Water Cooling with in-line peltier cooler?

[neuroticfox]

I was thinking about how inefficient it would be to have a peltier on top of the cpu or gpu die directly, this led me to looking on amazon, and you can find these small-ish water cooler type plates. I figure someone could power those with the 12 volt cables of a molex or pcie power cable and thermal epoxy them to the sides of it with a few big heatsinks to chill the water. Sure it probably wouldn't get down to sub ambient but it could probably improve the effectiveness of a small form factor loop or something. Anyone got any ideas about this type of thing?

The product I saw:
https://www.amazon.com/DIYhz-Aluminum-Computer-Graphics-Endothermic/dp/B07TDQ1M32/ref=d_sims_dp_d_dex_ai_speed_loc_sccl_3_5/145-3870753-7988936?pd_rd_w=gv3GK&content-id=amzn1.sym.29adec37-ce27-44ef-b467-ea0aa144e49d&pf_rd_p=29adec37-ce27-44ef-b467-ea0aa144e49d&pf_rd_r=VEA651JV20YTMAV7TX7A&pd_rd_wg=j1sfw&pd_rd_r=68070296-7cb3-4008-aa7c-eaa86ee371e6&pd_rd_i=B07TDQ1M32&psc=1

[Eigenvektor]

If a Peltier-element making direct contact with a CPU is inefficient, why would it become more efficient if you remove it from the CPU and put water in between? The amount of heat it can remove doesn't change. The amount of power it requires to do so doesn't change. The amount of cooling it requires in turn doesn't change.

 

Peltier only really makes sense in scenarios where you don't need a high temperature delta, don't have room for a better solution and efficiency isn't a primary concern. If Peltier cooling made sense for PC's, it would already exist in some form of commercial cooling solution.

[jaslion]

Wheter it is on the cpu directly or on a loop doesn't make it magically work better.

 

It has a set heatload it can exchange and that stays constant. You moving it further away from the heatload only adds resistance and makes it work worse.

 

47 minutes ago, neuroticfox said:

it could probably improve the effectiveness of a small form factor loop or something. Anyone got any ideas about this type of thing?

How? A peltier needs cooling too. It wouldn't help as it's basically just adding another part into the whole setup that needs cooling.

 

Keep in mind peltiers are incredibly inefficient too so you'd need a BEEFY psu and guess what there is no space for in a itx system

[Somerandomtechyboi]

I mean i guess you could cool the water down first with normal rads then feed into the peltier then the cpu

 

Still pretty stupid and not gonna work for sff, also just inefficient on a normal build when you can ditch those dinky pc rads for external massive car rads or atleast a massive mora rad which basically eliminates the use of the tec and it wont even bring much temp drop anyways even if you add it. Both will be inconvenient but one will outperform the other and by a pretty huge margin assuming the cpu/gpu can even transfer all its heat into the water

[vsteel]

Pelteir systems are designed to be direct contact because they work better that way.  Adding more interfaces that you have to transfer heat across to would make the system less efficient. 

[AnonymousGuy]

Sigh....it's like every year or so someone thinks they've discovered peltiers.

 

The end of this story is just use a water chiller if you want slightly cooler water.  It's not going to improve your overclocks by enough to make it worth the electricity cost.

[thrasher_565]

you have to buy a tec if you want it to work. they cost a but thow...

[Zinda]

Why not attach the Peltier to the outside of a return water tank. Especially if an aluminum metal tank could be used. Of course running it using a solenoid, and powering with an outboard psu, solenoid would connect to comp psu to start and stop it as well as power its fan. Working from the psu directly (if capable) would be best but would require a high output psu that would cost far more than buying a good cooling system to begin with. 

 

Chilling the tank would replace the need for a super inefficient radiator(s) (by comparison, especially when mounting space is limited).

A radiator can only cool to ambient temps, period! Never able to stay ahead of the cooling needs and thats why they tend to be useless for super hot CPUs. Drawing in cold air in front just loads up the case with more heat causing the board, GPU(s) and CPU(s) to run hotter (a lesser effect on CPU and GPU thats water cooled). 

 

The ability to get the tank water to match ambient temps would be ideal, this would not create any condensation on any metal parts. If you run plastic tubes I think thered be a good chance of running lower than ambient  since the block would heat the water at the connection point and be too hot to collect and moisture.

