Can A Wish.com Water Block and a $4 Peltier Beat a 360mm Radiator?
ฝัง
- เผยแพร่เมื่อ 5 ก.ย. 2024
- This video was a requested by Onno via twitter, he wants to know if you can replace a radiator with a peltier cooler. I thought this was interesting so I decided to find out, Can A Wish.com Water Block and a $4 Peltier Beat a 360mm Radiator?
This is a terrible idea. Keep it up!
its really not, the only problem i see here is the surface area the radiator covers vs the waterblock. the water goes through so many small tunnels with a bunch of fins on the side that cool the water instantly and efficiently with a radiator and the waterblock well its just a block of copper. try cooling a bucket of water with an ice cube.
@@rookm13 Actually it is a terrible idea. The TECs power consumption would need to be higher than the thermal output of the CPU and since the TEC also produces heat, the cooler would need to get rid of the combined heat of CPU and TEC. I am to lazy to calculate this, but with modern CPUs that's a lot of heat. Probably way north of 300W.
@@andreasvogler1875 efficiency is for people who don't live under the warm loving glow of a nuclear power plant.
@@ARVash ;-) I wasn't even talking about efficiency. The TEC he is using is probably too weak and even if it wasn't, the air-cooler definitely can't dissipate that much heat.
@@andreasvogler1875 ah, maybe he could put it in line with the existing radiator. That way he could extract heat from an "already cool" line
The Peltier element has a certain cooling capacity, once you put enough heat on its cold side it gets overwhelmed and the heat stays trapped in the water loop.
Off the top of my head, when I was cooling Pentium III CPUs which had a TDP of around 35W, my 60W Peltier (12V 5A) couldn't keep up and I got worse temps than without the TEC. I would say you need to put at least 3x TDP's worth of power into your TEC, or across multiple TECs, in order to get any cooling at all.
So for a ~90W TDP CPU you would need at least 270W worth of TECs. That would be 5 60W elements. You should go for at least 7 or 8 to have some margin, and maybe even to get sub-ambient at low loads.
Is this expensive and outrageously inefficient? Yes. But it should be cool.
EDIT: I should also add - don't stack your Peltiers, keep them at 1 element per tower cooler, and chain the water block+Peltier+tower cooler assemblies in a water loop. Maybe add a radiator without fans into the loop to provide a sort of heat buffer, without a radiator the total thermal mass of the loop is small and you might get large spikes in temperature.
IIrc they are about 1-5% efficient. so at 2%, you're looking at a 12710 (154w) cooling 3,08w
which means you'd need 35 to cool a 3950x (105W)
though you could bring it close to ambiant with a 3x3 radiator, and use like 2kw of tecs to bring it down a few degrees below ambiant.
but the best option is probably salvaging an old ac unit and make a chiller with it. (or cool antifreeze with it.)
Der8auer did something like that in this Video: th-cam.com/video/YaOjrk98vPI/w-d-xo.html
yeah, there's a reason Cooler Master abandoned their TEC cooler plans (See V10 cooler)
Would it just be more efficient to add a radiator to the loop to reduce the load on the peltier?
@@metaleater9 That wouldnt hurt, but he really needs more cooling on the hot side of the peltier. And if he were to get below ambient like he thought he was a radiator would actually increase the temperature of the loop
6:05 "Yeah, we're doin' things!" should be the slogan of the channel
t shirt
Major Hardware take my money
True
"...pretty much everything I do on this channel doesn't make sense"
But it sure is interesting entertainment
I love the authenticity and enthusiasm of this channel.
Could you run a radiator directly out of the cpu, then somewhere later in the loop run a peltier water block?
This is the only logical solution I don't know why he and linus didn't do it
Or maybe using a car radiator like some people did years ago.
Unless you're using an incredibly underpowered pump, loop order doesn't matter.
@@mikee9167 any youtuber this that in that way? i didn't see that atm.
Or use multiples setup like this in series... probably won't work as peltier don't seems to do well as cooling...
Maybe if a peltier was integrated with a rad directly it could work wonder... but impossible to test here :p
Turn the thing upside down man, convection is a thing.
i was looking for someone to comment on that. the heat pipes have to be a certain orientation to work properly.
Ohh, you mean right side up.
You guys get it. Not only is the Peltier under powered but the orientation is causing the air cooler to actively work against itself by heating the top side preventing any condensation that might have occurred due to the peltier heating the pipes.
