Alternative TEC Cooling?

[Perturabo]

Originally Posted by XionEternum

This is when the cold side is applied directly to the source of heat. It directly counterbalances the heat transfer rate of the TEC and if generating more heat than it can transfer, it looses and even reverses efficiency. I am talking about air cooling a TEC at a wattage that air cooling can handle, but then chilling the water flow on the cold-side. Big thermal dynamic difference if the air cooler can keep the hot-side cool enough to have a delta balanced to about ambient. (IE. hot-side = X and cold-side = -X) Since neither side is being hit with a thermal source, it stands to reason that it would act as a chiller, or this sort of thing wouldn't work in fish tanks, yes? :P

But you are completely ignoring the heat being added to the water by the CPU are you not? What you are saying only holds true if there is not a heat source present in the loop...The reason in works in a fish tank is that the fish output less power as heat than the the TEC, leaving it able to control the water temperature.

Now you are saying you also have rads in the loop to remove the heat from the CPU and then the TEC cools the water further correct?

This will not work because rads work both ways, if the TEC cools the water below ambient then the rads will absorb heat from the air and add it to the loop. So you will be back to ambient again.

[XionEternum]

Originally Posted by Perturabo

But you are completely ignoring the heat being added to the water by the CPU are you not? What you are saying only holds true if there is not a heat source present in the loop...The reason in works in a fish tank is that the fish output less power as heat than the the TEC, leaving it able to control the water temperature.

Now you are saying you also have rads in the loop to remove the heat from the CPU and then the TEC cools the water further correct?

This will not work because rads work both ways, if the TEC cools the water below ambient then the rads will absorb heat from the air and add it to the loop. So you will be back to ambient again.

No, I am not ignoring the heat output by the CPU. Please reread my original post, which was edited before you posted to include information that invalidates your counterpoint. Not to mention another responder already gave me a link to someone who was successful in a similar attempt via PM, which I responded to also in the thread since I can't reply directly on a new account. Since you have a problem reading every post in a one-page thread, the loop sequence is: Res, TEC block, CPU block, GPU block(s), Rad(s), Repeat. Also presume that the rads are capable of cooling the water to ambient, and that condensation is accounted for. Now admittedly the sample that was linked to me was involving a 400w TEC with ~15 degree drop in temps, compared to my 80-250w concept. Knowing it works means I will be testing it myself with various wattage TECs when I have the means. Now pay attention since you seem to have issues with procedural processes: The coolant starts at ambient, hits the chilled TEC block and cools below ambient, hits the CPU/GPU blocks and warms up, hits the rads and is rebalanced to ambient, and repeats. What I do not know and intend to test when the opportunity arises, is what wattage TEC is ideally balanced for this application. I also intend to test whether or not the TEC is chilling the water so far below ambient that the CPU and GPU heat doesn't raise it above ambient, by having fan-speed control and turning them down on the rads during testing. I doubt it'll help without once again having a higher wattage TEC than the combined TDP of the CPU/GPUs in the loop, but it's worth testing.

TECs are an odd thermal dynamic. I have seen and understand the formula for when the cold-side is applied to the CPU directly. In this practice, say you have a 150watt CPU and a 300watt TEC. The estimated delta between the cold and hot sides of the TEC is about 30-35 degrees. If your cooling solution can transfer 450 watts or more of heat, then it's fairly efficient. Now this is directly because the TEC can only transfer heat from one side to the other at a certain rate dependent on wattage. More watts is faster. This also generates heat and raises the center-point of the delta between both sides. Now applying a direct heat source to the side that is drawing heat, reduces the delta by a fraction of the two wattages. Now, what I am suggesting, and has been validated by someone else, is removing the heat generated by the CPU/GPUs before it ever touches the TEC so there is no inefficiency imbalance, and you get the full efficiency regardless of wattage, as long as you can keep the hot side cool enough.

