good purpose. Yes, it results in the total heat output being more, but the point is that it will keep your CPU much cooler, even though it's using more power.

    • > which would probably be another, larger peltier.

      Perhaps this is the solution for Pentium 4 cooling.

      It's peltiers all the way down (up).
    • Erm, I fail to see how that's an argument. Just because you've got a hot side to cool doesn't mean it's a poor idea. A Prescott processor from Intel runs at an incredible temperature, how's cooling this any different from cooling the warm side of a peltier? Yes, you'll destroy your TEC if your cooling solution fails, but you'd crash before damaging your CPU, and you could apply that argument to any cooling sysyetm - if your CPU fan fails then you'll crash just the same.
      Oh and I submitted this story two da
      • This is a vast understatement. I used to run a TEC cooled system with a waterblock cooling the hot side of the TEC. A failure of my water cooling pump cause all the coolant to boil off because of the incredible heat generated by the proccessor combined with the "heat pump of the tec". The proccessor failed and the system shutdown, but the Heat from the tec melted the solder that held the two halves of the cooling block which then dripped all over the back of the video card... The Tec eventually reached some
    • People are all commenting about strap-on Peltier "coolers" or TECs that overclockers use. They have failure modes that include condensation leading to oxidation and power failure, leaving an insulating brick on top of your processor. An integrated peltier junction layer in a Silicon On Insulator processor could not fail separately from the processor itself, it's just a solid state layer of semiconducting materials deposited as part of the fab process. So the power is always on when the processor is being
    • Problem is, off-chip peltier units are separated from the hot parts of the CPU by packaging materials, at least some of which are thermally insulating. This is fine when thermal output is roughly evenly distributed on your CPU, but this is becoming less and less the case. The on-chip peltier helps smooth out the thermal distribution, allowing the external cooling device to keep the chip operating properly without requiring liquid nitrogen.
    • The problem with peltier coolers is that if it breaks down

      It's a couple of dissimilar materials in electrical contact with a small current going through them - fairly easy to do reliably. Some combinations will work above 1000C in fairly nasty environments, it's just a thermocouple run in reverse. There are limits into how much heat you can move dependant on the size of the thing, but you design for the range of expected conditions.

      so anyone using this would have to use a cooler that can draw the heat a

  • Peltiers? (Score:5, Interesting)

    by polecat_redux (779887) <spamwichNO@SPAMgmail.com> on Sunday October 24, 2004 @01:17PM (#10614877)
    I have had limited experience with Peltiers in the context of CPU overclocking, and I must say, my impression was that they're not all they're cracked up to be. Yeah, with a large enough cooler, you can drop the temp of the CPU significantly, but the effect is dependent on your ability to remove the heat from the other side (which is in excess of that given off by the CPU to begin with). A good water-cooling solution works well enough without the need for the extra drop in temp.
    • Yeah, Peltiers are only about 5% efficient at best, which means they generate a whole fuckload more heat than they take away on the cold side. Water-cooled peltiers seem to work out pretty well.
      • Water-cooled peltiers seem to work out pretty well.

        Yeah, they do. I had such a setup running an Athlon 750 at 950MHz at one point. As I recall, even at close to full load on the CPU, the temp was significantly below ambient. Though, I learned the hard way how important it is to effectively protect the chip against condensation...
        • Though, I learned the hard way how important it is to effectively protect the chip against condensation.

          Hmmm. Pure water (condensed) doesn't conduct electricity... It will however corrode metal leads.
          • Hmmm. Pure water (condensed) doesn't conduct electricity... It will however corrode metal leads.

            Yeah, the proc still worked, but it did turn several shades of green, blue, and orange.
          • Water + a trace of impurity + 2 electrodes of potential difference > 3volts, you are running a eletrolysis experiment. The same effect is observed whether it is held in a beaker or on a PCB. Umm.. the copper trace turned green already.
    • by fireboy1919 (257783) <rustyp.freeshell@org> on Sunday October 24, 2004 @01:43PM (#10615006) Homepage Journal
      Well, the Peltier does accomplish one thing: it covers that unsightly CPU. I made my Macrame peltier (actually, I prefer the term "CPU Cozy") using simple macaroni, construction paper, but decorated with markers and some stickers from the craft shop.