I tried this way back when my AMD K7 was bleeding edge. It cooled the chip well, really well, then melded the fan on the radiator.
Try using a second waterblock with an air cooler on it and compare it to direct air cooling and to a radiator.
I would love to see one or more of these hooked up in series with a radiator in the same loop
I’m to the point now where I just wait for your videos to be uploaded the rest of TH-cam is just boring. Awesome work. I love watching these awesome experiments
Peltier can be used as a supporter to the radiator. It can be in series to the radiator before and/or after.
What about adding the peltier to an existing water loop?
I mean, have the colder return hose to the CPU be connected to a peltier device and see if that super cools the water in addition to a reg loop setup. How much lower could you get in terms of temps with and without the peltier device?
Different peltiers have different heat dissipation. I would opt for a 250w should work just fine
Oh hey, I just recently graduated with a thesis based on trying to cool a CPU with Peltier. The result of that is pretty simple:
Unless you have an IHS the size of a Peltier that is capable of moving MORE heat than the CPU produces - you cannot cool the CPU. Heat transfer rate of such Peltier module is way too low, hence it must be stronger than the heat source. And the moment you start pumping say 100W from the CPU you have over 200W of heat to take from the hot side. I've used a 40x40x3.2mm (TEC1-12715) module with Qmax of 144W, at peak efficiency at 13.2V (powered from a PSU that supplied my PC, had to use DC-DC boost converter :D) it I still got a score of 59C under 65W of load, as a reference I've used Liquid Freezer II 360mm, which managed to keep that temperature under 95W. And boy, was the radiator of that LF II warm when it cooled the Peltier under this 13.2V.
The only way of "efficiently" using Peltier module is for a burst stress testing, such as a CPU-Z validation. Also, sides of the Peltier are terribly wavy, you need to spread the paste manually.
200W of heat from the hot side? that's generous, I've seen efficiency numbers around 2-5%, so that'd be more like 2000-5000W.
manaquri I think you can get 10% efficient to get a few degrees advantage, but it requires under-volting them (so you might need twice as many). But that’s compared to say my over 100% cooling efficiency of my phase cooler (where it uses about 350w when cooling a 350w CPU at -35c, which is the hottest I have ever seen it get, just a basic Vapochill LS and I was testing an FX-8350 at 1.7 Vcore but my garbage board with only a 4 phase VRM could only get it to 5.5 GHz rock solid, or 6 GHz single core, but it would basically be 9900k at 6 GHz at that temperature for modern comparisons with similar power draw since they get much more efficient at these temps, I’d guess the FX was 35% more efficient because with the power draw scaling it would have been around 500w if by some miracle a waterblock could have cooled that volcano below thermal shutdown). I think it can handle a little more heat load (especially if it’s just a burst (maybe 400w cooling capacity at some temp maybe like -25c), but it’s far too small for my Threadripper (which sits happily on a 120mm AIO chugging through compression encodes all day long at 450w with it overclocked). I do want cascade sometime that can handle 1kW with TR-64 core but I need to have some parts made like the expansion block which I have modeled up but I don’t have a CNC machine yet.
@@jakegarrett8109 I always find it funny that phase change heating can be over 100% electrically efficient.
so you're using 350w total for the pump but probably outputing a bit over 700w at the heatsink. (in that case it's only 100%, because the cpu is in the same room, but if you cool outside air you can get up to 300-400% efficiency.)
last time I tried something in the >300W range, I fried a psu, it didn't like the phenom II and the 7970 overclocked at the same time, with furmark+linpack.
300W cpu + 400W GPU on a 500W no name psu, do the math :p
@@satibel @manaquri hot side gets a heat output equivalent to the sum of Joules heat and Peltiers effect, and you'd sooner get 7GHz stable than get even 500W of heat on a working, lower perfomance module.
For a phase changer you either go with multiple TECs and use them on low voltage or you have no way of cooling units.
@@satibel that "over 100% efficiency" is just an artifact of how you are calculating that efficiency. If the system is at equilibrium and we are assuming it is an ideal system then if you are putting 350W in at the pump and getting 700W out of the hot side then you are pulling 350W from whatever it is you are cooling. And the proper way to calculate efficiency for this type of device is, desired/input in this case the energy transfer of the cold side is our desired metric so: 350W/350W=1.0 or 100%. Just knowing work in at the pump and energy out of the heatsink isn't enough to calculate the actual efficiency of the system.
