back to article IBM fills chips with water

Oh, those crazy kids at IBM's Zürich lab. They can't get enough of processor cooling technology. Two years ago, IBM announced a breakthrough with carving channels into chips to let thermal paste sink down nearer the silicon. Then, in 2007, IBM Zürich patted itself on the back for cutting more channels into the cooling unit. …

COMMENTS

This topic is closed for new posts.
  1. Tony W

    One thing wrong, what about the rest?

    "10 times greater than the heat generated by a hotplate"? No, typical hotplates are one to two kW. Always makes me doubtful when a press release has a mistake that obvious. What else has the PR person misunderstood?

  2. E

    Watts / square CM

    Is 180 W/cm2 hotter than the surface of the sun? ;-) Also what is this number in ergs per cubit?

  3. Moss Icely Spaceport
    Thumb Up

    Why not oil?

    I prefer my chips deep fried afterall.

  4. Dick Emery
    Paris Hilton

    Furry pipes

    Surely not just plain water? Well not the stuff that flows from my pipes at any rate unless you want a severe case of limescale.

  5. Anonymous Coward
    Flame

    @Tony W

    <<Always makes me doubtful when a press release has a mistake that obvious. What else has the PR person misunderstood>>

    Well Tony - the 1 kWatt power on a chip seems to be a bit off. The typical CPU is about 40 W - or less - with an area of about 1/4 cm2. Soooo, maybe 1kW/cm2? (Even that seems a bit high - but at least it MIGHT be within reason.)

    Had to use the flame - as it is a HOT subject.

  6. I. Aproveofitspendingonspecificprojects

    > Also what is this number in ergs per cubit?

    3.76284 penguins.

  7. mark
    Flame

    obvious mistake?

    @Tony W

    so your typical hotplate is 4cm2 in size?

    i think your mistake is obvious...

  8. Anton Ivanov
    Boffin

    Re: One thing wrong, what about the rest?

    Per cm2 most likely. A typical hotplate is more than 100cm2.

    Anyway, the ex-chemist in me keeps wondering:

    What kind of water are they using and how often does it needs changed?

    My gut feeling is that this setup requires at least double-distilled and deionised water. Tap water is definitely out of the question here. If we ignore the lymescale bit in this setup water is actually in contact with the semiconductor itself and if it starts conducting things are bound to get really funny. The only way to make water non-conducting is to deionise and double-distill which is not cheap. On top of it the ghastly thing does not want to stay in that shape. It will scrape ions off the walls of the pipework and literally suck them out of thin air. By the time you turn around its dielectric constant shows that it is a conductor again and full of "pleasant" ions that will do their best to solidify in the most inappropriate place.

    Frankly, something less capricious and temperamental than H2O is probably a much better idea for this one.

  9. Dominic Kua
    Boffin

    Water's got its advantages

    Look at the specific heat capacity, and availabiltity of the basic resource. You can buy a water cooling kit for overclocking your CPU now, so it's not exactly completely new technology.

    Given the size of the pipes it'll be flowing through and the heat it'll need to shift, you're going to want as high a specific heat capacity as you can get, or you'll be pumping masses of fluid through the processor.

  10. Steve
    Go

    hot plate

    It doesn't mention watts/cm2 in the comparison with the hot plate

    "3-D chip stacks would have an aggregated heat dissipation of close to 1 kilowatt - 10 times greater than the heat generated by a hotplate"

    Which is just nonsense, hot plates don't run at 100watts.

    As for not being new, it's barely even comparable to existing water cooling. At the moment you strap a heat sink on top of your chip, which you cool with water.

    This design is talking about pumping water directly through a 3d stack of transistors. That's like saying the car wasn't new technology because wheel barrows already existed.

  11. Adam Foxton

    @George Schultz

    40W for a processor? Newer AMDs are up at the 90W levels, the mid-age P4s topped 100W IIRC, etc etc.