 

I have made a few custom water blocks that ill be testing tonight. I have not bothered to incorporate the Peltier yet since I'm mord interested in the actual block design first.

After seeing exactly what happens inside the blocks and knowing the fins are prone to collecting a layer of gunk on them quickly, my idea of running tubes that contact directly to the cpu surface seems like a much better approach.

 

I have developed 2 blocks. 1 incorporates water fed from a distribution header (I found a bunch of them for cheap on line, some have control valves for each output, intended for air but I see no difference for water, stainless steel or plated copper while plastic headers have no valves the lines can be controlled by simple choke clamps on each tube to equalize water distribution across the block and the ability to increase flow at hot spots).

 

The 2nd is a looped design that runs the lines across then up a few inches then looped back to the cpu in 5 loops. Of course the water will get hotter as it passes across the cpu. Buy I have planned on using 2 small intel fans from way back in the P4 days. Making the top of each loop flattened for a few inches and spreading them apart will allow a fan mounted beneath the top to cool thr water, plus adding another fan facing downward will further cool the cpu by fir ing air onto the top of the block.

 

The fans are tiny but spin incredibly fast and move lots of air. The problem is they arent easily mounted due to their design and most have been tossed years ago. I hung onto a few just because they were such monsters at moving air for their size.

 

I have to go get a few more parts to be able to do some testing. Both set ups are very rough and crudely made but these are simply made to see if im on the right track.

 

Something will be more precisely made if I find any gains over a premade block using the exact same system for moving the water using a lower flow pump or a higher flow pump. If the low flow works then ill be happy since the cost of pumps is about ⅓rd the price of the now needed higher flow. That way replacing a pump will never be an issue. Since there's less restriction through the lines and possibly no need for the radiator if a Peltier is used, plus cooler tank temps should all increase pump life.

 

Nobody seems to include heat as a major cause of pump failure. You'll find that pumps are rated for max and min temps, some will shut down at certain temps, Ive seen most rated around 65C (very few are rated at 90C and higher), thats right when its needed the most. The temp ratings aren't always available and might require some research, you might be surprized to see the ratings on most pumps and it will explain why so many pumps fail prematurely. Having a Peltier cooling the tank could be a huge factor to a successful and long running water cooling system. Its too bad so many find ways to shoot down anyone who has considered using them. Possibly since most cant shake the idea of mounting them directly on the CPU which is obviously never going to work, or mounting it anywhere but the tank will be costly and infective due to the way they work and the surface area required for their transfer to be effective.

 

The slow moving water in a tank is perfect for the application and even better if an aluminum tank can be used, that way it will distribute cooling as far as the metal will allow and not just at the spot where the pad touches the tank. Most tanks being plastic would not benefit much unless the pad is mounted to a larger copper plate to further distribute the cooling but there's a higher price to pay for the copper over an aluminum tank or even 1 side of the tank being aluminum. Or have the chip attached to a rod that goes inside the tank and is simply sealed with a grommet and mounted on the cover. That way the radiating heat would dissipate quicker and be away from the case and tank.

 

I will post pictures of design after I get everything working and properly registered in case I see potential to start manufacturing or sell the design. If it fails, ill post the numbers I got when compared to a premade fin type block thats been used for 2 months without backflushing or cleaning. I have 2 finned blocks, 1 I have gutted the plastic or rubber pieces from it since they were deformed and causing restriction and it cools better than the one that has the factory design. Both had failed pumps after 1 years use and temps 0hezwere higher than an upgraded heat pipe cooler running on a old AMD phenom 2 x4 965B with low overclocking @ 4x 4.2KHz (3.6KHz stock).

 

Now I'm running an older Intel I7 4790K on a Z97usb3.1 Asus board and it runs hot with no overclocking. Hits 80s with 100% stress test, with high flow pump and 90s with a smaller pump. Single radiator with push pull full-sized fans mounted in rear of case with a ¾ gallon tank running with about ⅚ gallon of coolant, water gets pretty warm in AZ summers when no AC is on ambient will be in the high 90s when AC is running its not much different due to ths location of the comp, heat builds up in the area around it quickly. Thats what lead me to try some new approaches, the system runs a single thread faster than nearly any other CPU made other than 1 according to tests ive done with no overclock at all. Id like to be able to OC it and see whats possible. Mem is 2400, GPU is GTX1660Super never hotter than 78C running 102% its the slowest part of the system but never uses all its resources either.