In actuality the efficiency is likely closer to say 0.9 so if you are putting 350W in at the pump you will be pulling 315W from whatever it is you are cooling.
"the perfect amount of thermal paste... " Love this channel! Hey, link that break out board, if you will.
One thing that may have an affect is the direction of your air cooler. The heat pipes are designed to bring heat away, but it might be more difficult with the air cooler upside down. The fluid in the pipes will rise when it gets hot.
I think if you had the fin stack the right way up, helping with convective heat transfer it would probably help a little better.
you need more peltiers than 1 in order to cool a cpu. I'd say get 7 more and then try again. But this will work you just need more peltiers
It won't peltiers are pretty garbage
Don't believe me? Check Der8auer's vid it's a few months old
@@racistpandagod No they will and saying peltiers are garbage just means you have no knowledge about them. Yes peltiers are very inefficient but they take little to no space at all and as a bonus they act as a generator if the temperature difference is there.
@@racistpandagod Also I have seen his vid and the peltiers on his vid are TERRIBLY COOLED!
@@racistpandagod Done it with two peltier of good quality and 3 waterblock. I was able to move 344w of heat load. Electric bill thru the roof (700W of load to move 344w only)
@@racistpandagod He used an awful setup and cheap, low-power peltiers. There are peltier chillers out there that will easily go sub-ambient, even sub-zero - it's a matter of designing the system correctly. Check out the channel Making Stuff Awesome to see how it's done - the guy designed peltier waterblocks for the military to cool equipment in areas where reliability was more important than efficiency, seeing that peltiers are solid state with no moving parts, unlike compressor-based chillers.
Dell offered a water loop with a Peltier block called H2C. The XPS 710/720/730 had it as an expensive option. It was regulated to not go below ambient temperature. If it goes below ambient temperature the radiator starts adding heat into the loop. This also avoided condensation issues. It had 2x Peltier chips after the radiator. Peltier can easily go well below ambient if you have enough of them. But then condensation will need to be considered.
This would be interesting to see if you hooked up several tec/air cooler setups in series just for giggles and see if passing the water over several cool side tecs would make it work.
De8auer used 12 peltiers/air coolers. th-cam.com/video/GZ3SwM2-rD4/w-d-xo.html
Betteridge's law of headlines. Betteridge's law of headlines is an adage that states: "Any headline that ends in a question mark can be answered by the word no."
I don't even know where to start... Lets start from the basic. 1. For cooling purpose DO NOT use 12volts. 2. There was ZERO insulation between hot and cold side, so the hot side was warming the cold side. 3. There was zero insulation in water pipes etc, when you want to go close ambient or lower, you will need good insulation everywhere because the room is heating up all the surfaces that are not insulated. Here are couple ideas for you, which i have tested and can confirm working, use 2x12706 stacked and run them at 5volts. This way you can easily reach 40-degrees under ambient temp, measured from surface of peltier. This method also uses WAY less watts (20watts) than running 1x12706 at 12volts (72watts) and it also makes MUCH LESS useless heat in the warm side. Currently my best result is using: 12710+12708+12703 stacked, running them all at 5volt, about 45watts and the cold side goes 57-degrees under ambient temperature. In a nut shell: 1.Drop voltage 2. Use stacked method 3. GOOD insulation and you can see some serious schit to happen. ;)
In the realm of techtubers, you've definitely solidified your place as the Mad Scientist between stuff like this and the ongoing fan experiments. I love it!
Revisiting this with a better peltier and a larger/better waterblock would be a worth it. Maybe place your reservoir between the blocks so the CPU block is receiving the coldest water?
So a 100w peltier will provide ~10w of cooling and 90w of heat. The tec also provides quite a lot of thermal resistance. It would work a lot better to put the water block straight on the air cooler
The fact that you ACTUALLY got something from WISH is in itself a miracle! I JUST got some filament that I ordered in MAY last week! (if I had ordered it from Amazon, and had it shipped from the China warehouse) in 14 days!
I had an all in one, CoolIt peltier setup years ago and that definitely worked. It had 3 peltiers sandwiched between 3 little water blocks (joined together in a loop) on one side, and a round central doughnut like radiator in the middle, on the other side. One end of the radiator doughnut was closed and had a waterpump bolted to the back side, the other open end had a fan sucking air out the back of my case. It used a lot of power but my temps never exceeded 20 degrees celsius.
