Why did'nt I think of this

LINK

what a neat idea if it works :twothumbs

It's a neat sales gimmick, but not really anything practical. Yes, a stirling engine is a way to convert heat to work and using one to drive a CPU cooling fan might save a miniscule amount of electricity. But it will be at the expense of the effectiveness of the cooling. To generate the mechanical force to drive the fan, the stirling engine has to let the heat build up until enough force is generated to move the fan. This defeats the purpose of a CPU cooler, to remove excess heat from the chip. It might work, i.e. run, since CPUs can tolerate some heat buildup, it's still just an overly complicated gimmick.

I'm wondering if some sort of fluid would be more efficient than air...

It doesn't take that much heat to get a Stirling engine going. You can set one over a cup of coffee and it'll start going. You can even just set one in the palm of your hand, and the temperature differential between your hand and the ambient conditions is enough.

Also, it looks as if the fan is blowing over a liquid filled radiator.


Edit:
http://video.google.com/videosearch?q=low+temperature+stirling+engine&hl=en&sitesearch=

Just what I need. A computer that goes thumper - thumper - thumper.

Now to figure a way to mount one of them on my flashlights once winter is over.

DonShock said:

It's a neat sales gimmick, but not really anything practical. Yes, a stirling engine is a way to convert heat to work and using one to drive a CPU cooling fan might save a miniscule amount of electricity. But it will be at the expense of the effectiveness of the cooling. To generate the mechanical force to drive the fan, the stirling engine has to let the heat build up until enough force is generated to move the fan. This defeats the purpose of a CPU cooler, to remove excess heat from the chip. It might work, i.e. run, since CPUs can tolerate some heat buildup, it's still just an overly complicated gimmick.

Click to expand...

Agreed.

A chipset cooler fan draws maybe 1-2W total. On a system that's using 100+W, such savings are miniscule, and don't justify making a simple system unnecessarily complex. Plus, I would imagine that the tiny Stirling engine wouldn't last NEARLY as long as a well designed chipset cooler, and would be WAY more expensive to boot.

A better solution would be to use a larger heat sink with more surface area. It would cost less, require less maintenance, and provide better cooling.

MSI's claims of 70% efficiency are dubious and misleading. Someone needs a basic thermo lesson here. The carnot heat engine is the most efficient heat engine possible. Its theoretical efficiency is 1-Tc/Th (not counting friction and thermal losses) [1]. If you let the hot side get to 60C, and keep the cool side at 40C (pretty typical chip/case temperatures), that translates to 333K and 313K respectively, so the maximum possible efficiency for any heat engine between those two temperature ranges is 1-313/333 = 6%. Again, this is for a Carnot engine. The efficiency of a Stirling engine is lower than a Carnot engine, and more difficult to calculate [2].

In fact, what they probably MEANT to say was that it achieves 70% of the theoretical maximum efficiency that a Stirling engine can achieve. That's a much more realistic thing to come to terms with.

Let's take an example: the Intel P35 series chipset. From here, http://download.intel.com/design/chipsets/designex/31696803.pdf page 15, the TDP of these devices is between 13 and 16W. Let's be generous and say the stirling engine can convert 5% of the heat transferred across it to mechanical power - note that the Stirling engine isn't even transferring all that heat, the majority is being transferred by the heatpipes below it to the heatpipe radiator - only a couple W are actually being transferred by the Stirling engine. So we're only looking at maybe 100mW total. That's not going to move a whole lot of air.

Finally, a stirling engine needs a temperature differential in order to operate. Since the point of a heat sink is to cool something as much as possible, requiring a larger temp. differential than necessary, by design, seems silly at best. This is along the same lines as people thinking you can use a peltier or other thermoelectric junction to get extra power from the heat produced by an LED to help power said LED. Sure you can, but you're going to run the LED near its thermal maximums in order to create a large enough delta-T to get any meaningful amount of power. At that point, you've caused the efficiency of the LED to drop more than the power you will recover via the peltier. You'd get more light just by running the LED cooler in the first place. Remember, TNSTAAFL.

So, in summary, if you really want to eliminate the chipset fan, you'd really be better off slapping a big passive heat sink on the chipset than going this route. Seems like an awful lot of work to save a measly Watt or two. It's as effective as replacing the bulb in your fridge with LEDs to reduce your power bill.


[1,2] Prentice Hall Physics, 2nd Edition, pg 540, 543.