thermoelectric Is it possible to power a case fan out of the GPU heat?

[Dumbo]

So here are the ingredients:

• 1 GPU with back-plate (RX 480 NITRO)
• 10 cheap peltier elements (40x40mm) 
• 1 120 mm case fan ... or maybe smaller

Preparation:

Apply the "heating" side of the peltier elements on the back-plate of the GPU with a thermal tape/paste. And then, naturally some heat-sinks on the "cooling" side of the peltier elements.

Hook up the cables and connect them to the fan.

Put the GPU to work and the thermometric elements power up the fan from the thermal transfer. Preferably the fan blows near the peltier's heat-sinks or in them so it can make a bigger thermal difference.

I am actually preparing to try this on my machine, haven't found the time to even take measurements yet (back-plate  temps most importantly) and calculate if it can actually work. But as soon as i do anything I'll share. In the  meanwhile i thought second opinion might be... discouraging and stop me from making a thermal paste mess in my case  .

 

 

[Electronics Wizardy]

You won't get enough voltage. You will need a boost converter. Also you won't get enough power unless you put the peltier unit on the die. 

[Mira Yurizaki]

2 minutes ago, Thread212 said:

I am more worried about condensation.. u may need to control your ambient temperature especially if u live in canada you know its snowing in there.

and u need at least 5V to make it performs better.

The person is wanting to do the opposite of the Petlier effect. Condensation wouldn't be an issue here.

 

Though if I were to make a comment, the system would only work in cold ambient temperatures where cooling doesn't really matter as much. You want the system to work in hotter ambient temperatures, except in hotter ambient temperatures, there's less of a temperature difference and therefore, less power to work with.

[Ykno]

just use mb headers or fan spliter on a mb header and speed fan and set it to work on the gpu heat why over complicate it?

[Dumbo]

11 hours ago, Ykno said:

just use mb headers or fan spliter on a mb header and speed fan and set it to work on the gpu heat why over complicate it?

Well... for since!

 

Seriously Im interested how it will work. The case is actually cooled very well. And I have the power to apply more fans.

>You want the system to work in hotter ambient temperatures,

 

11 hours ago, M.Yurizaki said:

You want the system to work in hotter ambient temperatures, except in hotter ambient temperatures, there's less of a temperature difference and therefore, less power to work with.

 

That is the whole point actually, I now i can pull enough for an old 60mm fan, from 2 40x40mm Peltier's if the temperature difference is around 50C.
I will try to figure out if it can power an 120mm fan with 10 Peltier elements with 20C difference, how fast would the heat-sinks disperse the heat etc.
 

And the main benefit should be: When the gpu is not under load, fans don't spin as fast or don't spin at all.

[Ykno]

8 hours ago, Dumbo said:

And the main benefit should be: When the gpu is not under load, fans don't spin as fast or don't spin at all.

Speedfan has that ability. just need to know how to config it. speedfan is powerful

 

stop trying to reinvent the wheel.

[Dumbo]

1 hour ago, Ykno said:

Speedfan has that ability. just need to know how to config it. speedfan is powerful

 

stop trying to reinvent the wheel.

I don't get why are you convincing me not to try this...

 

But I will go for it. 

 

And dont "sell" me monitoring software and cables. This is my 9th machine, I have everything a modern rig has, mb has aromatic fan control (smart mode, manual mode), case has fan controls, I had made a special separator in the case so GPU and CPU air wont mix. And I have the CAT S60 - so can thermal image the whole case, I can tell you the temps of the back-plate of the GPU...

Point being I don't need help with fixing an issue on my rig. I need opinions and help on this future project of mine. I like making stuff - I like playing with tech.

[Mira Yurizaki]

Well most cast fans are 12V 0.2A.

 

So there you go. I don't think you need 10 Peltier elements to generate enough power.

[ThomasD]

Let's rephrase your question.

 

You are seeking to capture energy coming off the CPU in order to dissipate the energy coming off the CPU.

 

See the problem now?

[Konj]

4 hours ago, M.Yurizaki said:

Well most cast fans are 12V 0.2A.

