Slashdot Mirror
MSI Develops a Heat-Driven Cooler
V!NCENT tips us to a write-up about an addition to MSI's Ecolution motherboard which harvests heat from the chipset to power a fan. The device is based on a Stirling engine. The heat from the chipset expands a trapped gas, which pushes against a piston to generate power. The article contains a YouTube video of how the device works. According to MSI, the device has 70% efficiency.
otherwise all that waste heat would be wasted.
MSI just threw this together so that their lead engineer could finish his bitchin' Steampunk case mod.
It has to heat itself to ... cool ... itself?
Goddamnit, I hate recursion.
Because I thought to get 70% efficiency there would have to be a couple of thousand degrees C difference between the hot and cold sides. Or have AMD decided laptops are not their core market for the next generation of chips?
Deleted
It's a fucking fan, not a cooler. In this house we obey the laws of thermodynamics.
A fan can't draw much more than a few watts. What's the point? It seems like a complicated array of technology just to save a few watts of power. You'd be better off buying a more efficient power supply if you wanted to be "green".
AccountKiller
So, the Sterling engine runs on the temperature gradient between the chip and the ambient environment. It uses this energy to...do what...increase the gradient some more? By pulling in cooler air from outside the case I guess?
Seems like it would work best when it's needed least, and vice-versa.
Yo dawg, I heard you like the Ackermann function, so OH GOD OH GOD OH GOD
It has to be hot for the fan to run, but the fan makes it cool, so.... huh? I can't seem to wrap my brain around this one.
The moving part is cute, of course, and gives a bit of visual
tension to the apparatus you see through your peekaboo case.
Still, it's a bit of a clunker compared to the old-tech way of
making a no-moving-parts air pump powered by waste
heat. I refer, of course, to the 'chimney'.
But even better would be if the energy loss could be decreased in the first place. Heat produced by a computer is actually only annoying.
The Stirling engine was invented by Reverend Dr. Robert Stirling.
If builders built buildings the way programmers wrote programs, then the first woodpecker would destroy civilization.
How many times have I heard rather short-sighted types say things like the Chinese/Japanese/Indians/whoever only copy us?
This is something we should have been looking at years ago - but no doubt someone other than we Americans will get this technology's price point down to practicality before any of our fearless CEOs get up off the R&D bucks that they fear will come straight outta their quarter to a half billion dollar annual compensation packages...
Orwell: "In a Time of Universal Deceit, telling the Truth is a Revolutionary Act"
...to appreciate how FREAKING AWESOME this idea is; I have to say, I'd buy one.
:D Oh, and show case it, of course!
Since I don't actively build PCs anymore (though this is cool enough to build a machine just to use it) I'd have to mount it on my bald head to cool my crazy off in the summer
Am I right on thinking this is the first viable, potentially wide-spread use of a Sterling Engine?!
PS: I don't reply to ACs.
Especially if it only uses waste heat to drive itself.
How much waste heat can they get from a modern power-efficient CPU? Let's see the thermal dissipation:
AMD Athlon x2 BE2300 or Inten Penryn. Both at about a few Watts at idle, and 60 (AMD)-90 (Penryn) Watts under load - so average let's say is 30W, assuming a box idles more.
30Wx70% = 21W for a fan. That's PLENTY for moving a fan - if the CPU is doing work.
However, at idle, you may only get 4 Watts if you're at 70%. However the fan speeds don't necessarily drop by that much in a normal computer that you see. Probably due to an engineered safety margin, but the fan is not getting a lot.
So unless the heat charges a battery and the fan is drawing from that battery, they may not be able to produce enough fan speed at idle.
And of course, using a battery lowers your total efficiency to around 40%. Again, even if you cut the 21W to 13W, it's still plenty to drive a fan. So the question is, how they're going to use the excess energy to charge a battery to use when the CPU is idle.
Oops I released the magic smoke...
Why is it that they found the use of this on an Nvidia chipset "amusing?"
I'd rather spend the extra watt on a constant fan that risk damaging the CPU in cooling/heating/cooling/heating cycles.
Peltier-based CPU coolers are quite available - see for example this amusing misuse of manufacturing resources.
I bet it sells by the truckload if it's UV reactive!
Seriously, it's just another toy for people with too much money. Anyone concerned about power usage should just get a slightly less powerful chip.
Besides, the sterling engine is the wrong direction for regenerative power, and the extra complexity of the unit probably just impedes airflow, making the chip run hotter or the fan work harder. That hardly improves efficiency.
This cooler is on the northbridge, not the CPU. One would assume the northbridge to have more stable loading conditions. However it's a heckuva lot better to use a larger passive cooler whenever possible, not add more moving parts in addition to the fan. Those dinky motherboard fans are usually the first to die.