What wattage is that TEC cooler? Sorry if you mentioned it and I missed it, but maybe you could stack a bunch of high wattage TECs. 😂
just 60W for this little guy
Maybe a 250w pelt next time! If there is a next time...
@@doubleas88 that would make the total amount of heat that he would need to dissipate 250w plus the power of the CPU. You can't cool that easily.
@@ryanmickelwait1521 Pelts get stupid inefficient as you max their voltage, which is why going by the ratings always leads to these results. What you'd want is something like this: www.amazon.com/BQLZR-TEC1-12730-Thermoelectric-Peltier-Cooling/dp/B00EQ1X5EC
But only run at 7 volts/20 amps instead of the full 12/30
@@MajorHardware Oh yeah. That's a little baby TEC. I don't know what but I have a weird fascination with TEC
You can put the peltier in the reservoir of the 360 loop and see if we can squeeze more performance out of the system.
Rather put ice in it really
it's gonna freeze on the surface of the TEC
now the question is how many peltier do you need to beat a 360mm radiator
5 with noctua fans
in my home i use 5 15A peltiers 120mm in cold and 240mm with beefy 2x4000rpm san ace fans must say removes the heat produced by my pc and removes the need of requiring a ac
@@hyakinthos_0902 what are you doing with that massive amount of heat though? Are your radiators outside?
@@vikmanphotography7984 yes outdoors
@@hyakinthos_0902 fair enough.
i have been waiting for you to do projects with peltier coolers :).. i have a little idea.. try put the peltiers on the radiator , or peltiers in a closed box and then have the heatsink with the hot side stick out of the box so that does not generate heat inside the box, and for better cooling performance you could add a heatsink or some other metal block to the cold side on the peltier inside the box with the radiator i think that might think that could work , also remember to add rgb light for extra cool performance
"Yeah, we're doin things!"....lol. Great stuff man :D
Since you already have all the components try this simple modification:
CPU block -> radiator -> peltier block -> back to CPU block
Let's see if is possible to achieve sub ambient coolant temperature. Added bonus, use your water cooled air cooler to cool the peltier for extra lols! And if you use the cooling tower to cool the coolant of the water cooled air cooler it would just be PERFECT!
I had a spare 4770k lying around still in its old mobo with a 850 watt power supply. I had just learned what peltier tacs were that day and instantly got the idea to cool a cpu with that day, went on ebay and grabbed a 450w peltier tac and went to work on stripping couple psu cables when it arrived I slapped a hyper 212 on it and booted it up, idle temps were 4-5c and load temps were just around 30c, but after 10 mins I didnt realize it was condensating and then it got into the socket and short circuited the board and cpu lol. Good times
There was actually a TH-cam video on someone doing this a good couple of years ago on the "Then Blazing Fast" AMD Athlon. They had the xavt same issues, and resorted to conformally coating their MB in something to avoid destroying their setup.
If recall correctly, they were averaging around the 5° mark as well 🤔
Now im curious as to how that would work without the Peltier element.
I used to have a pair of 40x40 TECs to cool a water loop. I customized a DangerDen 5.25" Acrylic Tank by removing the bottom piece of acrylic and replaced it with a 1/16" piece of aluminum sheet metal. The TECs were attached to that with a couple of CPU coolers and fans. The loop was Tank -> CPU -> Pump -> Radiator w/Fan -> Tank. The idea being that to get the coldest coolant (read automotive antifreeze basically) possible it needed to spend a lot of time resting on the TEC and be as close to ambient room temp as I could get it. I think I managed to get about a reasonable drop over ambient room temps but it was a PITA and these weren't the cheap wish.com style ones either. Honestly, I wouldn't run that type of setup again, modern water cooling systems are much more efficient than the early 2000s when I used to tinker with trying to OC a CPU to within an inch of its life and such.
You'll need more peltier elements, to make it cool. one is just too ineffiecient, so if you make a loop with like 10-14 peltier elements, it could work
On this vain, it could be limited by voltage, I have 2 peltiers and they're rated between 12v to 36v
Unfortunately no. Peltiers are only effective in large applications. The devices he is using are only capable of a couple of watts each of thermal power and as such are easily overwhelmed by a CPU. These are more effective as thermoelectric generators. You can even see them on some stoves connected to a fan motor to stir the air.
