back to article Mobo maker builds 'powerless' processor cooling fan

Mobo maker MSI has come up with a novel way of keeping processors cool and conserving energy at the same time: the CPU to drive its own fan. It's actually a new implementation of an old discovery. MSI's Air Power Cooler uses the energy inherent in the expansion of air as it warms up to drive a fan. MSI's Stirling Engine …


This topic is closed for new posts.
  1. Kevin Johnston

    smart engineering

    Good use of some simple principles here. I assume there will be a trickle fan or similar to top-up the cooling but even so it helps break to loop with more power used to cool down then more power needed to cool the PSU then...

  2. Anonymous Coward
    Thumb Up

    The PC needs to be in a Gothic Case

    for that authentic steampunk look.

  3. Campbell


    Let me be the first to say, Cool.

    Got my coat.

  4. Sam


    Beat me to it...

    Is it quiet as well, or does it sound like a certain animated steam train?

  5. Joe K


    So the more you overclock, the faster it'll go, thus reducing the temp, so reducing the fan, so the CPU gets hotter, thus speeding up the fan, which'll cool it down, so the fan slows, and so it gets hotter, which then gets the fan going......

    I'm lost here.

  6. James Le Cuirot

    Perpetual motion

    "It's not totally efficient, which is why it won't run in perpetuity, but it is kicked into motion simply by the heat generated by the chip."

    I was gonna say, wouldn't this be a case of the theoretically impossible perpetual motion. Now it makes sense.

  7. Simon Day

    Are you sure its a CPU cooler?

    The CPU slot looks quite obviously empty there?

    That looks more like an oversized chipset cooler....

  8. Bronek Kozicki

    CPU ?

    From the picture it seems like it is not the CPU, but north (or maybe south) bridge being cooled by this curiousity

  9. G R Goslin

    Nice idea, but...

    .. the only snag with this type of Stirling Cycle engine is that it is not self starting. You have to spin the shaft for the thing to work. Given that it's buried in the innards of the machine, this does'nt seem altogether practical. They're not particularly quiet, either, given the number of linkages and reciprocating masses. But points for a nice try.

  10. Dazed and Confused

    Now that's cool

    OK, I couldn't resist it.

    Mines the oily one over there with the torque wrench in the pocket.

  11. Tom

    I don't think so...

    "MSI's version can transfer over 70 per cent of the heat power to motive power, the company claimed,"

    I think they have an extra zero on that...

  12. Tim Schomer


    ... can I get my hands on one??? Can I imagine that any duff boards would make good money on Ebay as geeks buy them for the heat pump toy?

    (MSI should market these on their own, they'd make a fortune!)


  13. Bob

    Nearly Perpetual Motion...

    Really, this translates 70% of the heat power into mechanical energy in the fan? There must be a mistake. This is a heat engine, and heat engines are never 70% efficient. Large electric generation systems --- some of the most efficient heat engines out there --- are only maybe 40-60% efficient MAX.

    I would believe that this Stirling engine is 7% efficient.

  14. Tim
    IT Angle

    One or two problems

    With the way this is setup it will either blow hot air towards the ddr dims adding heat or suck the warm air back over it's own heat sink.

    most north bridge coolers vent at 90 degrees to the board to allow the case fans to expel the heat.

    Story says cpu fan when intell socket can be seen on photo?.

    If it does work on the cpu how will the system start as most boards sense the cpu fan at startup . you could plug in a case fan into the connection but how will that save power?.

  15. Ash


    This *IS* an IT website, isn't it?

    Someone's obviously in need of "Motherboards for Dummies"

  16. jeffrey
    Thumb Up

    do it in reverse

    wonder why they don't try running it in reverse, i.e put a motor where the fan is and turn it into a stirling cryocooler (heat pump), that way it should "pump" the heat off the chip and into the air.

    I hope they start making/selling one as it would look good on my northbridge

  17. Parax

    generic title

    bit dangerous if case heat builds up.... say in july, in a non aircon'd room..

    the best stirling engines are only less than 40% efficient (these are the free piston electro generators from sun power aka NASA @ Glen) the mechanical versions like this are very poor say 15% energy conversion. perhaps just a big alu fin might work better. will be interesting to see the stats..

    Now what if they used a high efficiency free piston style electo generator and used it to power a case fan to produce through flow .. that would be really cool...