[Perturabo]

I have read your posts, I just don't agree, thats all, no need to get defensive.

This is what I am thinking will happen, im sure you will agree with most of it but I will go through it all for the sake of completeness.

Say we have a loop like you said. CPU/GPU blocks -> rads -> TEC -> back to blocks.

For a reference I am first going to assume there is no TEC. Water in loop starts at ambient, CPU and GPU loaded and produce heat. Heat transfer from block to loop is proportional to (T_blocks - T_water) this difference starts high and heat flows to the water increasing the temperature a small amount. Then the water hits the rads and is cooled at a rate proportional to (T_water - T_ambient). Temperature difference is small at first so little heat is removed. T_water is not slightly higher so heat transfer from the block is slower so T_water increases again but by less than the first pass. Now (T_water - T_ambient) is larger so the rad can remove heat faster. So over time the water temperature increases, the heat flow from the block decreases and the heat flow from the rads increases. So at some point an equilibrium is reached and the water temperature reaches equilibrium.

So that part is all standard stuff that we both agree on. Water in the loop reaches an average temperature that is a certain amount above ambient, say ambient + 10 C for example.

This is what I think will happen when you add a low power TEC to the same loop.

ambient + 10 C water goes into TEC block, some heat is removed, T_water drops slightly. At the CPU/GPU blocks (T_blocks - T_water) is now larger than at equilibrium so more heat can be extracted from the blocks. At the rads (T_water - T_ambient) is now smaller so the heat flow out of the rads is reduced slightly. The water returning to the TEC blocks is cooler than it was on the first pass, say ambient +9.9 for example. As the process repeats heat flow out of the CPU/GPU blocks increases and heat flow into the air from the rads decreases. Thus another equilibrium is reached, say ambient + 5 C.

So I agree that a low power TEC will decrease average water temperature and thus CPU/GPU temperature but the water will remain above ambient.

Next you try a higher power TEC, the process above repeats but this time the water temperature reaches ambient. In this case you now have zero heat flow out of the loop from the rads as there is no temperature difference between the air and the water. At this stage you have the TEC removing all the heat from the loop and the rads achieving nothing. So you may as well remove the rads and then you are at the position I was talking about originally. A high power TEC doing all the cooling and no need for rads.

Assuming you kept the rads in the loop and chose an even more powerful TEC you would then get to the point where the water temperature starts to drop below ambient. At this point the sign on the rad equation reverses and the rad starts pulling heat into the loop, counteracting the TEC more and more as the temperature decreases.

So I agree that a TEC WITH rads would help for trying to push your temps all the way down to ambient but if you want to go sub ambient I think you need to remove the rads and just have a high power TEC, like is the conventional way of using them.

Sorry for the long post but I wanted to try and get my point across clearly. I find discussing new ideas like this interesting and im not just trying to shoot you down. If you can come up with a reason that I am wrong in my understanding of the situation I am more than happy to admit that I am wrong.

[XionEternum]

Originally Posted by Perturabo

I have read your posts, I just don't agree, thats all, no need to get defensive.

This is what I am thinking will happen, im sure you will agree with most of it but I will go through it all for the sake of completeness.

Say we have a loop like you said. CPU/GPU blocks -> rads -> TEC -> back to blocks.

For a reference I am first going to assume there is no TEC. Water in loop starts at ambient, CPU and GPU loaded and produce heat. Heat transfer from block to loop is proportional to (T_blocks - T_water) this difference starts high and heat flows to the water increasing the temperature a small amount. Then the water hits the rads and is cooled at a rate proportional to (T_water - T_ambient). Temperature difference is small at first so little heat is removed. T_water is not slightly higher so heat transfer from the block is slower so T_water increases again but by less than the first pass. Now (T_water - T_ambient) is larger so the rad can remove heat faster. So over time the water temperature increases, the heat flow from the block decreases and the heat flow from the rads increases. So at some point an equilibrium is reached and the water temperature reaches equilibrium.

