TRUE TEC Module (with pics)

[scifikg]

As Vapor said, each heatpipe has a load for it, most around here are 35W or 50W. Past that they become quite useless.

OK, Lets say each heatpipe in the TRUE can move 40W just to be conservative. 40W x 6 = 240W. Can you honestly tell me there is a heatink without heatpipes that can move that much heat?

I'm curious does anyone know how much heat the TRUE can move?

[n00b 0f l337]

True has 35W I believe.
And any heatsink that has just metal, like the old switch peltier heatsinks, do not have a heat maximum. They will transfer heat through metal up to the fins or pins, and then you can cool those. There is no max.

[systemviper]

ahhhh so is that why most of the old production models that are around use large hunkers of metal with fins to cool the tec's. I always wondered why that was all they did and didn't figure out another way..

plus you say the true ceiling is about 35w per heatpipe.

[n00b 0f l337]

Bingo, exactly, right on, and you got it. Though a thinner hot side plate would definitly have helped those old units.

[scifikg]

True has 35W I believe.
And any heatsink that has just metal, like the old switch peltier heatsinks, do not have a heat maximum. They will transfer heat through metal up to the fins or pins, and then you can cool those. There is no max.

The problem with that is you can only fit so many fins or pins over a heat source. The great thing about heat pipes is that they have great thermal conductivity. With heatpipes, you can add many more fins or pins and therefore get much better cooling performance. Even though heatpipes have a maximum, TRUE 6x35W=210W, a heatpipe cooler will move way more heat than a heatsink without heatpipes. A heatsink with no heatpipes may have no max, but if it can't move the heat as well as a modern heatpipe heatsink, what is the point?

Also, isn't probable that old heat sinks didn't use heat pipes because nobody was making them to cool electronics back then. I don't think that I saw a heatpipe in a PC till the late 90s

[n00b 0f l337]

Way before then, heatpipes have been used for a long time. But they have a max and in this case he is exceeding said max.

[hollo]

Thanks, but lets be reasonable. Is it really fair to compare my cold plate to data from that review article? I can see some really important differences between them

1. That data was made with a single heat source but my cold plate has two.
2. The heat transfer coefficients are at least an order of magnitude to high.
3. 300W is too much for an e8400
4. The 8x8cm shape and the cold plate shape are very different

I think heat spreading is just way too complex to to declare that my cold plate is too thin based on that data.

well, when i looked at it the cold-plate screamed "too thin" (mostly because of the metal taken off the edges, and because copper just isn't such a great material for heat-transfer over long distances), i wasn't gonna say but i remembered that link for comparison so i thought i'd throw it in
but regarding comparison with the link-
1 the cold-plate has one heatsource (i wasn't talking about your hotplate)
2&3 for any given line the graph shows the relative effects of vertical transfer deteriorating with thickness, while horizontal heat transfer improves. i'm pretty sure any constants used for heat-load or transfer coefficient would show the same scaling with plate thickness
4 your cold-plate is narrower but longer than a 8x8cm plate - the linear heat transfer between the CPU and the ends of the peltiers would be sub-par at 12mm thick, and it's not even that thick around the edges

[scifikg]

well, when i looked at it the cold-plate screamed "too thin" (mostly because of the metal taken off the edges, and because copper just isn't such a great material for heat-transfer over long distances), i wasn't gonna say but i remembered that link for comparison so i thought i'd throw it in
but regarding comparison with the link-
1 the cold-plate has one heatsource (i wasn't talking about your hotplate)
2&3 for any given line the graph shows the relative effects of vertical transfer deteriorating with thickness, while horizontal heat transfer improves. i'm pretty sure any constants used for heat-load or transfer coefficient would show the same scaling with plate thickness
4 your cold-plate is narrower but longer than a 8x8cm plate - the linear heat transfer between the CPU and the ends of the peltiers would be sub-par at 12mm thick, and it's not even that thick around the edges

hollo, I'm not trying to be mean or insulting because your critical of the design. If anything, I welcome the criticism. Tell me my design is steamy pile of dog sh**. I don't care. Just be prepared to back it up.