  18. Edwin

    Cool but useless

    Having (at one point) built a Stirling engine, I think this is a pretty useless gadget. Cool, but still useless. I suspect it's just another passive heatpipe cooler with extra gadgetry to make it appeal to people that enjoy looking inside their PC.

    On the 70% discussion - I would guess that the marketting muppets got the technical documentation, decided 7% was a typo and wrote up 70% without consulting the engineers.

  19. Name
    Thumb Down

    Green - not

    Of course the w**kers that pay extra for this 'green' innovation will no doubt have cases with windows and internal illumination using far more power than this sterling engine 'generates'.

  20. Anonymous Coward
    Anonymous Coward

    Would this work on the vertical?

    Seeing as how most PCs are towers and the motherboard is mounted vertically, makes me wonder if orientation mattered.

    Still pretty cool... er... hot... cool... hot...

  21. joe K
    IT Angle


    @Joe K earlier

    There's something called Negative Feedback, standard to almost all typical control systems. Assuming an increasing in put causes an increasing output, and a decreasing input a decreasing output, then connecting a portion of output to input means you can stabilise the output. In summary, perturb the input and the system will stabiise itself.

    Then again, you might be a professor of control systems and were trying to make a joke, which I missed........

    And more importantly, i'm joe K with a lowercase 'j', didn't know there was another joe K on here, what are the odds....

  22. matt
    Thumb Down


    The northbridge could almost certainly be better cooled by sticking a big heat sink on it

  23. Anonymous Coward
    Anonymous Coward

    old idea too...

    from 2005..

  24. This post has been deleted by its author

  25. Nìall Tracey

    Running in perpetuity...?

    "It's not totally efficient, which is why it won't run in perpetuity,"

    Oh dear.

    Given a big enough thermal gradient between the item being cooled (the chip) and the surrounding medium (air), a Stirling engine should run indefinitely.

    This thing won't stop due to *inefficiency*, it will stop once the CPU has cooled sufficiently.

    Theoretically, if the chip is perpetually on, the engine will run in perpetuity, even though it is not a perpetual motion device.

  26. Anonymous Coward

    Stirling CPU cooler not new

    Y'all might want to check out the following book:

    It talks about a company called SunPower, which developed, among other things, a Stirling CPU cooler. Since Stirlings are also used to produce liquid helium, I think we can safely say they get REAL cold. The one they developed allowed the CPU to be overclocked 50-100%.

    No, that's NOT a typo.

  27. Stephen


    Is anyone else as fed up as me with the constant typos on thereg? Near enough all software including browsers have built in spell checking software now.

    Also a habit seems to be repeating things, e.g "then when you when you add the chip it.." (not from an actual article but you get the idea).


  28. Anonymous Coward

    @bastard, it's pretty quiet, but if you listen closely, you can hear it muttering "I think I can... I think I can..."

  29. Anonymous Coward
    Anonymous Coward

    why only a one use sterling driven fan

    why only a one use sterling driven fan , why not put a small brushless dynamo on there and generate a very limited trickle charge to a small ups or something

  30. Mark Allen
    Thumb Down


    Have MSI never looked inside the average PC? How is the dust build up on those blades and shaft going to effect it?

    Looks nice though :)

  31. Mark

    Looks like Mr. Smith misread the press release

    MSI's website and the 3 or 4 other "news" stories I read called it a chipset cooler as well, sitting on the North Bridge.

    Don't slam Tony too hard, he's reworded a lot of press releases.

  32. Chris C

    re: UGH!

    I'm with you. When I commented on it in another article a while ago, another commenter stated that getting the "information" was more important that the ability to successfully read said information. Repeated words/phrases make it very difficult and confusing to parse the sentences. That, together with the misspellings and other grammar problems (actual problems, not US/UK differences) make it painful to read many of the articles now.

  33. Anonymous Coward
    Anonymous Coward

    @ Bastard

    Reminds me of the two patent clerks: one turns to the other and says "if I hear you say 'why didn't I think of that' one more time I'm going to kill you".

  34. Adam Foxton
    Paris Hilton

    Am I looking at a different picture to everyone else?

    It occurs to me that 70% is entirely possible. Please read the whole post before arguing with it.

    The CPU puts out (for example) 100W. I know modern chips aren't doing this, but just for the sake of easy maths lets assume that they are.

    70% means that only 30% would be left to accumulate, i.e. like having a 30w heater rather than a 100w heater.