So that part is all standard stuff that we both agree on. Water in the loop reaches an average temperature that is a certain amount above ambient, say ambient + 10 C for example.

This is what I think will happen when you add a low power TEC to the same loop.

ambient + 10 C water goes into TEC block, some heat is removed, T_water drops slightly. At the CPU/GPU blocks (T_blocks - T_water) is now larger than at equilibrium so more heat can be extracted from the blocks. At the rads (T_water - T_ambient) is now smaller so the heat flow out of the rads is reduced slightly. The water returning to the TEC blocks is cooler than it was on the first pass, say ambient +9.9 for example. As the process repeats heat flow out of the CPU/GPU blocks increases and heat flow into the air from the rads decreases. Thus another equilibrium is reached, say ambient + 5 C.

So I agree that a low power TEC will decrease average water temperature and thus CPU/GPU temperature but the water will remain above ambient.

Next you try a higher power TEC, the process above repeats but this time the water temperature reaches ambient. In this case you now have zero heat flow out of the loop from the rads as there is no temperature difference between the air and the water. At this stage you have the TEC removing all the heat from the loop and the rads achieving nothing. So you may as well remove the rads and then you are at the position I was talking about originally. A high power TEC doing all the cooling and no need for rads.

Assuming you kept the rads in the loop and chose an even more powerful TEC you would then get to the point where the water temperature starts to drop below ambient. At this point the sign on the rad equation reverses and the rad starts pulling heat into the loop, counteracting the TEC more and more as the temperature decreases.

So I agree that a TEC WITH rads would help for trying to push your temps all the way down to ambient but if you want to go sub ambient I think you need to remove the rads and just have a high power TEC, like is the conventional way of using them.

Sorry for the long post but I wanted to try and get my point across clearly. I find discussing new ideas like this interesting and im not just trying to shoot you down. If you can come up with a reason that I am wrong in my understanding of the situation I am more than happy to admit that I am wrong.

Now this is what I like to see, thank you for being detailed and my apologies for being presumptuous of your position.

I had already 'planned' for the possibility that the rads will be rendered ineffective at some point with the intention of applying fan-speed reduction, even to the point of shutting them down. The reason my initial example is based around a low power TEC is because I want the rads in the first place, and because I am considering a decent $50 CPU cooler. My Antec Lanboy Air is the inspiration for that since it's about the only thing that would fit. Larger chassis such as the Switch 810 would easily support an NH-D14 on the back. I could even still go through the effort of a dedicated dual-loop using one Monsta triple rad (yes, will fit my Lanboy *wink*) in a stand-alone loop for the hot-side of a high-power TEC sandwiched into the primary loop for the cold side. Now this application may, and I say that with a grain of salt, prove more efficient than applying the TEC directly to the CPU as is common practice.

I do know that application of a 400w TEC using a cheap Antec 625 AIO LCS to cool the hot side is plenty efficient applied to an existing basic loop with about a 15 degree drop in CPU/GPU temps (provided the source itself wasn't misinformation), however I am curious how efficient that is compared to different TEC wattages in the same setup, for one. For another I'm curious how efficient it will be with a decent air cooler instead, since the one tested in the 400w TEC sample linked to me was next to near junk.

Edit: I would also like to point out that unless you are running Prime95 or something similar, such a balance will never happen in the water. CPU usage spikes and dips and for the most part is idle most of the time, with partial usage while playing games, and full-usage while video encoding or 3D rendering and the like. So, I'm mostly concerned with that, and far less concerned with getting another 100Mhz out of the CPU. I'm also more concerned with making the CPU last as long as possible while still getting the most out of it. In that sense, colder is generally better especially under load.

[Mytheroo]

hi all, I also have wondered about a TEC cooled waterloop for many years, but the explanation just given as to its inneffectiveness seems pretty robust.