I guess he knows that, his dilemmas must be around the whole balance of the thing.

OP: This is actually pretty cool! But I think you could end up with to much amps or to little current, at least if you get similar peltiers as mine and if i remember the stats accurately.

Be sure to comeback and post some test results.

[Mira Yurizaki]

4 hours ago, ThomasD said:

Let's rephrase your question.

 

You are seeking to capture energy coming off the CPU in order to dissipate the energy coming off the CPU.

 

See the problem now?

OP is relying on the thermoelectric effect.

[Konj]

4 hours ago, ThomasD said:

Let's rephrase your question.

 

You are seeking to capture energy coming off the CPU in order to dissipate the energy coming off the CPU.

 

See the problem now?

No...

Whats the problem? Your rig needs that energy (thermal) for something else?

As he phrases it; it is excess energy. Turbo in diesel engines have the same principle only with excess pressure, uses the gas pressure from the exhaust to turn a turbine which is connected to another turbine that compresses air to go in.

[Konj]

23 hours ago, Dumbo said:

So here are the ingredients:

• 1 GPU with back-plate (RX 480 NITRO)
• 10 cheap peltier elements (40x40mm) 
• 1 120 mm case fan ... or maybe smaller

Preparation:

Apply the "heating" side of the peltier elements on the back-plate of the GPU with a thermal tape/paste. And then, naturally some heat-sinks on the "cooling" side of the peltier elements.

Hook up the cables and connect them to the fan.

Put the GPU to work and the thermometric elements power up the fan from the thermal transfer. Preferably the fan blows near the peltier's heat-sinks or in them so it can make a bigger thermal difference.

I am actually preparing to try this on my machine, haven't found the time to even take measurements yet (back-plate  temps most importantly) and calculate if it can actually work. But as soon as i do anything I'll share. In the  meanwhile i thought second opinion might be... discouraging and stop me from making a thermal paste mess in my case  .

 

 

Dude you just gave me crazier idea! 

You can make a an electric heat loop:

You do the same set up on the GPU as you mentioned. Only instead of connecting the peltiers to a fan, you take the cables outside the case and connect them to another 2-4 peltiers.
You stick the "cooling" side on something like a square tube and you install that tube just in front an intake fan.
That way you get cooler than room temp air to be taken inside by the intake fan and you keep your condensation outside the case.

If it works you might need to tip the tube a little and set up some drain system. But this can really work and might even show better results if you are in some area that has hotter periods of year.

edit:
nvm , bad idea, you will still have condensation issue with the fan at least when the cooling stops.

[ThomasD]

5 hours ago, M.Yurizaki said:

OP is relying on the thermoelectric effect.

Which is why I used the term energy.

 

Or do you think the thermoelectric effect is somehow immune to the laws of thermodynamics?

[Mira Yurizaki]

15 minutes ago, ThomasD said:

Which is why I used the term energy.

 

Or do you think the thermoelectric effect is somehow immune to the laws of thermodynamics?

OP wants to use the heat energy generated from the GPU to generate electricity via a Peltier unit. If you can cool off the other side of the unit enough, it will generate electricity.

 

So no, there's no perpetual motion machine thing going on here.

[ThomasD]

24 minutes ago, M.Yurizaki said:

OP wants to use the heat energy generated from the GPU to generate electricity via a Peltier unit. If you can cool off the other side of the unit enough, it will generate electricity.

 

So no, there's no perpetual motion machine thing going on here.

Well sure it might be possible.  But the problem remains.  The cumulative inefficiencies will render it entirely ineffectual even if it does function.  In order to get any appreciable fan performance you would likely need an uncomfortably high case temperature.

 

Which only begs the question of why bother?

 

Proof of concept is great if the concept makes sense...

[Mira Yurizaki]

50 minutes ago, ThomasD said:

Well sure it might be possible.  But the problem remains.  The cumulative inefficiencies will render it entirely ineffectual even if it does function.  In order to get any appreciable fan performance you would likely need an uncomfortably high case temperature.