There is a german channel that tried this with several radiators and several high rated peltiers and a ton of fans and it still failed
@@SoranoGuardias the also work well in those coolers
@@sinformant it was @Der8auer and i watched that viedeo too
That B-roll... Damn man, love your channel, its easy to see the work you put into your channel and its a blast to watch
Surface area, plain and simple. Cool video, sir!
If you want it to work then either you want the Peltier modules to be directly on the CPU and have a water block on it or you could have an fan cooler on it.
I am really hoping you go further with Peltier units, I'm curious about this stuff and no one goes to any actual lengths in their videos to make a proper effort with them, I hope to see a continuation of this!
Try looking for the "Curious Scientist" he has done so many videos on peltier cooling (Even though his setups are geared more towards how inefficient they would be as air conditioners, he has more data than I'm sure you would know what to do with 🤷🏼♂️).
Yay.
The idea behind it was based off one of those ancient fad coolers that had a peltier built in. This was before there were dedicated water loops so I was wondering if the new water loop tech and new heatpipe tech would make it worth giving peltiers another look. I was thinking it would at least allow for supplemental cooling or relocating a big tower cooler. I am bad at math, and a lot of the comments below seem to confirm this.
Running off 12v just doesn't seem like enough juice and the thing just gets saturated too quickly. To get around that (if still possible, which it's probably not) would require additional crap that would defeat its intended purpose.
I am just glad this didn't set you back too much.
Peltier works. I remember even Cooler Master adding one on their V10. It was a gimmick at that price, but it added something. It just that even if it can make a temperature difference on both sides, it's limited in how much thermal power it can transfer.
So the reason why this would never work is that what's happening in the radiator is that the hot water is being cooled via the surface area. The fans push air through the radiator to better dissipate the heat, but really it's just making the radiator work at maximum efficiency.
If you wanted the Peltier to cool the water in any meaningful way you'd need to slow the water passing through the Peltier, in this case, you'd need a LONGER Path over the Peltier to increase the surface area of the water.
Additionally: The CPU block is designed to draw heat up to the water for the water to then carry the heat energy away, not for the heat energy to be absorbed back down. Basically, it doesn't work going the other way (with the heat source going through the fins and down to the Peltier).
So to get this to work, you'd need a way for the water to Zig/Zag over a larger Peltier in such a way that the Peltier has time to cool the water enough so that it no only cools the water down completely, but also can soak the heat energy created by the CPU.
As soon as I read Peltier in the title: Nope. Doesn't work. Whatever it is, it doesn't work.
The homemade water cooling kit I bought used for my K7-800 back in 2000 came with a peltier that looked excactly like yours.
I couldn't get the setup watertight, so I never got to test the efficiency against my insanely efficient Alpha Slot-A cooler with 2 60mm fans. I mounted a temp probe behind the heat spreader, as closely to the CPU core as I could, and for some reason my Asus K7V reported sub-ambient temperatures.
Ah, the good old days when you had to hack into the encasement of the CPU and void the waranty to mount a Goldfinger-device to alter the multiplier. If I remember correctly, I got the K7 to run stable at 950MHz with the Alpha.
Good times... :)
You do super cool video ideas. It's just a matter of time before you hit something that blows up your channel
I've seen so many experiments like this, the result is always the same: Peltier cells works but they are NOT EFFICIENT enough at the point that you'll have better results just removing them from the loop. EDIT: but we always need Major Hardware to do experiments like this, to look and find if the things the vendors sell us are the better ways to do them, or if they could be improved in some ways! Thank you Major, you're our hero! Greetings from Italy
Here's a challenge for you: somehow merge two water blocks into a single double sided one and put one of those air coolers on each side
I always thought the best way to do this would be to use a peltier + air cooler in addition to the radiator. AFAIK the peltier's goign to be sucking X number of watts out of thermal energy out of the water regardless of delta, right? So have the water go out of the cpu into the radiator first to dump as much into the air as possible that way, then when the delta is lower use the peltier to push the water temperature down even further.
The way i used the peltier was doing a chilled water setup. Hooked up the peltier on the side of a reservoir with an air cooler attached and a thermostat controlling the peltier. Placed the thermistor inside of the reservoir. It did work to cool the water down to 5 degrees celcius.
