CPU heatsink mod, stacking heatsinks with Peltier

[tomaatvk]

Just like with my case build (link in my signature), I do this because I can and am bored, not because this is the most efficient way to dissipate heat.

 

Lately I've played a bit with my CPU (i7 3770k) in terms of cooling and switching between passive through downclock (3.8GHz) and active through some fans after an overclock (4.8GHz) but I am still hitting a thermal wall at some point. Around half a year ago I did some school experiments with Peltier-elements (electrical heatpumps) and I still have all the pieces lying around in my room somewhere. This gave me an idea as to how I could possibly dissipate even more heat from my CPU. However, I don't want my cpu to burn up if the element dies (they are isolators so the CPU won't get rid of its heat if the element is mounted directly on it) so I have to build a buffer system so it'll still get cooled if it dies.

 

Currently I have my computer setup like a testbench so the only things that limit my working space around the CPU will be the RAM, my GPU and the CPU power cable. The CPU power cable is far enough out of the way to not have any effect and the RAM is low enough to not bother me (valueRAM FTW). As for the GPU I will probably need to buy a 20cm PCI-e x16 riser cable for around €30 to move it away from the CPU. This should give me enough space to almost mount a big, flat heatsink on the CPU.

 

My current plan is to bend open my Coolink Corator DS LGA2011 (got it for the experiments). Directly on the CPU I will mount a Streacom HT4 (I have yet to order this for ~€25) and on the top of the Streacom I will mount my currently used Deepcool Gamer Storm Lucifer. In the middle of the Streacom (so on the gray part of the CPU mounted plate) I will mount a 12715 Peltier-element that I will either be off, run at 5v or at 12v (depending on my CPU temperature with some relays). On top of that element I am going to put the bent open Corator DS with the cooling fins alongside the top part of the Streacom and the lucifer.

 

Coolink Corator DS LGA2011

 

Deepcool Gamer Storm Lucifer

 

Streacom HT4

 

12715 Peltier-element

 

 

I will also have no other options but to just latch everything together with zip-ties and hope it stays in the correct position.

 

Before I do any of this, does anyone have any idea if this will work, like at all or will my temps get worse than with only the Gamer Storm Lucifer?

 

Power usage is not my concern and neither is the extra heat kicked into my room.

[TheRandomness]

In theory it should work... kinda did when I stacked 3 heatsinks without a peltier device  bending heatpipes is rather hard I must say...

[KemoKa]

Honestly, I have no experience with this. I've heard stories about this working, but you're gonna need to pump a lot of power through that pad to cool the CPU down enough.

[vorticalbox]

so you're putting it on the CPU and using the heat sink it get rid of the plates heat?

[tomaatvk]

In theory it should work... kinda did when I stacked 3 heatsinks without a peltier device  bending heatpipes is rather hard I must say...

 

I know bending heatpipes is rather hard, with the kinking and closing they can be a b*tch but I only need to bend them 45° and not really in a short corner I guess that will work and if not who cares, I'm not using Corator DS anymore anyways.

 

About just stacking heatsinks, are you serious, does just stacking give better performance? Makes sense I guess but is it noticable?

 

 

Honestly, I have no experience with this. I've heard stories about this working, but you're gonna need to pump a lot of power through that pad to cool the CPU down enough.

 

I don't think it actually needs to pump a lot of heat, any heat pumped through there will already make a difference compared to not having it at all (at least I think, idk) because it doesn't have to be dissipated through the main CPU cooler.

[Stefan1024]

The plan sound unteresting, but your peltier is onderpowered.

The overclocked CPU can kick out ~140 watts, so you need at least a 200 watt peltier element.

Also keep in mind with your cobfiguration you will hardly get sub ambient temperature.

[tomaatvk]

so you're putting it on the CPU and using the heat sink it get rid of the plates heat?

 

I am putting the Streacom HT4 on the CPU and one side of it will be standardly cooled by the Lucifer and the side directly on the CPU will get some heat pulled out by the Peltier and the Corator DS heatsink will be there to get rid of the heat generated by the element and of the heat pumped by the element from the CPU.