I was wondering what you folks thought about this:

Pump > TEC (cold) > Blocks > TEC (hot) > Rads > res > pump

this would mean the coldest water hits the blocks (keeping it above the dew point though) and then the hottest water hits the rads maximising the cooling there.

I know it's logical to think it can't work due to the TEC adding more heat to the loop, but this principle does work in A/C systems where the R134 gets squished back into a (hot) liquid by the compressor.

I appreciate your thoughts

[XionEternum]

Originally Posted by Mytheroo

hi all, I also have wondered about a TEC cooled waterloop for many years, but the explanation just given as to its inneffectiveness seems pretty robust.

I was wondering what you folks thought about this:

Pump > TEC (cold) > Blocks > TEC (hot) > Rads > res > pump

this would mean the coldest water hits the blocks (keeping it above the dew point though) and then the hottest water hits the rads maximising the cooling there.

I know it's logical to think it can't work due to the TEC adding more heat to the loop, but this principle does work in A/C systems where the R134 gets squished back into a (hot) liquid by the compressor.

I appreciate your thoughts

No, there is a specific reason I wanted to avoid that in my example, which is detailed rather well by Perturabo: The water eventually obtains an equilibrium.
This means that the water in the primary loop will balance out to about the same temperature across the entire loop, not getting warmer due to the rads (and TEC if applied as I suggested), and not getting colder due to ambient temp and the blocks (and TEC if applied as is standard). Since it balances out between the heat generated and removed by the rads, and since the TEC only transfers heat from one side to the other, the addition of a TEC in that way will have theoretically no effect due to equilibrium.
While in this particular thread, there is a theoretical inefficiency in TEC usage, it is more efficient than applying the TEC directly to the CPU. As things stand, using a TEC with a lower TDP than the CPU will actually heat the CPU up under load since it can't transfer heat as quickly as the CPU is generating it if applied directly to the CPU. My theory (which was confirmed in a link provided via PM) was to chill the flow somewhat by pulling some of the water's heat out of it as it flowed through a block on the cold-side of a TEC, and have either an air-cooler or a second loop for the hot side. The separation was meant to keep the heat from the TEC out of the primary loop entirely. It's been done, and works. I'm looking forward to actually building it eventually. I am aware that as expensive as such a build would be, that getting a pre-built phase-change system would cost as much and perform better. It is nowhere near as aesthetically pleasing, so I'll pass on phase-change for now.
As for the A/C example, I was under the impression most A/C units used phase-change. Mine does and it's a cheap $100 window unit for my PC room. The only application of TECs I know of outside PC liquid-cooling is temperature control for fish tanks.

[Mytheroo]

yes indeed it is phase change, but for this to occur you have to add heat into the loop. The reason it works is that at the radiators there is a large delta in temperature, and lots of heat energy is dissipated to the (outside) air. This leaves a high pressure cool liquid that when it hits the low pressure zone drops rapidly in temperature (-51C or whatever). In a sealed system the phase-changing doesn't remove heat, it's just a great way of creating a large delta in controlled places. A TEC does the same (creates a large delta) though is less efficient i think

So I'm suggesting a system that also has a wider range of water temperature than normal, colder at the blocks and hotter at the radiator. Yes this will be an equilibrium, but with the min and max temps being further apart.

What I am wondering is if the extra heat introduced by the TEC will be countered by the extra heat radiation/conduction/convection at the rads.

[XionEternum]

While I have not tested it, I still doubt there would be much difference unless the radiators are more than capable of purging all the heat accumulated throughout the loop. At that point, there may be a noticeable variance potentially along the same theoretical level as my concept. Due to the thermal transference per watt on TECs, I would suggest at least ~250w minimum for any appreciable effect. You may also want to consider AquaComputer's Airplex Modularity System radiators; perhaps the dual-circuit model so you can pre-cool the water before the TEC, then back into the second circuit in the rad. It's what I've been planning recently.