Peltier modules are nowhere near as efficient as sterling engines. Using a peltier module, you would be lucky to get enough power to light a small LED from the typical chipset to atmosphere temperature differential. They work fine as heat pumps since you've already got a big sink strapped to the hot side, but when you start trying to use them the other way around - to generate power from a temperature differential - their inefficiency shows through.
such as "desktop" CPUs. Why are these still being produced, when the "mobile" variants of the same models are much more efficient? For example, look at these two:
http://processorfinder.intel.com/details.aspx?sSpec=SLA98
http://processorfinder.intel.com/details.aspx?sSpec=SLA49
Both of them are Core 2 Duos, 65 nm process, 2 GHz, 2 MB cache. But one of them is a "desktop" model, and I wonder what the hell it's doing to waste almost double the power of the "mobile" one.
Escher was the first MC and Giger invented the HR department.
Normal fans are prone to outages. This system may be a nice back up on that. As to how green it is, well, they would be better off getting all OSs to push systems to sleep.
I prefer the "u" in honour as it seems to be missing these days.
the efficiency may be only about 10% or so. No violation of the second law of thermodynamics - just move along people.
..........FULL STOP.
as I understand it, single-cylinder Stirling engines are not self-starting. I wonder how MSI gets around that.
I also notice that the heat-pipe going to the radiator is on the hot side of the Stirling engine. So as the heat is dissipated by the radiator, the hot side of the engine cools down, which causes the engine to slow down. I know eventually it will reach some equilibrium speed, but I wonder how hot the CPU must be to power the engine. Maybe it would be better to put the heat-pipe to the radiator on the cool side of the engine?
There is also about a $100 price difference between those two chips. I imagine they are manufactured to different quality standards. The mobile chip probably has less leakage or something to that effect.
Why are these still being produced, when the "mobile" variants of the same models are much more efficient?
The T7250 appears to be about twice the price of the E4400.
I can get an E4400 for 129.00. Best I could find on a T7250 was 262.50
You're absolutely correct that an ideal Carnot engine would have to have about a thousand degrees if it rejects heat to room temperature.
Typically what's done in these cases is to compare the efficiency of the engine to the Carnot efficiency. So the claim of 70% efficiency really means that the engine is 70% as efficient as a Carnot engine at the temperatures it operates between. The real efficiency then is n_carnot*n_engine. Their real efficiency claim is therefore probably closer to 4.9%.
But that's not the only convention. Sometimes the comparison is made to the ideal version of whatever cycle they're using. For a stirling engine, I believe the ideal still approaches Carnot efficiency, so that wouldn't affect their claims, but you can see how some shady math can be used to get people excited.
Can you be Even More Awesome?!
A Carnot engine is a theoretical device, a heat engine,
with the uniqe property of having the maximum
theoretical heat effeciency a heat engine can have.
(The carnot engine is very impractical.)
No heat engine operating between two heat resouars
with temperatures T and t, t less than T, can have an effeciency
of more than e = 1 - t/T. (t and T are in kelvin)
For an engine operating at 70 % effeciency,
e = 0.7, so
T = t/(1 - 0.7). If the cooler works at room temperature,
we can probably set t = 20 degrees centigrade = 293 kelvin.
Therefore, T = 293/0.3 = 977 kelvin ~ 700 degrees celcius ~ 1300 degrees fahrenheit.
Solder melts at about 200 degrees celcius.
TFA boasts: "MSI 'Air Power Cooler', more then just saving watts!"; are they trying to appeal to the leet, spelling impaired internet people?
Other than (sorry, then) that, isn't the fan really small? It's kind of strange to see two heatsinks and such a small fan together. But the fact that the Stirling engine is both draining heat from the chip and using the power for the fan is kinda cool. Also that means the fan's speed is autoregulated: no temperature difference, no fan noise.
Anyways, it's for the motherboard chipset; those usually don't have a fan at all!
Don't take my posts literally; it's just code to control my botnet.
It's not that clear from the article how they are defining efficiency, but if they are claiming that they are converting 70% of heat energy into air flow energy, simple thermodynamics can tell you that that is bullshit. If you assume an ambient temperature of 300K, then you need a temperature of 1000K MINIMUM to achieve 70% energy conversion. The Carnot efficiency is 1-Tc/Th, which is the best efficiency possible for a heat engine. Typical heat engines are a fraction of that efficiency.