Also, a little advice for using peltiers. The cooler you can get the hot side of it, the colder the cold side will get. If the temperatures that you are trying to cool are too warm for the cold side and you have inadequate cooling on the hot side, the hot side temp will start to bleed through. This will turn it into nothing but a glorified heater. Also, try not to run the peltier at the maximum 12 volts. Run it at say 9 volts. I used a dc to dc converter for this. I also insulated the cold side to prevent the ambient air from interfering with the performance of the cold side.
Better idea? Maybe: put the Peltier cooler on the loop between the rad and cpu, so the rad cools the hot cpu coolant and the peltier cools it down further? Worth a shot!
Try running through a rad to drop the temps to near ambient, then run it through the peltier to reduce temps even more.
I really think if you get some decent waterblocks and a couple better peltiers and but them in a series to cool the water, you could get some cool results
Nothing wrong with experimentation . That fan that you were printing seem intriguing . Add ,remove blades. Looks cool
This project can be done to create a water chiller. The main problem is having enough surface area on the cold side. And a massive slab of copper would be quite expensive right now. I actually had a idea similar to this utilizing and dual loop and a heat exchange block in between. Planned to cool the pelts with the alphacool cape cora which is a classic favorite of mine which I have bought. Never got around to the project / extensive video editing.
That's interesting cause by design, thanks to the tech, it should be able to go under ambiant temp! Try again with lots of Peltier in series !
peltier coolers were all the rage in early 2000's ...even had freezer towers as well :) however they were woefully inefficient
One thing pretty much always overlooked is that the peltier cooler will require more cooling than the cpu alone. You're effectively trying to dissipate the heat of the cpu, plus the heat created by the tech. Also, the bigger the temperature delta, the less efficient it gets, and rather rapidly. So good cooling of the tech is essential.
So with your setup here, you're effectively cooling the cpu with just the air cooler, plus several thermal transfer layers, plus a heater.
Your bottleneck is the noctua cooling the TEC. The generated heat must go somewhere so that the Pelt can sustain the diff in temperatures between cold and warm face.
You got a 12706. That means it will use 6 amps and remove about 50W of heat. If you use a 12715 it will use 15 amps and remove about 130W of heat. A 360mm radiator can easily dissipate 300W-600W of heat as long as you get it hot enough above ambient and have good fans.
Peltiers can move heat using electricity, but they're very bad heat conductors. This means that while yes, they can cool down things by transporting heat away form them, they do it at a very slow speed, therefore they only really work for things that don't generate heat by themselves, like the inside of a cooler, or a can of beverage. You would need a massive surface area covered with dozens (hundreds? thousands?) of Peltiers to cool down the wattage of a CPU!
possible with two TEC of good quality (Made in USA, not the Chinese ones with wrong specs) and a strong Qmax. Did it with three waterblocks, two to reject heat with water and the center block get cold from both side. Not power efficient!
The peltier chip doesn't directly transfer energy from the water block to the fin. If you're gonna use peltier chips with water blocks you want to use them to keep the liquid cool but they'll need their own way to exhaust heat.
I think the solution is a combination. Open loop, exhaust water from the cpu goes through a rad to cool down a bit, then, connect a series of cpu blocks with peltier/cooler modules, to further chill the water, then send it back to the cpu.
Use a rad to cool the water coming from the cpu, then a Peltier+air cooler to chill the water below ambient before it reaches the cpu's waterblock
I would be more interested how the radiator + tec in series would perform. The radiator would cool the water to ambient temperature and the tec could bring it below that. Additionally you could cool the hot side of the tec with your evaporative cooler.
2 x peltier to cool both sides, you sir are a mad scientist 🤣👍🏻
This is something to revisit when you can get your hands on an Ice Giant Prosiphon. And a better peltier. Or maybe stack peltiers.
the biggest issue with setup not even taking the peltier into account is the surface area of cooling. an air cooler and the fins of a radiator are basically similar in cooling capacity.
I guess now i'm curious to see a radiator replaced with a copper block with 3 air coolers tied to it.
I think part of the problem that it didn't work too well was because the air cooler was upside down, meaning the vapor chambers of the cooling system weren't properly utilized. Although, it is still a Peltier so... they're not too awesome if you can't disperse the load..
