This Cooler DRAWS 545W!!? Bad Cooling Ideas #2 

+2 points for the solidworks flow demo and real engineering talk. More of that please tyvm.

Everything goes right - Linus: "See I told you!" Everything goes wrong - Linus: "See I told you!"

Alex: pays thousands of dollars towards an engineering degree Also Alex: Chop chop the bottom bit

I love Alex's face when Linus is speaking to the camera. The face of a kid who's waiting for their parent to finish talking with their friend.

You'd be better off putting the TEC cooling on the radiator to get lower than ambient, then you'd need another water cooler just to cool the TEC plate. You could get the water as cold as you want and not even worry about condensation.

"There's no way we can just finish like this" *Finishes just like this

Your peltie uses 900W to transfer heat. That means you have to be able to cool at least that amount with the water block. If you don't, both side start to heat.

16:10 their funnel has a label saying "FUNNEL"

I did my HS physics project on TECs and I found that one of the biggest problems with them is that their faces are not flat at all and since they are ceramic they are difficult to flatten so I had to compensate with globs of thermal paste and even then it was bad plus hot side temps matter almost as much as cold side since a high delta caused by insufficient cooling leads to heat leaking to the cold side lowering efficiency. Future improvement could be done by flattening the TEC with an end-mill or facing on a lathe, stacking TECs, and direct contact vs an additional copper plate(that slot for the thermal-couple wasn't helping anything.

Linus: "It was a learning experience" Alex: "Yeah" Linus: "About listening to your boss next time" . . . . . . Alex: "WHAT?"

I like to think that Linus knows all of his sponsors off by heart and all he needs is someone to tell him what the sponsors is.

May I suggest you try using a logic level mosfet next time instead of a solid state relay, Alex? Because as you probably found out during the making of this video, you cannot use a solid state relay to switch DC 🙄🤣 They can only be used to switch AC waveforms that cross through zero. A solid state relay (henceforth SSR) is nothing more than a triac with an additional zero crossing detection circuit. Although they are very different, the triac component in a SSR performs a function similar to that of a mosfet; where, for example, an active logic signal causes a current to flow. A very important difference between a mosfet and a triac is that once latched (and conductive) the triac cannot unlatch (and return to a state of isolation) unless the AC waveform crosses through zero, where as a mosfet can unlatch at will. This is the function of the zero crossing detection circuit; all it does is ensure that if the triacs logic signal becomes active somewhere halfway through the AC waveform, the actual latching of the triac is delayed until the waveform once again crosses through zero. Latching at any other point during the AC waveform has the serious potential of resulting in an enormous (potentially triac destroying) current surge immediately upon latching. Unless you're lucky, and the triac happens to latch precisely at the zero crossing point of the AC waveform, after which the current is allowed to rise gradually with the voltage. _[Important note: just as a triac is unsuitable for switching DC, a mosfet is unsuitable for switching AC!]_ I hope this clears things up a bit 😁 I know the chance you'll read this is so low I might as well be talking to a wall, but perhaps there is a slight chance another soul will come and stumble across this information and find it useful somehow. Who knows, they might potentially even learn something 😊 (If I made a mistake please correct me in a reply to this comment _in a friendly manner please_ and I'll happily modify this comment!)

"There's no way we can just finish like this" [HARD CUT TO SPONSORS]

I dunno but linus is addicted to pc cooling

I think it's hilarious that linus doesn't know who the sponsor is and then someone tells him and he always immediately knows what to say about about it.

18:42 My favorite part of the video, it was totally necessary!

For future reference, get a CO2 extinguisher. It will still put out the fire, but wont hose your expensive GPU.

Why not cool the water instead of the processor with a bunch of peltiers?