      The inside of my case looks much better now. Except that it keeps catching on fire when I play games.
  • Power... (Score:5, Informative)

    by PaintyThePirate (682047) on Sunday October 24, 2004 @01:17PM (#10614881) Homepage
    Just imagine the amount of power required for something like that. Each core of a dual core 65nm CPU will need at least 70 watts of power, giving 140 watts total. In order for a peltier to be useful, the wattage has to be greater than that of the CPUs. So with only the CPU and Pelt, that is a minimum of 300 watts, with something between 400 and 500 being more likely. That is an absurd amount of power and heat.
    • In order for a peltier to be useful, the wattage has to be greater than that of the CPUs.
      uhm... Are you sure you have to put more energy into the Peltier cooler itself than into the device it's supposed to cool? While the Peltier cooler does use some energy itself, IIRC in order to transfer heat from a 100W device the cooler itself does not consume more than 100W.

      • *uhm... Are you sure you have to put more energy into the Peltier cooler itself than into the device it's supposed to cool? While the Peltier cooler does use some energy itself, IIRC in order to transfer heat from a 100W device the cooler itself does not consume more than 100W.*

        if that were true we could be building some pretty intresting mini powerplants.

        • No. Air conditioners only require 10W to move 100W of heat to a different location. No laws of physics. Think amplifier here.
  • by Anonymous Coward on Sunday October 24, 2004 @01:25PM (#10614931)
    hot on one side, cold on the other. sounds like my bed.
  • by doormat (63648) on Sunday October 24, 2004 @01:40PM (#10614995) Homepage Journal
    is not just the total amount of heat they put out, but the fact that they put out that much heat over an area of about one square centimeter (on the 90nm process at least). As the physical piece of silicon shrinks, the thermal density increases. More transistors switching on and off in a smaller area, and the drop in Vcc isnt enough to counteract the increase in density (we were at 1.8v or so with the 180nm process, and now at 90nm, we're at 1.4v or so - some chips dynamically change voltage and multiplier based on demand). I'm not sure this will do a whole lot of good if you just try to disapate the heat from the processor and the heat introducted by the peltier effect over the same square centimeter. You'd need to disapate the heat over a much larger area, say 10 sq cm. They you can stay in the realm of air-cooling instead of watercooling.
  • by LiquidMind (150126) on Sunday October 24, 2004 @01:47PM (#10615026)
    "A peltier is a devices that gets cold on one side and warm on the other..."

    come up with your own shit. my g/f patented this 'technology' years ago
  • by Brane2 (608748) on Sunday October 24, 2004 @01:49PM (#10615041)
    AMD is patenting this as a way of *getting around* of SOI disadvantages. SOI means silicon on insulator , which is in this case SiO2, which is also excellent thermal (not only electrical) insulator. AMD says that SiO2 conducts heat at least hundred times less than silicon.

    What they are saying is that transistors on SOI might behave better, but they are certainly running hotter than their classic countepairs, since layer of SiO2 stands between them and the cooling system.

    So AMD is proposing several schemes of embedding TEC device into the insulating layer in the silicon. This layer would:

    1. Decrease overall thermal resisstance of the cooling path

    2. When powered on, offer bigger thermal diferential, since it could cool embedded side of the TEC significantly below the cooler temperature.

    It is unclear if they intend to use this on the whole chip, or just the especially hot areas...

    • I can't believe this is marked as informative. The SOI layer is very thin, so it contributes very little to the overall conductance of the Si substrate. And, one of the biggest advantages of SOI is that you get less leakage current, i.e. less wasted power, and less heat dissipation.
    • " AMD is patenting this as a way of *getting around* of SOI disadvantages."

      "US patent number 6,800,933 was filed on the 23rd of April 2001, with the following abstract:... "

      I doubt its primary purpose was to address an issue with technology not in use until several years later.
  • by ozzee (612196) on Sunday October 24, 2004 @01:59PM (#10615089)

    I remember that SGI once used peltiers and they had to recall them because of failures due to corrosion due to condendsation because the device temperatures fell below the dew point.

  • by caseih (160668) on Sunday October 24, 2004 @02:06PM (#10615119)
    Maybe I'm wrong here, but it doesn't sound like AMD would be using the peltier as a replacement for the fan and heatsink, but rather building in a peltier into the silicon itself to pump the heat out of the CPU core itself faster, so that the heatsink and cooling fan on top can keep the core cooled. As someone mentioned, as we increase the density of the cpu die itself, the thermal density is also decreased and thus the problem becomes getting the heat from the core of the silicon wafer out to the outside of the chip or wafer itself. If we put peltier material into the wafer, we can electronically pump this heat to the surface where traditional cooling devices can disappate it into the air
  • mistakenly read the first sentence of this article [xbitlabs.com] as:

    "Advanced Micro Devices, one of the world's leading makers of central heating units, has patented a technology...."