I know I'm super late to the party, but put that block in-line after your radiator. You stand a small chance of hitting sub-ambient temps. If you're using a $4 peltier its probably able to suck out about 15W of heat, at the absolute best. I've got two "225W" peltiers around here somewhere, and their curve shows they'd be able to move about 90-100W of power. If you wanted to cool a modern CPU you'd need about 3 of them. Also, keep in mind, if you do get sub-ambient you'll need to seal things up to keep condensation from dripping on your board.
Im no scientist but I think the reason this didn’t work is because the surface area of a water block is so much smaller than one of a radiator maybe if you put the then cooler between the cpu and block it would work better
how to make it work: you need a res with the peltier cooling water connected in parallel to the main loop with some kind of valve and thermostat control. the peltier comes on and cools a quantity of water as low as it can go. the thermostat opens allowing the cool water into the loop and new water into the res. repeat. only way to increase the heat capacity is more powerful peltier. if it could rapidly cool the fluid in the res (im thinking 5-10 seconds) and the res holds a similar quantity to the main loop you should be able to do some good cooling.
i did peltier cooling for a science experiment like 5 years ago. i got a monster of a pad like 300 watts and attached it to a h105 aio with some 9000rpm fans. i was able to keep a pentium 4 at 1c for extended periods. possibly lower since i dont think the motherboard read temperatures below that. loads of fun to do but 100% totally impractical. im surprised how well the motherboard held up considering i was a total noob at sub ambient cooling back then. i put kneading eraser around the socket to help but there was still so much frost buildup on the backside. i still have the cpu and motherboard and they still work!
I think using the radiator with the peltier would be interesting.
So as help for the radiator.
Thatd be cool
Try to incorporate it into a loop with radiator and place the thermoelectric heater before the cpu.
Now hook this up to your air-cooler-turned-liquid-cooler.
"Yo dawg, I heard you like coolers. So we put an air cooler on your liquid cooler on your air cooler"
You should try connecting the peltier to the exit of the radiator and run both of them together to see if it does make a difference
For starters, you've got it hot side down, heat rises, you're heating the cool side with the hot side that way. It would likely have worked far better without the water block, directly attached to the CPU with the same heatsink on the hot side and a high speed fan to cool the hot side. The cold side can get cold enough to freeze things, so if you're not getting enough cooling out of it, you're doing it wrong.
Should pick up one of those 200W Peltier plates, it should work then! Just gotta make sure the air cooler you're using is up to the task
I don't like the phrase "suck the heat out", peltier cooler doesn't do that, it actively cools the water by consuming power. So weather it's going to work or not purely depends on the cooling power of the peltier. You'd need it's cooling power to exceed the heating output of the CPU. The cooling power is not the actual power consumption of peltier, the actual power will be about 50% more as peltiers efficiency is about 65%.
Well, there's y'er problem!
The Peltier will fail. It's not if, it's when.
You'll need a custom water block for this one, and some mini radiator fins.
The cold plate of the water block contacts the CPU (with paste, of course). On top of the water block, protruding into the water block is the tec cooler with mini radiator fins poking into the water stream (would be cool if they were heat pipes lol -- See below for another idea). On top of the Peltier is an air cooler, of sorts, or whatever method you want to use to cool the Peltier.
The water cooler pump (not in the water block) circulates the water, and the water loop's radiator(s) cool the water.
Set a fan curve that stops the fans if the water temperature is at or goes below ambient. You don't want to pump heat into the system accidentally.
When the Peltier fails, the water loop can take over cooling until the Peltier recovers. Then start again.
Here's another idea (completely unrelated). What if radiator fins had several, small heat pipes to help spread the heat away from the water tube? That would increase the heat distribution on the fin allowing the fans to better move the heat away from the radiator. And what if the radiator fins, themselves, had curvey ridges? That would also increase surface area and time of contact of the air with the fins an transfer more heat away from the system. Wanna give it a go?
Now I have two questions:
-what would happen with the water lock directly attached to the air cooler? Would it be better than the radiator?
-could you chill the water by doing CPU->rad->Peltier? The rad would bring the water closer to ambient temp, then the Peltier could chill the water to maybe achieve sub ambient!
You need more peltier modules. They're not rated for much thermal transfer each, but with a few of them, they can get some work done.
Der8auer got peltiers to work. It took many of them in series, each with their own water blocks and heatsinks. Would recommend checking out those vids.