[TheRandomness]

I know bending heatpipes is rather hard, with the kinking and closing they can be a b*tch but I only need to bend them 45° and not really in a short corner I guess that will work and if not who cares, I'm not using Corator DS anymore anyways.

 

About just stacking heatsinks, are you serious, does just stacking give better performance? Makes sense I guess but is it noticable?

 

 

 

I don't think it actually needs to pump a lot of heat, any heat pumped through there will already make a difference compared to not having it at all (at least I think, idk) because it doesn't have to be dissipated through the main CPU cooler.

It sortof worked, it was better than the stock cooler (It was 3 stock coolers, one on top of each other )

[tomaatvk]

The plan sound unteresting, but your peltier is onderpowered.

The overclocked CPU can kick out ~140 watts, so you need at least a 200 watt peltier element.

Also keep in mind with your cobfiguration you will hardly get sub ambient temperature.

 

First of all, I'm not aiming for sub ambient, this is one of the other reasons I want a buffer heatsink and that it won't always run on full power, I don't want any condensation to start forming at all. Secondly, the Peltier is only there to lower the temperatures a little more, the system doesn't rely on the Peltier for it's cooling so I don't think it'll have to provide enough cooling for the whole CPU, any heat left should be dissipated by the other cooler (I think).

[iamdarkyoshi]

I would suggest trying it out on an older machine. Test the death of the peltier unit by unplugging it. See if the system survives

[tomaatvk]

I would suggest trying it out on an older machine. Test the death of the peltier unit by unplugging it. See if the system survives

 

I have 2 older machines, one broken Pentium 4 pc (the CPU still turns on but only generates about 40 watts of heat) and an old, crappy laptop that is just old and crappy, doesn't kick out any heat. Testing it on an older machine won't really work I think, I guess I'll have to take it carefully with my 3770k starting from a really big downclock of which I'm sure it can handle even with an awful heatsink (so 1.6GHz with like 0 vcore).

 

 

It sortof worked, it was better than the stock cooler (It was 3 stock coolers, one on top of each other )

 

Currently testing that theory, I just put my Corator DS on top of my gamer storm Lucifer (testbenchish setup so it stands stable) and the max load temp without the Corator is 94°C and with the Corator DS 93°C, seems confirmed enough to me This is without thermal paste inbetween the 2 heatsinks and no mounting pressure at all so it could, when mounted proppely drop to 92°C.

[Stefan1024]

First of all, I'm not aiming for sub ambient, this is one of the other reasons I want a buffer heatsink and that it won't always run on full power, I don't want any condensation to start forming at all. Secondly, the Peltier is only there to lower the temperatures a little more, the system doesn't rely on the Peltier for it's cooling so I don't think it'll have to provide enough cooling for the whole CPU, any heat left should be dissipated by the other cooler (I think).

This may actually work somehow.

But keep in minde every contact surface will add some thermal resistance. So to bee as good as mounting the cooler directly, the peltier needs to do something.

What about just skipping the peltier and use two coolers as you show in the sketch?

It is very important to have as less contacts as possible as every one adds a lot thermal resistance. But when the surfaces are plane and large it's not too bad. But it's still chalangeing. This nearly caused my CPU to overheat. Look in my signature for more information.

[tomaatvk]

This may actually work somehow.

But keep in minde every contact surface will add some thermal resistance. So to bee as good as mounting the cooler directly, the peltier needs to do something.

What about just skipping the peltier and use two coolers as you show in the sketch?

It is very important to have as less contacts as possible as every one adds a lot thermal resistance. But when the surfaces are plane and large it's not too bad. But it's still chalangeing. This nearly caused my CPU to overheat. Look in my signature for more information.

 

About skipping the Peltier and just mounting the 2 coolers like in the sketch, I think that will be worse than what I have now because of the increased amount of contact surfaces, I will need something to compensate for that and that should be the Peltier.

 

Also, I'm planning on sanding all the slightly rough surfaces out as much as possible to get better contact but you never know.

 

I have no idea of how well it will work, I guess it's just a matter of finding out.

 

Ps. I read the silent cube build log already.