*Linus* - "tech cooling is and was a bad idea" *Intel / cooler master* - "Hold our beers"

Interesting. I'd often wondered about using a Peltier for PC cooling, but not directly on the CPU. My thought was to use it more passively, in conjunction with a fan at the front or rear of the case to cool the air flowing into the case by 5 or 10 degrees while drawing only minimal power. The idea being that if the air being supplied to your standard air- or water-based CPU cooling rig is slightly cooler than ambient, you would see some increase in the performance of your cooling rig overall. Kind of like keeping servers in an air conditioned room, but in this case using a little modular box on the back of the case to create the AC effect. It would be interesting to hear your thoughts on this, Linus.

Forget the thermoelectric effect. Hook that sucker up to the compressor on an actual AC unit.

Who buys an expensive tap then doesn't use a vice?

Instead of trying to cool with the peltier directly you should try to use a standard water cooled loop setup but after the water has cooled in the radiator it should pass true the peltier then and get cooled to ambient or even sub ambient.

"award winning templates" I 100% want to go to the awards ceremony for best web developing templates.

Linus: "we need it to spread the cooling" Alex: *silently cringing in the distance*

The relay itself is the problem: First, it's for AC power and expects the voltage to turn zero at some point. It turns off in the zero crossing. DC voltage never turns to zero, so the relay will never turn off. Also, it has a 1.6 volt voltage drop, so the peltier gets only 10.4 volts. Please use a MOSFET next time. It will work way better, and we get to see even more sketchy LTT electronics. I love these videos!

You're right about thermodynamics is in play here, specifically the latent heat of vaporization of H2O. There's an inherent heat storage capacity to water that isn't there for the Peltier device. There's also the phenomenon of the difference between laminar flow vs. turbulent flow of water in a heat exchanger. If you have laminar flow through your water block, you can generate micro bubbles of steam in the water; that steam infused water can then be pulled away from the water block's heat exchange surfaces. In turn, the micro bubbles of steam infused in the water collapse and spread and disperse the thermal energy in the water downstream from the water block. That's the "inherent" heat storage capacity of water I was referring to earlier. That's what makes water coolers superior to Peltier devices or tech cooler devices, imo.

LOL I haven't programmed a PID in two decades now, but looks like they're still just as PITA as ever. Sad to see it didn't even help. BUT (dun dun dun!) maybe putting the Peltier cooler on the CPU itself was the bad idea. What if it was used as a secondary cooling phase (so placed after the rad) to cool the water going into a normal CPU water block? So CPU - AIC - Peltier - CPU? So two sets of radiators. An AIC for the CPU and a massive rad for a janky Peltier. Arguably (not really) pointless as you COULD just use the massive rad alone to cool the CPU in a normal watercooled setup ... but where's the fun in doing things the sane way?

"There is no way we can just finish like this" DO YOU NEED A BEAUTIFUL WEBSITE WITHOUT ALL THAT HASSLE?

It’s rated at 32 amps, and we’re expecting 32 amps... Yeah, I’m just going to stand back here, behind the plexiglass, with the fire extinguishers. Always use things that are rated for a higher load than what you are expecting. Helps keep things from burning and exploding.

Have you never heard of the 437w Arctic Web TEC? I ran one for years on my C2Q9300 (and still have it). Used a heater core for a radiator with an electric car radiator fan running on 5v to cool it. Fed the 24v peltier with ~15v and got a below-freezing idle temp, with about 15C full-load temp with a 1Ghz overclock. Insulated the motherboard around the CPU (both front and back) as well as coating it with dielectric grease to protect the board from condensation.

Linus TEC Tips

20:52 this whole part seems very similar to the chernobyl accident...

Linus 2019: 'A PID controller is like an On/Off switch' Linus 2020: 'A computer is like a spinich'

Love that discussion about confidence at 16:21 :D

As suggested by some commenters, it would be way better to cool the water using the TEC and stay within its maximum efficiency zone. More power doesn't mean more cooling. You'll have to dump that heat the TEC is generating somewhere or it will start to heat up. Cool the CPU loop radiator with multiple TECs and cool the multiple TECs with another loop. That way the CPU block should receive cooler water and the TECs dump their heat into another radiator so it doesn't interfere with the CPU loop. In theory, this _should_ work as long as you don't overdrive the TECs.