You would need something like a 250W Peltier to cool that cpu, efficiency is terrible on peltiers, and if it heats much, at certain temperature, it actually isolates the heat transfer.
Excellent video! Looking foward to the next!
Peltier cips are not designed to work under constant power. You should control them with PWM according to the delta temp between hot side and cold side. Also the water block should have a longer water loop inside due to the low thermal inertia.
Try attaching 3-6 to the rad, with those attached to air coolers. It may seem useless but "effective", but if you have it located in another room with thet tubes running from the PC to it (like mine was) it does a really good job at silent cooling with better stats than any AiO. Or rather than using a radiator, use a 100% metal waterblock. Usually used in scientific applications, but you can find long ones that'd they fit to on Amazon pretty easy.
I've wanted to see this done for ages. Awesome man
the peltier setups that tend to work usually have two loops, cold side and a hot side with multiple peltiers, its NOT efficient at all but hey, sub ambient is doable.
I believe it'd be more efficient to go Radiator > Peltier > CPU > back to rad for going sub-ambient. But yes, ya need a better water block to use with the peltier, maybe an aluminum based one.
More Peltiers and a larger waterblock to use with them for lower temps, but then I'd suggest monitoring the temp to ensure the liquid doesn't freeze.
Thermodynamics says that the cooling capacity of this loop is determined by the Noctua Cooler. There is nothing you can do to make it cool more than the capacity of the Noctua and any intermediate steps only add heat or reduce its dissipation. Its like plumbing an A/C unit back into the room it is cooling.
Peltiers that size are usually too wimpy for that, you need to be able to move the cpu heat and still maintain a negative temperature gradient, you could check that in the peltier's datasheet. Then you need to cool that from the peltier, but now you could have a grater temperature difference than the cpu. The best configuration for that would be pet directly on the cpu, waterblock to radiator. (you will need to dissipate the cpu heat plus whatever the pet is using to keep the gradient)
As pet aren't very efficient you would likely have more power to the peltier than to the cpu, so you are likely to have three times more heat than the cpu itself. For over 100W to an overclocked cpu you would easily need a 200+W pet that's the size of the cpu, good luck with that, and easily a 360 rad to keep all that heat out.
If you really want to pull this out using many peltiers in parallel should be used, but then you need a heat spreader, like 2 thick Aluminum plates with the pets in the middle, take the heat from the cpu to the cold side, and keep the hot side cold enough to give it a fighting chance.
Maybe add a big reservoir (bucket?), let it cool down over night below ambient, then run shortish benchmarks. You could call it a success for limited use cases. ... Like starting Word at 5.3 GHz or so.
multiple thin fans stacked on top of each other in a single frame with varrying blade sizes and numbers similar to a turbo fan compressor
Hi mate!, Would love to see a rad immediately following the CPU to dump most of the heat, then have a TEC to cool the water before it heads back to the CPU. TECs can create good deltas, but dont manage large heat loads well.
you should try ading that peltier + air cooler to the rad. loop to see if it make any difference versus the rad alone
This is pausible but single peltier won't cut it. Last year, German Overclocker der8auer tried the same thing by using 12 peltiers and water blocks in order to cool his system. He used cheap air coolers to cool down those peltiers.
TECs are such incredibly inefficient thermal devices that Linus Tech Tips had to use multiple 500 watt Peltier modules just to make a CPU reasonably sub ambient during idle, connected to heatsinks that would've cooled the CPU better under load directly anyway.
If Peltier cooling were in any way remotely efficient, fully silent air conditioners would be a thing, and cooling a CPU could be done in a much more compact form factor.
It's best you don't go down the peltier train, unless you're prepared for disappointment. Instead, let me offer you a quick fix idea: Same waterblock to air cooler idea, but cut the Peltier module out from between them. The TEC is actually working more as a thermal barrier, than it is as a thermal transfer accelerator, or otherwise a tool that makes the hot intake cold and the cooled finstack hot.
Waterblock to air cooler could genuinely work without the TEC in the way, though.
What I think is better is the Peltier element as a supplement to the cooling. I mean, cool the liquid down to ambient temperature with a radiator and cool down with the Peltier element below ambient temperature.
i been tryina tell linus and jayz two cents about doing this for 3 years now, but instead of just having the peltier you add a rad before it so it chills the water to the cpu and the heat gets dissipated through the rad