    So you're looking at- say- 100W over a 4cm2 area. That makes the power 250kW/m^2 by my reckoning. A 33cm x 33cm hot plate at 2kW would be 18kW/m^2.

    So yes, processors are far hotter than hot-plates.

    And for those wanting to check my figures:

    4cm2 - say 2cmx2cm. 50x each side for a sq meter. then multiply by the power of each one of these 2cmx2cm cells you've just tiled.

    So 50x50x100W.

  12. ShaggyDoggy

    Hot plates

    I think he meant that if multiplied up to a 'normal' hotplate size it would be etc etc

    ... or maybe I'm being too kind here

  13. hugo tyson
    Paris Hilton

    Hotplate?

    They might mean the hotplate under a coffee percolator, that's certainly not anything like as powerful as an actual 1kW electric cooker ring? To me a hotplate means a device for keeping warm something you already cooked, at low power, isn't it? Anything you cook on is a ring.

    Paris because she'd have no clue in a kitchen too.

  14. S Stewart
    Happy

    Pasta anyone?

    So I could build a super-computer into my hot-plate. I could then exploit the excess processor heat to cook my dinner...and 10 times quicker too?

  15. Secretgeek
    Coat

    Energy re-use.

    How about running an ickle CHP plant with the hot water. Ok so it'll be larger than your average box and massively more expensive but hey in 45 years time it'll have paid for itself.

    I'll get my coat.

  16. Anonymous Coward
    Anonymous Coward

    Amazing how people get hung up on words...

    Processors are hotter than hotplates, not quite 10 times as much, but still hotter. Why? Because the 90W odd that they pump out over their size (400-900 mm2) is, when extrapolated to the size of a hotplate, definitely more than a hotplate.

    The 2kW hotplates tend to be 31,400 mm2 (r^2*3.14 = 100mm ^2 * 3.14 for a 20cm dia hotplate), 35 times the area of a processor. A processor would have to pump out less than 57W for it to be less hot than a 2kW hotplate. Generally processors pump out up to twice that.

    You do the rest of the math.

  17. Anonymous Coward
    Alert

    Ethylene Glycol... a real sweet coolant!

    Is the answer to all your problems unless you want to make a curry to bump of your spouse. Yes I know its an antifreeze but that antifreeze mix that you use in your car is also desigend to prevent corrosion. the question is will you buy it at $4 a bottle from the garage or $400 from IBM...

  18. mike brockington
    Boffin

    @AC

    "Amazing how people get hung up on words...

    By Anonymous Coward

    Posted Friday 6th June 2008 11:05 GMT

    Processors are hotter than hotplates,"

    I would imagine that most electric cooking rings reach temperatures of well over 200 degrees centigrade / celsius, as they need to be hot enough to transfer heat to cooking oil.

    I don't think many processors, or other electronic devices, can survive even 100 degrees centigrade / celsius.

    Temperature != Power

    /pedant

  19. Aodhhan

    Missing math

    I still don't buy it, and the math is off due to missing variables.

    Wattage doesn't create heat, resistance does.

    Hotplates and processors are both built differently for different purposes. Processors are using the majority of their watt energy for processing, therefore the resistance heat isn't as efficient per watt. They are also built of material which doesn't hang on to the heat for a long period of time.

    Where as a hot plate is using watt energy and creating as much resistance with it as possible. Therefore it is more efficient in creating heat with each watt. It is also made of material which is designed to hang on to heat as long as possible.

    A brick holds heat better than a sponge or a block of silicone.

    Now, reconfigure your tests using the above information, and you'll forget about wattage vs size.

  20. John Angelico
    Joke

    Water Cooled Processors? From IBM???

    [Henry Crun:] Mngh, water cooled, eh? Er-m, mngh, Min isn't that something like the old System-360 we had at our lovely little inn down at Brighton? Where we kept the records of all those Mr & Mrs Smiths. Do you remember?