Thought for sure they would cause a brownout on the west coast.

GEOTHERMAL cooled pc. It might be just crazy enough.

Watching back these older vids it's interesting to see how your production has improved; use of wide shots for conversations, audio mix with the background music being lower, and less filming of computer monitors to explain concepts all help the videos flow better, and probably make the editors jobs easier. Just goes to show that even the best still have room to improve. These videos are great, but your current stuff is even more refined.

The "Oh my god!" moment of the overflow was the first genuine laugh I have had in days.. wasn't watching LTT for that experience, but thank you, I needed it!

Overpowering peltier module raises its temperature instead of dropping it if heat removal isn't ideal.

Also, if you push too much heat into the TEC, it becomes an insulator.

The best method I have found linus is a thermostat plugging in the power supply for the peltier. Having the thermal couple in the water, preferably in the reservoir of your loop. Simple as setting a target temp for you water..

Truly hope this series keeps it up

Next video: Can you cool a PC with radiators from the ISS?

"we have a tap wrench." *Proceeds to use it without a vise.*

Now this are video's that i like to see! You just almost perfectly proofed the working of it and made and designed yourself! Loved it👌🏻👌🏻 We used 4 TECs with a thick copper plate on a oc'd Athlon 1700.. there is now way that one TEC is going to cool a cpu of this time. PS. 15 years ago I used to go to the junkyard, get a piece of copper, made it fit, bore holes and placed plexi with my own soldered leds in to it😁

Glad I found this video. I was about to try this. Saves me the trouble.

There's no way that little water flow can stop it from overhearting

Alex: There’s no way we can just finish like this. Linus: *smiles* *AD STARTS PLAYING*

I'll be honest, I think Whole Room Water Cooling was one of the best things ever, and I want a new version of it. I don't know what they could do at this point, but I think it was an awesome idea, and could absolutely work. I've seen similar setups work extremely well with only a couple computers, and LTT has the capacity and means to do a "whole room" unlike so many others. Some day I hope they do a 2.0, or have a reason to. Maybe just not the editors den..

I want another channel that is run by Alex where he just does design and engineering stuff in a lot of detail. It doesn't need to be this kind of camera or editing quality, I'd just like to see... DETAILS.

im getting whole room flashbacks with those fans and radiator, the cable mess and evrything..

"So the block design is working as intended- OH MY GOD WHAT IS GOING ON HERE!"-Linus in a nutshell

Beautiful design and test great video!!!

Did Peltier cooling back when it was needed, more than 10 years ago. Had one of the massive aluminum Lian Li cases with room for 12 hard drives on the bottom. After the cpu or gpu you send the water to a radiator. Which then feeds the reservoir. This brings the water to roughly room temp. Then you cool the water, with a different pump system cooling the peltiers. All in all, 3 x 3 fan radiators and 1 x 2 fan. With a 1 gallon res. 1250 watt EVO galaxy PSU and 450 watt Thermaltake side by side psu. Peltiers took about 600 W in total. Anyways, give the system the same cooling as a radiator first then cool it with peltiers. Don't remember the Intel CPU. but it was stock clocked at 3.2 with 4 cores(best cpu at the time was 3.6x4). With radiator alone water cooling could hit 4.2 stable. With peltiers the way I stated, it would go up to 4.8 stable and all the way to 5.4 (unstable). The only reason I would pull my old case out of storage and do this again, is I had money to throw away. Air cooled with a decent set up works today. These are no longer the days where Crysis made your system have a mental breakdown.

You know it's going to be a good video when Alex puts on is engineer uniform...