    The Stirling Engine is 7% efficient. This means that if it takes in 10W there's only 0.7W left at the output to use, rather than it takes 7% of the available power and converts it 100% efficiently to mechanical motion as some people seem to be suggesting. Every single watt of power being used to turn the fan means that the Stirling engine was sucking up 14 or so watts of heat from the heatsink. And a small 12Vdc fan can take a few watts of power to drive. Say 5watts of power is used to rotate the fan and push all that air- that's 70W of heat being sucked off the processor by the stirling engine (and dissipated by its body)

    So it's sucking up quite a bit of power on its own to turn the fan.

    The fan is blowing over a dual-heatpipe fed radiator. This will suck away even more heat than the stirling engine, what with having cold(er) air blown over it and all.

    So yes, 70% seems to me to be entirely possible _for the whole system_. Though a 70% stirling engine is clearly wrong.

    Paris because (insert obligatory "that's hot" reference)

  35. Brian Miller

    Needs tube audio section

    It just needs a tube audio section to be complete! Maybe a Jacob's Ladder, too...

  36. Adrian Lai


    A Stirling Engine is just a particular type of heat engine.

    Heat engines have a maximum theoretical efficency of 1-(T_c/T_h),

    Where T_c is the "cold" temperature, and T_h is the "hot" temperature.

    Given, say, temperatures of 20 and 60 degrees celsius, you'd get a maximum efficency value of about 12%.

    The 70% figure is presumably 70% of 12% (~ 8.5%).

  37. Steve VanSlyck

    Not Perpetual

    It's not totally efficient. No problem. But shouldn't it still run so long as power is provided (i.e., hot processor) and the parts don't wear out?

  38. Anonymous Coward
    Anonymous Coward

    re: one or two problems

    So motherboards don't start up without a cpu fan connected? How do you think watercooled set-ups work then? I haven't had a cpu fan connected in months.

  39. Ryan Stewart

    Its like a clutch fan in an auto...

    Good thinking. Its basically self regulating once its calibrated. Its just a matter of how long the piston seals last.

    Otherwise its a niftly idea. Once the underlying chip gets to a certain temp it starts driving the fan until they reach an equilibrium. Should help keep the temp consistent (which is good).

  40. Steve Browne

    I love this business!

    Where else can you continually reinvent things and be acclaimed for it.

    How about a bimetallic generator, which emits current as it heats up and use the current to drive a small fan. I actually think fans are rather passé and that a punkah could waft air over the chip, so cooling it and automatically switching off when it has cooled sufficiently to generate enough current to run it.

    Heath Robinson lives on in all of us!

  41. Pete W
    Thumb Down

    Temperature Coefficient

    Stirling engine performance is based on the temperature coefficient, in this case, chip set surface temp. vs. ambient case temp. If the case is properly ventilated, and assuming the room is reasonably cool, then this thing should operate fairly efficiently.

    What happens in situations where the case is poorly ventilated, or the room begins to get warm... Seems to me that the chip set surface temp will have to rise (and continue to operate at a higher temp.) to make this work in a warmer environment.

    Factor in the piston seals and the complex mechanical nature of converting reciprocating motion into rotary motion in a generally dusty/dirty environment...

    I'll stick with my 12VDC fans with RPM monitoring for all my PC cooling needs.

  42. James Pickett


    ..but unnecessary. My video card has heat pipes and a big passive heatsink. No noise, and Newton's law of cooling ensures that it works better as it gets hotter...

  43. Ian Rogers

    @ Efficiency

    "theoretical efficency of 1-(T_c/T_h)"

    Except that the temperatures in this equation are measured in kelvin not centigrade.

    Look up "Carnot Cycle" in google.

  44. Dave Bell

    Efficiency revisited

    Adrian, shouldn't that temperature calculation be in degrees Kelvin?

    (Which pocket did I put my slide rule in?)

  45. Adrian Lai

    @Dave Bell & Ian Rogers

    Yes, I used Kelvin for my calculation, as quick and simple use of a calculator would demonstrate. It should be obvious from sight that 1-(20/60) is not going to equal 0.12 in any case.

    However, for readability, I provided celsius equivalents.