Thermoelectric effect works better when there's a greater temperature difference, hence a lower case temperature is better.

50 minutes ago, ThomasD said:

Which only begs the question of why bother?

Because OP specifically asked to not convince them otherwise, even though there's plenty of fan control utilities.

[Stefan1024]

On 12/19/2016 at 10:59 PM, Dumbo said:

So here are the ingredients:

• 1 GPU with back-plate (RX 480 NITRO)
• 10 cheap peltier elements (40x40mm) 
• 1 120 mm case fan ... or maybe smaller

Preparation:

Apply the "heating" side of the peltier elements on the back-plate of the GPU with a thermal tape/paste. And then, naturally some heat-sinks on the "cooling" side of the peltier elements.

Hook up the cables and connect them to the fan.

Put the GPU to work and the thermometric elements power up the fan from the thermal transfer. Preferably the fan blows near the peltier's heat-sinks or in them so it can make a bigger thermal difference.

I am actually preparing to try this on my machine, haven't found the time to even take measurements yet (back-plate  temps most importantly) and calculate if it can actually work. But as soon as i do anything I'll share. In the  meanwhile i thought second opinion might be... discouraging and stop me from making a thermal paste mess in my case  .

 

 

I did this in my "silent cube" build and I recovered about 100mW, but with a boost converter and a super capacitor as a jumpstart I was able to push it to 5 volts and charge my smartphone.

However, the charge rate was very low and after about 10 minutes the cold side overheated and the temperature difference dropped a lot and the energy output was basicly zero. But in theory it works.

The effiency of the peltiers as a generator is about 1 to 5 % and you are not going to extract more than 5 to 15 watts from the backside of the GPU, so don't expect doing more than ligthing up a low power LED.

[mark_cameron]

On 19/12/2016 at 9:59 PM, Dumbo said:

So here are the ingredients:

• 1 GPU with back-plate (RX 480 NITRO)
• 10 cheap peltier elements (40x40mm) 
• 1 120 mm case fan ... or maybe smaller

Preparation:

Apply the "heating" side of the peltier elements on the back-plate of the GPU with a thermal tape/paste. And then, naturally some heat-sinks on the "cooling" side of the peltier elements.

Hook up the cables and connect them to the fan.

Put the GPU to work and the thermometric elements power up the fan from the thermal transfer. Preferably the fan blows near the peltier's heat-sinks or in them so it can make a bigger thermal difference.

I am actually preparing to try this on my machine, haven't found the time to even take measurements yet (back-plate  temps most importantly) and calculate if it can actually work. But as soon as i do anything I'll share. In the  meanwhile i thought second opinion might be... discouraging and stop me from making a thermal paste mess in my case  .

 

 

You'd need some sort of means of converting heat to electricity.

 

This would potentially increase the PC's energy efficiency ....

 

I think for any DIY solution you'd have to pump the hot air in thermally lagged/insulated pipes to an external unit to properly get any real benefits. Probably combined with excess heat from a fridge or similiar appliance that is creating heat.  Plus scaled up phase transfer type system.

 

On 20/12/2016 at 4:10 PM, ThomasD said:

Let's rephrase your question.

 

You are seeking to capture energy coming off the CPU in order to dissipate the energy coming off the CPU.

 

See the problem now?

It's no different to 'combined cycle' energy generation units it's about recovering heat energy in order to recover electricity that was lost as heat. Basically to improve efficiency.

 

In theory it's an excellent idea especially given how much power a multi GPU system uses these days. Extracting the heat quickly to recover the heat would make PC running costs decline substantially. Mine costs GB£400 to GB£500 and comes second to my actual economy7 electricity heating costs!

 

The main issue is the amount of heat recovered. As you are losing heat to the room by radiation and convection.

 

However the idea is sound. Application more difficult.

[mark_cameron]

Basically I'm saying. Trying to do this inside the case will be a waste of time and create more issues. Designing the case to funnel all heat to another unit to recover energy would be the way to go I think.

[Cyracus]

Enjoy your science experiment, remember to bring us pictures if anything explodes, or a video if it works. either way fun stuff