Tecs get less efficient as the power consumption goes up so cranking it up to 900w might not be as effective as keeping it at 200w. Also expect a 10c delta across the tec so for sub ambient multiple tecs should be used with a larger cooling system

I loved the ambition of the "wole-room" series. I mean you've come a long way in knowing why that can't work but it was still a brilliantly disastrous test.

I am designing a cooling system for a giant telescope mirror. I never thought that LTT would be a resource I would use to design the thing. You guys are the best!

10:56 "It only needs to be straight enough" - Linus, 2019

Alex&Linus videos always feel like a what-if articles in a video format. And I absolutely love every second of it.

Yeah, I like my NH-D9L. All of these TEC plate experiments and my experience using water cooling, it's hard to beat an efficient air cooler. I wasn't even a believer in the Noctua hype until I tried this one either.

So, I tried to do something similar a few years ago. I was really not up for condensation, but still wanted to get things as cool as possible. You can grab an arduino and a humidity + ambient temperature sensor and calculate the condensation threshold on the fly. Then use the arduino to talk to the relay/peltier. I'm not sure what the word for it is, but my problem was "thermal momentum" (my name for it) where you would have things snowball so fast that it would pass the condensation point. I gave up when I had to pull out my old calculus books to start predicting where the snowball was going.

Thought I'd see my *grandkids* before this video.

Luke was the king of janky projects

A couple of tips: 1. you could have just measured the temperature at the edge of the cold plate, as that's the area where condensation would be a concern - you can solder the junction of a thermocouple directly to a metallic surface so long as your TC readout has an isolated input or the cold plate has no conductive path to ground. 2. As others have mentioned, an overpowerd TEC just becomes a resistive heater... you have to do testing to find that point and target a power somewhat below that. 3. If you're going through all the effort of designing custom blocks, you might as well add a dry gas purge around the cold plate to go as cold as you want without worrying about condensation 4. As others have mentioned, you'd be far better served by having "hot" and "cold" loops interfaced by the TEC than trying to dump the TEC directly onto the cpu.

The PID controllers are pretty good things. Often used in labs and metalworking industry to control the temperature in muffle furnaces. Heat treatment for steel or otherwise simply maintaining a stable unchanging temperature within the chamber. BTW, that's why it doesn't read negative temperatures, it's intended for high temperatures, such as from the 600's to over 1k Celsius.

Alex is a genius! This is exactly what LTT should be all about! MAKE IT WORK!

Man, they should do some of this awesome invention videos with ElectroBOOM, it would be awesome.

16:12 Oh yeah I remember that haha, that was quite a leak for me because I was smart and forgot to put the fitting on the top of a rad because "it'll never make it through there" It did

EKWB finally released a good version. comes out next month. It was the "mystery box" linus used to get the 10900k to 5.7 ghz. EK quantumX Delta TEC cooler.

if you do water cooling aswell, then I would rather use the Peltier to cool the water and not the chip directly

Linus: But hey it was a learning experience. Alex: Yeah. Linus: About listening to your boss next time. Alex: *WHAT?*

Loved this episode, I tried this back in 2006, pretty much did exactly what was done here minus the temp regulator and my water loop started boiling, The pipe burst off the radiator and that was the end of that idea. I also first tried with a tiny peltier cooler which was useless but then ordered an industrial grade peltier and the outcome was the boiling water loop as I mentioned. It would be awesome if this kind of technology actually worked.

I just like that they connected PID to a relay.

Why not use the peltier as a "radiator", cooling the water, and not directly on the processor?

13:00 you could just flip the polarity

Go one step further. Standard water block -> large radiator to bring temp to ambient -> tec cooling array to bring coolant subzero -> back to water block. Fyi, peltiers have three energy balance equations including internal resistive heating, conductive thermal transfer and seebeck thermal transfer. As your temp delta goes up the tec efficiency drops and the resistive heating takes over resulting in your tech becoming predominantly a heater on both sides. Amping up the voltage and current just drives more resistive heat load. Also as the temperature delta increases, a reverse voltage is generated which subsequently further drops the seebeck heat transfer performance. Fun times 🤠

Generally with a TEC set up you use it to chill the water in the loop. Then pop an air cooler on the TEC. I've done it, worked great for me.