  • Nitpick: (Score:5, Informative)

    by drinkypoo (153816) <martin.espinoza@gmail.com> on Sunday October 24, 2004 @02:47PM (#10615337) Homepage Journal
    Peltier is the guy who discovered the effect. A peltier junction (sometimes called a peltier cooler, which is a stupid name because it's also a heater) is the solid-state heat pump. A peltier junction is not called a peltier.
    • (sometimes called a peltier cooler, which is a stupid name because it's also a heater)

      Double or nuthin'. Most coolers are also heaters. The more correct name for these are heat pumps, although the peltier junction can actually create electrical engergy from heat.

      • Yeah, I did note that it was a solid-state heat pump. Given that every reaction is inefficient, not only are all coolers heaters on a certain scale (for instance, more heat is generated making ice than heat will be "absorbed" (moved from one place to another) by the ice being reheated later) but all coolers are really heaters, which is why leaving the refrigerator door open and taping the close switch down will heat your house, not cool it. Some people have a hard time bending their mind around that one...
  • by MetalliQaZ (539913)
    I suppose this will be strictly for their desktop processors, since the Peltier effect truly consumes a very large ammount of power. I couldn't see this technology being used in any rational way on a mobile proc, and since the two markets are converging, I question this move from AMD. Anyway, it will be interesting to see in what form this technology is actually realized...
    • Consider a typical EER for a peltier air conditioner of 0.33 compared to 9.5 to 13 for a freon one. It's an interesting physical phenomenon, but a huge waste of energy.
      • by dbIII (701233)

        Consider a typical EER for a peltier air conditioner of 0.33 compared to 9.5 to 13 for a freon one. It's an interesting physical phenomenon, but a huge waste of energy.

        First, you don't use it if you can just move gas around. Since it's just a couple of dissimilar materials in electrical contact at two ends it can fit in very tight spaces.

        Second, the efficiency is going to vary enourmously depending on the temperature difference of the two junctions, so it's very hard to pick a number out of the air. In t

      • It's an interesting physical phenomenon, but a huge waste of energy.

        It's not amazingly effecient when you are using it for cooling, but why aren't any electric heaters using it?

        You actually get better-than 100% effeciency with Peltier heating, and freon certainly can't do a better job at that...
  • Apple's warranty experience with the G5 is going to influence what happens on the PC side. If Apple's water cooling solution doesn't incur any undue warranty claims, I'm willing to bet that AMD and Intel will simply mandate water cooling for their chips. I can even imagine having water channels embedded in the chip itself to cool the chip from the bottom as well as the top.

    Once Intel and AMD accept the heat losses associated with the smaller dimension gates because they can just pipe the heat away, the MHZ



  • I can't speak for AMD, but I'm positive Intel uses Minx pelts.
  • Peltiers are very inefficient and therefore will cause your computer system to need a lot more cooling. The cool chip stuff [slashdot.org]
  • As long as someone rids the world of thermal paste, I will be happy. There has never been a more annoying substance that that crap.
  • Thermodynamics (Score:2, Informative)

    by Roger_Wilco (138600)

    The heat dissipated by a heat sink (with a fan or not) is a linear function of the temperature of the surface of the heat sink.

    So if I have a CPU which puts out, say, 100W of heat, and I have a particular size of heatsink, the temperature of the heat sink will rise until the dissipation of the sink is 100W (or the chip melts). Say my heatsink dissipates 100W at 100C. (All numbers are made up.) Let it conduct perfectly, too, to simplify the discussion. If I put out more than 100W, the temperature of the he

    • by ahecht (567934)
      The problem is that the heat sink has to dissipate the heat from the CPU and the heat from the peltier itself. Lets say that, for example, the peltier used 20 watts to get that 20C (a real peltier would use MUCH more). This means that the heat sink has to dissipate an additional 20 watts, and as you said, this means that the temperature of the heat sink rises. Lets say it rises 20 degrees. Now, you have your 20 degree difference bringing your CPU temp to 100C, right where you started. All you have done
  • Editors? (Score:3, Interesting)

    by Max Threshold (540114) on Sunday October 24, 2004 @06:46PM (#10616637)
    Aren't editors supposed to catch and fix simple grammatical errors in submissions? Isn't that part of what editors have traditionally done?

    What do Slashdot editors do, anyway?

EARTH smog | bricks AIR -- mud -- FIRE soda water | tequila WATER

Working...