That being said. The cold plate does have 2 heat source. 1 from the CPU and 1 from the TECs. Think of TECs as a negative heat load. Your not going to convince me that the cold plate is "too thin" by referencing that article again. In my mind, your just comparing apples to oranges. One way you can prove me wrong is to create a fininte element model of the cold plate with the appropriate heat loads on the boundaries. A 2D model with different thicknesses would be sufficient. However I should tell you, I think a finite element model is overkill for this design, mostly because it was designed to fit the motherboard and case over thermal performance. I only mention it because you clearly have an interest in thermodynamics so it might be a good exercise for you and may even aid you in your own endeavors. I'd recommend the ANSYS program if you were interested.

[scifikg]

more eye candy


just making sure it fits right



Back side insulation and mounting



Insulation and thermal probe placement



Retention plate mounting



heatsinks installed

[hollo]

well i don't mean to be insulting either, don't think for a second that i don't find the entire setup pretty sexy

copper is a good conductor of heat compared to other metals, but it's still a terrible conductor of heat compared to a heatpipe or a water-loop. so i knew straight away that the cold-plate was too thin but i was hesitant to say anything without the link because a subjective opinion from a random person on a forum isn't necessarily convincing.

i think your point that the TECs need to be modelled as a seperate negative heatload is invalid
in the heat transfer situation the chart describes there is a heat-load on the bottom of a plate, and a heat transfer coefficient on the top of it. the heat transfer coefficient could be provided by TECs, could be provided by water, could be provided by a slab of frozen turd - in the end it's just a heat transfer coefficient. as far as i can tell from the pics you posted the TECs cover the entire top surface of the cold-plate except the fins at the side, so they're basically just providing an even heat transfer coefficient over the plate, the same situation as modelled by the chart. and thus i think the model represented by the chart is a good comparison to your setup except that their plates were square and yours is rectangular + 2 different thicknesses.

and i don't think the heat transfer coefficient of the top of the plate or the wattage of the heat-source would change the optimum thickness of the plate - i should have mentioned that.
imo the optimum thickness of the plate would be derived from a model of the heat-flow through the plate. the absolute temperature of any point on a 2d cross-sectional model of the plate would be a function of the temperature on the hot side and cold side, but the heat-flow is the derivative function of the absolute temperature function, and the temperatures of the hot-side and cold-side would be removed from the equation as constants.

uh.. i'm not sure if i described that properly, but i'm pretty certain that the heat-flow would determine the heat-spreading power of the plate, and that it wouldn't depend on the temperatures of the hot and cold sides, only on the dimensions of the plate and the thermal conductivity of the material used.
eg between the cold and hot sides of a symmetrical plate the temperature in the middle will always be (Thot + Tcold)/2
so as far as the heat transfer coefficient and heat-load they used being different from those in your setup is concerned, yes it's apples with oranges
but if i'm right they won't affect the "minimum" point of the graph, which is the point showing the best thickness for a heat-spreading plate, and that's what's important.

and fortunately they modelled a material with the thermal conductivity of copper - 400W/mK - at similar dimensions to your cold plate, so it should be a pretty good guide.

that's the basis for my comparison
hope it makes sense, and i'm not saying i don't have it badly wrong
i've barely studied thermodynamics or solid state physics except for one paper that i didn't really enjoy

either way i found myself thinking about your setup - copper is just too crap to properly dissipate heat a distance of 6cm with the high temperature deltas of a overclocking cooling mod - it looks like it would really benefit from heatpipes and it fully has room for some. how about getting a couple of heatpipes with low boiling point refrigerant in them and sticking one on each side with thermal epoxy?

maybe this guy could hook you up http://www.xtremesystems.org/forums/member.php?u=23113

[Jorlen]

I feel like I can mediate a little here.
Hollo is pointing out there are performance benefits to be made by working on the cold plate. And he is right.
scifikg seems pretty happy with the current setup and it works. So to call the cold plate too thing might be overreaching with the adjectives.
Also regarding what NOL said about TRUEs and heatpipes compared to a regular heatsink. A regular solid heatsink has less bumps in the C/W curve. Which means heatpipe based heatsinks "dry" out when they become to hot, the returning fluid in the evap-condense cycle runs low and thermal transport capacities die. So you are left with the passive capability of the copper tubes to transport surplus energy at that point. Which is useless.
And normal solid HS wont suffer the same fate when loads are increased. It has a almost linear C/W curve and to my knowledge only black-body radiation that comes with really high temperatures throw the curve off. But in a positive way, since C/W becomes lower because of it.