My only concern with this type of setup is what type of liquid/gas they'll be using inside the chamber to move the piston and also what the heat sink will be made out of. If the liquid/gas in the chamber's boiling point is too high the gas will start to expand _way_ too late.Not to mention the fact that depending on the gas they use you'll probably have to buy a new one every so often since the gas will escape if it is less dense than the heat sink material. Granted the solidworks youtube video is nice, i'm still not sure i'd buy this heatsink for my chipset due to the complexity of the chemistry involved.
BSD is for people who love Unix, Linux is for people who hate Microsoft.
...between the CPU and the outside world, why not just use a Peltier device in reverse to power a conventional electric fan?
Maybe it's just me, but I was under the impression that you couldn't move any more heat by using waste heat for power than you could by letting the waste heat move passively across the gradient in the first place.
Good point. But I'm still seeing problems. "Desktop" and "mobile" CPUs use different sockets, so you can't just invest in a "mobile" CPU to save power in your "desktop" motherboard. Common sense and standards would unify the two sockets, right? Just like you can invest in a fluorescent bulb to save energy in the long run, as it fits in the same socket as an incandescent bulb. Legislation in some countries is heading towards the banning of incandescent bulbs, so why not ban these space-heater CPUs?
Incidentally, I've been using Mini-ITX mobos with "mobile" CPUs for some time now, and I wish they would be more generally available, with global warming and all that. Funnily enough, it looks like "desktop" CPUs are the new Celerons; I'm thinking of the Celeron M that's cheaper than Pentium M but lacks some of the power saving features.
Escher was the first MC and Giger invented the HR department.
How does the piston ever retract, if the thermal gradient is putting constant pressure on it ?
I want to delete my account but Slashdot doesn't allow it.
...stands for Ecolution can only leverage unrealistically tailored internal operating nature.
(I hope not. I just wanted to give LordKaT extra recursion nightmares for fun.)
You can hold down the "B" button for continuous firing.
If you want a low power GPU and are willing to accept lower performance, then what's the problem with integrated graphics?
The hotter the chip gets, the faster the fan goes. Simple and elegant.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
So, for anyone who knows about low temperature difference stirling engines, they know that one cylinder engines are not self starting. This looks like a one cylinder engine, so you'd probably have to open up your computer and give it a spin to make sure it starts.
I wonder how much power this actually dissipates. Most recent desktop processors I've seen need at least a 80mm fan running pretty fast or even a 10cm fan and shrouding. This is a gimmick. A cute gimmick but a gimmick nonetheless.
Back at the turn of the century there were oil burning stirling powered fans. There is a company that makes them now, but these are inherently sensitive machines to work on a low temperature difference. This translates to expensive parts... Most are about $100~200, though you might be able to do it for less, it won't touch the price of a cpu fan.
+++ ATH0 +++
I speculate based on my knowledge of how chip fabrication works, that if they banned the "high power" versions of a chip, the price of all chips everywhere would skyrocket.
The most efficient running chips at a clock get cut first, less second, ones with defective cache marked down to lower versions and sold like that. If you took out all the ones that work but take more power doing it, you suddenly have a much smaller yield -- meaning a much higher price per unit.
you should know.
but I do like the idea. Current thinking is 'there's too much heat in your box, therefore add another (heat generating) electrical component to help cool it. There's something nice about 'turning the heat on itself' - even ignoring the cooling effect of the fan, the stirling engine itself is using some of that excess energy just to power itself (and therefore cooling even without the fan).
I read the title as "Microsoft Installer Develops a Heat-Driven Cooler". Was I only one?
I think a PSU-powered fan would be very much necessary still. It's unclear what they're measuring when they say "efficient" but if this means it can't remove 100% of the heat that exists, then it's only going to slow the rate of warming, not eliminate, stabilise, or control it.
What an absolutely wonderful idea. Let me see if I have any spare genetic material lying around...
Logi - I can do anything, but not everything.
It's good for Microsoft to invent such a machine. Now PCs may run Vista, at last. ...wait a minute...
It seems to me that, as your CPU heats, you get more power. Your fan goes faster, and as it cools your chip, the power output drops and your fan spins down.
Fully automated fan control, and as little noise as possible.
What a depressingly stupid machine.
A CPU manufacturer always needs to have a faster CPU than their competition. So it's speed first, efficiency later. The latest and greatest CPU will be fabbed in large amounts, so it cuts the costs. These large amounts of pwnage CPUs all need to be sold, and selling them is easy because the best CPU now can still be sold as mediocre CPU later. When you fab a mediocre efficient CPU you can only sell it at a later time as garbage CPU so this is risky. Not to mention you cannot sell the once-it-was-the-best-CPU anymore because there are also efficient ones now.
Here be signatures