    [Minnie Bannister:] Yes, Henry, there were so many of them too. We needed a serious database for the storage. Brown paper bags just didn't have the capacity, did they?

  21. John Dougald McCallum

    @ AC 6th June 2008 11:11 GMT

    Ethylene Glycol also raises the boiling point of the water it is mixed with,so that makes it even better than plain double distiled water as mentioned in an earlier post .IF this technology ever comes to market will it mean EVEN smaller laptops and PC's?

  22. Matt
    Coat

    processors live longer with Calgon! (TM)

    ill get my coat... the one without limescale!

  23. Anonymous Coward
    Anonymous Coward

    @mike brockington

    Mike, you may be correct, but do remember that without appropriate heat dissipation (whether bonded directly to a metal plate or via heat transfer), 35 processors will over the same area easily reach the necessary temperatures.

    Put 35 (based on 30x30mm) or 78 (based on 20x20mm) processors together on the same heat-transfer system, and you will start getting toasty. My laptop gets deliciously hot (not!) when I use it hard, and that's got an Intel mobile processor in it. The air that it gets rid of, is hot. We're talking 40 degrees here. Harness that wattage that's transferred by those processors and your oil will start boiling - eventually :-)

  24. Daniel B.
    Linux

    Re: Energy re-use.

    Heh. You just reminded me of the old computer lab at college. As I was a temp-admin there, I had the keys for the small private office, which gave me a cozy space to sleep when we had to do overnight stuff. The "lab" was actually a "converted" warehouse, so on cold nights it got mighty cold, and we had *no* heating facilities.

    So those days, I just fired up the 5-Pentium-class "cluster" and let it heat the small office. Nice energy re-use, even if the cluster itself wasn't really used after its initial demonstration phase.

    Penguin 'coz the cluster ran Linux.

  25. Britt Johnston

    Other than water?

    Recently losing a USB stick in the wash confirmed my inate distrust of mixing water and electrons. What's wrong with low-watt components?

  26. Owen Milton
    Go

    Perfect solution...

    All we need to do is run the water through the chips and then dump it via a heat exchanger to more water... a built-in coffee (or tea) maker in every computer! Mwhahaha, my plans to over-caffeinate the world will proceed at flank speed!

  27. This post has been deleted by its author

  28. Brett Leach
    Paris Hilton

    @Brit Johnston

    Next time something like that happens rinse it well under the tap and then leave it somewhere warm for a day or two to dry out completely.

    Might even work with the stick you already washed.

    I've rescued a couple of pocket calculators that got dosed with coke this way. Plus several watches that have gone through the wash. Remove the batteries first.

    Paris coz she knows all about fluids gumming up the works.

  29. stizzleswick
    Stop

    Re: Water is OK

    "Cray Research used Freon..."

    They also used liquid nitrogen for a time, which has certain advantages. Though they needed a whole basement full of cooling machinery to keep the stuff liquid in the case of the MPs. Darned superconductors.

    Unfortunately, the problem here does not call for coolants that evaporate -- that would blow up the stacked chip. So we need a medium that does not itself cool anything, but transports heat. And cooling the entire system down so that Freon or its antedecessors won't evaporate will most likely lead to unwanted side effects, like some places on the board becoming superconducting, thereby frying the nearest chip.

  30. Sirus Black
    Stop

    Freezer ?

    y cant we just put the whole cpu in a freezer ?

    or cant we just soak the whole thing in distilled water ?

  31. Anonymous Coward
    Anonymous Coward

    H2O

    "Frankly, something less capricious and temperamental than H2O is probably a much better idea for this one."

    In what way is H2O temperamental?

    A lot of people are shocked at water being in close proximity to electricity, and most of the time, that fear is justified, if you're talking 230 volt mains electricity and impure water.

    The atoms of Hydrogen and Oxgygen in water form covalent bonds, very pure water does not conduct electricity, it's the presence of impurities, salts, which create ions in the water that enables water to conduct electricity.

This topic is closed for new posts.

Other stories you might like