[scifikg]

hollo, I do like that heat pipe idea. Maybe when I'm done with the initial testing, I'll give it a shot. I think maybe soldering it the base rather than thermal epoxy


Jorlen, are you saying that if you heat a heat pipe above its maximum wattage it will

1. completely stop transfering heat except for it's outer shell

or

2. keeps transfering its maximum wattage and the excess heat must travel through the outer shell

[ZOMGVTEK]

hollo, I do like that heat pipe idea. Maybe when I'm done with the initial testing, I'll give it a shot. I think maybe soldering it the base rather than thermal epoxy


Jorlen, are you saying that if you heat a heat pipe above its maximum wattage it will

1. completely stop transfering heat except for it's outer shell

or

2. keeps transfering its maximum wattage and the excess heat must travel through the outer shell

My basic understanding is that once a heatpipe reaches its maximum heat removal power, you essentially cant get any more additional heat through it. So, if its 35w, at 30 youre fine, but once you exceed 35, you start to not get effective heat removal, and whatever youre cooling will exponentially heat up.

[evil-98]

none the less i think the design is awesome, and looks great

tho everything i read here appears valid i hope you see great results

good luck 5*s

[Sam__]

some loverly eye candy there... cant wait for temps... wanna see what the outcome to this discussion is. I wish i could have an input but sadly my knowledge is very limited when it comes to TECs.

[systemviper]

wow, that is great, I love the hardware.

I am thinking about a big hunker with some kind of heat pipe addition to handle the quick temp changes but won't put hte tec into overload when/if the hetpipe gets overloaded...

I'v been trying lots of different types of watercooling but not having ot cool the heatplate with water is the only real way to go,,,,,

[n00b 0f l337]

I just realized, your TRUE's are mounted to your aluminum hot plate? Why?

[evil-98]

the aluminum is what is clamping the TRUEs to the copper hot plate, no?

[n00b 0f l337]

Yeah, however it is slowing down heat transfer as well. You really want to have as close to ambient hot side as you can, and I think removing that alu plate, or making it much much thinner might give better results.

[Bloody_Sorcerer]

I think the TECs are TEC1-12730

Is that site legit? Those are some damn fine pelts that i'd like to implement in a future project... Also, any chance you know of a US distributer?

[[XC] MarioMaster]

at this power consumption you could easily run a half horsepower compressor which would give way lower temps than tecs, but i'm sure that's not what you were wanting or you would've done it that way

[scifikg]

Yeah, however it is slowing down heat transfer as well. You really want to have as close to ambient hot side as you can, and I think removing that alu plate, or making it much much thinner might give better results.

The Al retention plate might have a small affect on heat transfer. I could attach a thermal probe to it and see how temperature changes if you would like.

Is that site legit? Those are some damn fine pelts that i'd like to implement in a future project... Also, any chance you know of a US distributer?

I got them on ebay from thermal enterprises. It is the ones they claim are 545W. I asked them if they were the same TECs, but they just ignored my email so I don't know with 100% certainty that they are the same but the specs are almost identical.

[Sam__]

How much would having an ALU coldplate effect it all as opposed to copper? im intrested as im having real trouble finding copper cold plates in the UK.

[scifikg]

How much would having an ALU coldplate effect it all as opposed to copper? im intrested as im having real trouble finding copper cold plates in the UK.

For heat conduction, aluminum has a k value (heat conduction) of 250 W/m*K whereas copper has a k value of 400W/m*K so copper is almost twice as good as aluminum.

Did you contact a local machine shop? Tell them your interested in copper 110. They might not have it in stock but I bet they can get it pretty easy. Most shops always have some aluminum 6061 in stock if you go in that direction.

[systemviper]

thermal enterprises

seems to be the biggest seller on ebay for TEC's and makes the best claims. I personally have a bunch of them, i can't wait for some results to determine if they are bunk or good quality and close to their claims.

regards
steve