17:45 "Don't touch that!" D:

"If I have to blast a TITAN RTX I'm gonna be so pissed off" I'm sure there is a cheap as crap PCIe graphics card somewhere in the warehouse they can use until they are satisfied it's booting correctly before testing further

Always a great video Linus. Never give up,and I hope you continue to find great ideas for videos without losing your passion for pcs, and stuff.

I think people are using peltiers the wrong way. We are already good at getting water back to room temperature, we should keep our existing loops and focus on making the water that's going *into* the block cooler. if you make a reservoir that runs water over the cold side of a peltier there are a couple things that you gain: 1. The ability to only turn on the peltier when it's really necessary (Example, when reservoir temp is 5 degrees above ambient turn on turbo charge mode) 2. The ability to pick and choose a module that isn't quite as overpowered. You don't *need* the module to move the entire 300w of heat (or whatever). You could pick a 200 watt peltier that moves 50w of watts of heat and still get a massive performance increase. 3. failsafe. If the peltier dies or whatever, you have still have the vanilla loop.

long story short, linus tech tips still doesn't understand TEC's.

Der Bauer build a very capable TEC-cooler years ago. Maybe ask him for advice

Linus: "TEC cooling is and was a bad idea." Intel: "Hold my Beer..."

"We need to hit this thing with a Load." -Linus

Everyone gangsta till the cooler draws 696969 watts

1. If you don't use the right ssr, it won't turn off when using dc 2. You want a T type tc to measure sub freezing temps 3. look at the ssr derating curve. they only can handle their max amps when they are cool

Even though it didn't work very well i think this is my favorite episode

There's some more that goes into PID. Basically, instead of full on when the desired temperature is not reached and full off when the desired temperature is reached, it tries to predict when its about to reach the desired temperature and slow down so it won't overshoot or speed up so it doesn't undershoot. That way it pretty much stays exactly on point, there is no oscillation in the temperature.

"It's probably the jankiest thing we've ever made" Alex just reminding us he wasn't around for Scrapyard Wars.

5:05 on the bottom right side was really disappointing.

I run multiple peltier devices in series, thats also inline with a corsair 115i, to reduce my coolant temperature. As always, the peltier devices are not at all efficient, but its a nice little assist to the corsair 115i. Plus it was a lot of fun to build.

All this time I thought that Linus was using pre-recorded ad messages but because of the echo in the room you can tell he says it on the spot. Neat.

Man, alex's stuff is getting super impressive. Kudos to you guys

The amount of shortcuts taken in this video is unreal

Linus I love your content it really relaxes me and rest appreciate it a lot

i actualy ran a dual CPU system with this setup around 1999-2000. i used two celeron 400 (socket 370) a BP6 motherboard and two 55 watt coolers running flat out 24/7. you needed to have "GOOD" cooling. i used two Alpha PAL6035 running full blast. as long as the room i was in this worked just dandy. the problem came when i arranged this LAN party and the climate control in the building was set to turn of at night. and it became 30c+ at the second level. then i did not get the air out from between the gfx card and the PSU stack (i had to run two) so cardbord and gafa was used to direct the fan flow inside the case and i was ones again able to hold my hand on the case lol. btw.. you will ned a LOT of insulation to prevent condensation. as well. i used foam cut and stacked around the socket and the lower part of the cooler. for you to make this work your biggest enemy is heat on the hot side. i think if you made the ribs in the water block 2-3 times higher AND square (have the ribs covering ALL of the cooler element) and used two in and two out tubes + dual pumps it might work.. remember you MUST be able to remove 2000 watt+ and hold it under 35c. you can only expect a temp difference of 35-45c at this level on the hot and cold side. might go doen in the 20's. so the hot side MUST be kept cool!