[Stefan1024]

About skipping the Peltier and just mounting the 2 coolers like in the sketch, I think that will be worse than what I have now because of the increased amount of contact surfaces, I will need something to compensate for that and that should be the Peltier.

Also, I'm planning on sanding all the slightly rough surfaces out as much as possible to get better contact but you never know.

I have no idea of how well it will work, I guess it's just a matter of finding out.

Ps. I read the silent cube build log already.

Ok, I don't know which surfaces you like to sand.

The coolers are fine and the heat hipies on the thermal riser should not be sanded.

But just try and please show off the results. I'm actually pretty interested in the results.

[AnonymousGuy]

Tried it.  It'll work at idle but any load and you'll outrun the peltier and then overheat.

[tomaatvk]

Tried it.  It'll work at idle but any load and you'll outrun the peltier and then overheat.

 

Did you try Peltier elements directly mounted on the CPU or the exact system I'm designing which is pretty much the same concept of the cooler master v10 which worked for sure but then just more extreme.

 

 

Ok, I don't know which surfaces you like to sand.

The coolers are fine and the heat hipies on the thermal riser should not be sanded.

But just try and please show off the results. I'm actually pretty interested in the results.

 

I was gonna sand the rough surfaces that I would create by clamping the Corator to bend it.

 

As for the results, my funds are kinda low so I can't directly build anything yet, however, I am currently working on a 'simplified' model of the whole thing to get results without actually doing anything. It currently has around 30 variables/formulas in it already and I'm not even halfway done. It's simplified in the way that I don't have every single piece of data, however, I can make an educated guess wether it'll work or not and the results will probably be within 10°C of the actual test (if I were to do it).

[tomaatvk]

But just try and please show off the results. I'm actually pretty interested in the results.

 

I would suggest trying it out on an older machine. Test the death of the peltier unit by unplugging it. See if the system survives

 

Like I said I started modeling in coach 6 (some physics modeling software, I don't advise using it). I just finished my model of how it would be with the CPU, the Streacom HT4 and the Gamer Storm Lucifer (with MX-4 thermal paste inbetween the things). What it doesn't include: the peltier-element, the Corator DS (so it's basicly like the peltier is dead), heat impendance of the Gamer Storm Lucifer, some heat transfer contact points, heat impendance of the heatpipes of the Streacom HT4, some more, low impact stuff and the only heat loss is through the Gamer Storm Lucifer. The results without the peltier are only like 5°C worse than with the Gamer Storm Lucifer mounted directly on the CPU (with MX-4 ofc) and stay at 97-98°C. I did do a seperate model of only the Gamer Storm Lucifer with the same data to check if the results resembled the real world and yes they do, I only got a 1°C difference between real and theoretical.

 

Blue: CPU temp in °C

Green: HT4 bottom part temp in °C

Red: HT4 top part temp in °C

Purple: watts of heat being added to the CPU

Gray: watts of heat being added to the HT4 bottom part

Other type of green: watts of heat being added to the HT4 top part

Pink: watts of heat escaping into the air through the lucifer heatsink

At the start the CPU will generate 130 watts of heat and after 1000s it will output 0 watts.

 

 

Within a few days I think I can have the Peltier with the other heatsink modeled in there and get a working model of it.

 

This current model was achieved after about 3 tries, had some fundamental flaws in the first 2 resulting in either nuclear fusion or temperatures of 0K.

[tomaatvk]

The results are in:

Spoiler

 

This is made in a simplified model with a lot of variables left out and with a cpu heat output of 140 watts. For example, the heat transfer loss through the HT4 mounting blocks has been neglected but that should at max make 2°C of a difference which would still give clear results. Also difference in air flow has been left out and so has any heat tranfering to the air through anything but the heatsinks. Here is a picture of the model so you can get an idea of what it already became.

 

Spoiler

 

Conclusion, I might use this system but without the Peltier-elements, so just with stacked heatsinks, my model might not be accurate but it does show that it will have some damn nice results, now I just need to find a spare €30 and order the Streacom HT4.

 

Just a tag for @Stefan1024