  46. Steven Jones


    I think the point about efficiency has been lots here (and 70% is beserk figure - you'd need a huge temperature difference to make that possible). However, it doesn't need to be that efficient - just sufficient power has to be generated to turn the fan. If there's 100W of thermal energy available and 7% of that could be turned into mechanical energy then that's 7W to turn the fan.

    A CPU cooling fan I have is rated at 3W (although I have a chassis fan rated at just 2W). I suspect that the actual consumption is quite a bit lower - especially on temperature controlled cooling fans. So the energy saving isn't earth shattering, even if power supply overheads are taken into account. Just whether the energy savings will compensate for the extra used in manufacturing such a complex device is debatable (also it's quite liable to have a shorter lif if it is more complex).

    As for the overclockers, there's bad news here (not that they tend to care one iota about power consumption save their ability to extravagantly cool it). Successful overclocking depends on cooling the processor to much lower temperatures than is required for normal running. As the thermodynamic efficiency of this device is inherently related to the difference between the heatsink and the ambient air in the case then this efficiency goes down (albeit that total heat energy available increases). To give an idea, to get 7% theoretical thermodynamic efficiency the chip thermal pad temperature would have to be 20 degrees above the ambient temperature in the case at anything like room temperature. So if it's 30 degrees air temperature in the case you need that to be at least 50 degrees. If they get anywhere remotely close to the theoretical figure for a carnot cycle I'd be amazed. The carnot cycle calculations make no allowances for things like mechanical losses. It would not surprise me if you need thermal pad temperatures closed to 60 or 70 degrees to hit an overall 7% thermodynamic efficiency.

    It would be a nice little irony if the chassis cooling had to be boosted to make this little thing work. In all, I suspect this is an insignificant development and a much more elegant method of fanless cooling could be designed.

  47. Stan

    Now that... f***ing cool :) Got to get me one of them and one of those see through case things. How long before they come out with a V8 version?


  48. Gleb

    Nono, this is all wrong.

    I don't want a pansy little engine that could, I want a monster truck cooling my precious bits! And if it takes more power than the sun outputs in a year? So be it.

  49. Tim Bates

    Cools after power removed... but doesn't if the air's hot at startup.

    What would be really cool is that it would keep going for a while after the computer is turned off... Which would bring the temp of the chip down quicker than an idle heatsink.

    Of course the whole thing is pretty stupid. As has been said heaps, if the air temp inside gets too high, it won't work very well. At least electric fans move the air around in the case, so they get pretty much the best air they can. This would not move air if it got warm around it while not moving.

  50. David Sidebotham
    Jobs Halo

    April first?

    I am sure it will work but it looks like my friend Heath Robinson has been at work.

    There are a few issues:

    Noise, reciprocating even Sterling engines make noise.

    Heat, the engine needs heat transfer to make it work. If you take away the heat with a heat pipe what heat is left to operate the engine? This application is the equivalent of running up a very soft sand dune.

    I thought Sterling engines needed a kick start to get them turning.

    As I keep saying to my collegues, "Just because you can do it does not mean you should".

  51. Ishkandar

    @Steve Browne

    We, the undersigned, am having serious objections to your most learned assertion that punkahs can be adding to Mobos. It is becoming bad enough for us to be working in very, very cramped conditions and, now, you be having us work inside a PC case ??

    We will be having a pukka strike !!


  52. Ishkandar


    If your precious bits get that seriously hot, then either (1) you are not doing something right or (2) you've not got her sufficiently worked up and lubricated ...

    Suggested treatment: petroleum jelly

  53. Steve
    Thumb Down

    All this to save a couple of watts?

    All this to save a couple of watts?

    I think I'll stick with the electrically driven version - at least then the mobo can monitor (as well as control) the fan speed; better that my PC goes beep instead of bang!

    Why not pass the heat through a peltier effect pump? Some of that juice could be fed back to the supply so the PC takes less power - the Seebeck Effect (that's good for more than a couple of watts anyway).

  54. Anonymous Coward
    Black Helicopters

    I'd always wondered if a stirling engine would cool a CPU.

    Apparently not, since it's only on the north-bridge.

    I'd been tempted to try this myself with a unit from Kontax, but they cost over £100.

  55. Stylee

    @Ian Rogers

    In addition to criticising Mr Lai before checking his calculations, you refer to centigrade, rather than Celsius. The centigrade scale was abandoned in 1954 - you may not have been aware there was a difference.

    Look up centigrade Celsius in google.

This topic is closed for new posts.