Our DIY CPU Chiller From AliExpress is RIDICULOUS

Ryzen 7000 just launched which means one thing - we need to sub-zero it, and this time we're doing it on the cheap.

Purchases made through some store links may provide some compensation to Linus Media Group.

FOLLOW US
---------------------------------------------------

MUSIC CREDIT
---------------------------------------------------
Intro: Laszlo - Supernova

Outro: Approaching Nirvana - Sugar High

CHAPTERS
---------------------------------------------------
0:00 - A terrible idea
0:45 - Zoho MarketingPlus!
0:57 - LTT Intro
1:07 - How TECs Work
2:29 - Mounting 20 TECs
4:22 - Power, manifolds, tubing, soldering
7:02 - Filling hot side
12:04 - Filling cold side feat. Antifreeze
14:21 - Subzero Prep
16:05 - Overclocking
19:03 - Backblaze
19:53 - Outro

[tayloraldridge]

Alex's cooling ideas are honestly my favorite videos

[tronique]

The TECs did not draw the full amount of power expected because the supply voltage was too low. Looking at the data sheet for TEC1-12710, these modules will need up to 17V, but the supply used was only 12V.

[IAmCoopa]

As an engineer I absolutely love all of this whacky engineering content, I would 100% love to watch longer format videos that keep dive into some of this stuff more.

[Daniel-hjjt]

I spent a couple of years playing with TEC's, successfully building out a universal whole case cooler that dropped the CPU temp by 20c while maintaining a case temp of around 10c. So a few things, mounting pressure on the TEC plate is EXTREMELY important, you need even pressure across both sides or performance takes a serious hit. We ended up threading out own bolts with a very high thread density, counting the turns and confirming with calipers. You also need to have a neoprene gasket (cut a hole for the TEC and sandwich between the heatsinks) we used quarter inch high density neoprene, but your application may need something different.


The other important thing, you're much better off using high power TEC's like the 600w we used, and undervolting it. At lower power, TEC plates are much more efficient. Basically you're better off using 8 600w TEC plates, running 75w each than you are using 1 plate running full load. I would have to check my 15 year old forum posts, but i'm pretty sure we found that 30% load was the sweet spot in terms of efficiency, though i'm really not sure if that's a flat metric or simply applied to our 600w plates.

Either way, i'd suggest taking another run at this, you should be able to get much better results using much fewer parts and a less janky set up.

[zejjnt]

I absolutely love these things that us regular nerds just dream of like "this idea is dumb, I want to test this" but you guys actually get to do it :D

[StJimmy]

The Alex and Jake combo is pure gold definitely would love to see more of them together

[MattyRRG]

We need more Alex and Jake, such a great duo, both of them are smart and janky at the same time

[Wolfesbrain]

I love all of these "technically feasible but practically inadvisable" projects. Give Alex's mad science lab a bigger budget!

[Nils__pltr]

I think the whole bubble trapping thing could've been solved if they mounted each row in series and not all of them in parallel but i still love the idea!

[eduardocarrasco]

Alex, the way you connected the pumps is not giving you double the head pressure, you are basically connecting them in tandem. This mean that one pump discharges just to feed the suction tank to the next. There is not pressure transmission doing that, so basically is just the last pump working, the first is just feeding the second tank, that’s it. If you want to connect the pumps in series to double the head pressure you need to connect directly the discharge side of pump 1 to the inlet side of pump 2. If you want to double the flow, then both pumps must be connected to a discharge pipe at the same time, but the total head is equivalent to one pump.

[Steamrick]

I really hope that there's a followup video where Alex fixes the worst issues with this cooling apparatus... there's soooo much room for optimization.

[Nickgowans]

Alex went from nervous engineering nerd to full on video presenter in a few short years. It's a beautiful thing to see

[milktaster]

Alex doing silly cooling projects are genuinely my favorite videos on the channel, I hope we see more in the future.

[indigenous.rabbit]

I believe you ran the cooling to the peltiers in parallel, that would completely explain why the coolant was not getting there. Water always takes the easiest route so there was no need for it to go through the top ones, it could just go through the bottom ones. Also rethink the diameter or your water tubes, the tubes going to the waterblocks were really thick and the one going to the pumps was only a little thicker. No wonder there was almost no flow in some areas. You would probably be fine with those flimsy 5mm ish tubes going to the individual units on the hot side. Maybe for a second iteration of this split the cooling system up into groups of like 4 peltiers cooled in series with smaller separate radiators. Also have the water in the cold loop flow the other direction from that in the hot loop to make sure the water doesn't get reheated by the higher temperature peltiers (because they are in series the last one gets a lot warmer cooling water and thus will cool less far on the cold side). I think by hooking up the cold side in groups of 4 (or so) with their own pump and combining the output with a splitter to the CPU plus the aforementioned changes you could optimize this setup.

[NonSensicalCharacter]

I feel like in a previous video, it was established that if you want to do this, you should first run the hot water through a radiator to get the temperature down for most of the way and then run it past the TECs to go sub-ambient.

[Princess_kitty]

Alex was so preoccupied with whether or not he could, he didn't stop to think if he should

[kanetw_]

1. Use a few distribution blocks (eg Phoenix PTFIX) instead of soldering 80 wires. Or good old Wagos. You'll hate yourself less.
2. I would do a 5s4p connection instead of 20p to save on tubing effort while having some redundancy.
E: any manifold connection has potential for unequal flow so you need to watch out for that. Big benefit of series tubing is you have the same flow rate everywhere.

[koalaunknown]

I would love to see a sequel to this that uses the same idea but in a more optimized way.

[chod]

You need a separate res at the top of the loop for this to work. pump at the bottom, res at the top, then you're not fighting gravity. come on guys this is water loops 101!

[ral-xkqb]

12:14 "The D5 pump loosener has escaped prison and the police are on hunt to find him!"