Our best effort still SUCKS - Sketchy Heatsinks 3

[Granular]

20 hours ago, AlexTheGreatish said:

.......The best heatsink is an NH-D15, you need heatpipes.

LGA 1151 mounting holes are 75 mm apart.

Imagine a copper beam with an 80mm square cross-section and a length of 0.5 meters.

If one end of this beam is 50 degrees C above ambient and the other is at ambient (due to convection cooling), then the power conducted away from the hot end is at least

heat conductivity of copper * area of square * temperature differential / length of beam

or 386*0.08^2*50/0.5=247W

And okay, the CPU package is a 3.75 cm square and the silicon inside it is smaller still, so it's not going to be keeping the cores of a 247W CPU at only 50C above ambient and it would weigh 29kg, which works out to something around 200 USD, but I think it would do pretty good of a job cooling a 105W heat load, it would be fanless and stupid simple to make.

[JTrevail]

12 hours ago, darknessblade said:

On 7/18/2018 at 4:39 PM, JTrevail said:

That garbage you linked is only 1/3 copper by weight. You would have a heatsink that performed only marginally better that plastic

if you say that this is garbage why did you not give a link to better copper/metalic filament???

I don't think there is a filament worth consideration. The nearest you can get are powder based 3D printing technologies; those produce parts that are brittle and have poor thermal conductivity. Powder metallurgy can also produce porous parts, which would also make them bad for water cooling.

[bene20080]

Advantages:

-cheap (roughly 30 Dollars)

-should cool good, because it has at least 10 times the surface to dissipate heat, than the 212 EVO. Also the water does a good job in transporting the heat to the walls of the tubings and bucket.

-Can save a lot of heat (It takes roughly half an hour to heat 25l of water from 20°C to 60, assuming that there is no convection to the surrondings)

 

Disadvantages:

-danger of leakage

-ugly bucket is always standing in the way

-may be loud

-may forget to start the pump, before starting the computer

 

Ideas for improvement:

-long tubing with good heat dissipation, through the use of the right material and airflow.

-isolating tubing inside the flat, dissipating the heat outside the flat

-Make artificial roughness at the water inlet/walls in the cube above the CPU to make sure, that there is turbulent flow(better heat transfer)

[Tom is a Door]

I have a 212 evo laying around. Lost most of the mounting stuff. Wondering if I should get some zip ties and tie on top of my laptop gpu

[baRRacudax]

I think Linus should sell a few gtx 1080 TIs and buy a mill.

Watercooling is pretty easy to DIY and cheaper as that heatpipe mess they created.

20€ (new) VW radiator
10€ Aluminum, enough for 5 Waterblocks
10€ SunSun pump

One afternoon milling, drilling and cutting threads
(still had fittings, tubing, screws etc)

 

 

Was capable of cooling 3x AMD 16C Bulldozer Opterons (OCed) and 2x 6C LGA 1366 Xeons, with a notebook cooling pad strapped to the radiator. That's what i call bang for the buck

just had to sign up and post this after seeing the video

[JTrevail]

On 24/07/2018 at 12:51 PM, baRRacudax said:

I think Linus should sell a few gtx 1080 TIs and buy a mill.

Agreed; he did say on a recent studio tour that they are adding to the shop. Personally, I would love a HAAS VF-1.

[darknessblade]

On 7/24/2018 at 6:51 PM, baRRacudax said:

I think Linus should sell a few gtx 1080 TIs and buy a mill.

Watercooling is pretty easy to DIY and cheaper as that heatpipe mess they created.

20€ (new) VW radiator
10€ Aluminum, enough for 5 Waterblocks
10€ SunSun pump

One afternoon milling, drilling and cutting threads
(still had fittings, tubing, screws etc)

 

Was capable of cooling 3x AMD 16C Bulldozer Opterons (OCed) and 2x 6C LGA 1366 Xeons, with a notebook cooling pad strapped to the radiator. That's what i call bang for the buck

just had to sign up and post this after seeing the video

 

the only problem with your design is leaking. since 1 mistake with the design or 1 imperfection in the aluminium could leak water/liquid out of the seams.

 

it does need to have a lot of test runs before you can create a working heatsink that dissipates enough heat

[AlexTheGreatish]

17 hours ago, JTrevail said:

Personally, I would love a HAAS VF-1.

Lol I wish, those things start at like 50K.  I'd much rather have a good manual mill and lathe, and a smaller CNC mill

[mrstochastic]

48 minutes ago, AlexTheGreatish said:

Lol I wish, those things start at like 50K.  I'd much rather have a good manual mill and lathe, and a smaller CNC mill

I would agree, a used manual mill with a digital readout and a simple used lathe (with a spindle bore) is a good base for fabshop. If you aren't trying to hit tight tolerances and don't care too much about surface finish (can always polish matting faces) you can get away with a drill press (that can get down to 400 RPM) and a translating table off amazon for hobby materials (plastic, AL, wood, etc) in place an actual mill. Just a cheap bandsaw, desk grinder would have made the fab of your last heat skin much more pleasant to watch. My previous life as a mech eng I can feel the struggle trying to build something like that without the "proper" equipment. 

 

Speaking of heat sinks, if you are accepting ideas: Ice bucket water cooler, Peltier heatsink stack, and last one, sleeved cylindrical fin extrusion hooked up to the dyson (inlet at the at bottom, exhaust out the dyson).

[MattMatt]

Having heard snippets about Alex's background I believe he/you is told to make things sketchy just for entertainment purposes and for the most part it is funny but it would be good to see a real DIY engineer's attempt at something rather than Linus's idea of maximum views for minimum cost.

[baRRacudax]

12 hours ago, darknessblade said:

 

the only problem with your design is leaking. since 1 mistake with the design or 1 imperfection in the aluminium could leak water/liquid out of the seams.

 

it does need to have a lot of test runs before you can create a working heatsink that dissipates enough heat

It's sealed up with HT silicon. That stuff can keep my crankcase sealed so no problems on a waterblock. I also could have used gaskets but i didn't feel like spending so much time cutting them out. btw the system ran 24/7 for about 6 months without losing a drop (at least not on the waterblocks, the cheap chinese pump is another story)

 

1 hour ago, MattMatt said:

Having heard snippets about Alex's background I believe he/you is told to make things sketchy just for entertainment purposes and for the most part it is funny but it would be good to see a real DIY engineer's attempt at something rather than Linus's idea of maximum views for minimum cost.

I'd love to see more DIY engineering stuff, but i can't really imagine more serious projects with the equipment i've seen in the LTT workshop.

I'm planing on doing a Laptop heatsink next week. Maybe i'll post something on this forum, if i don't mess it terribly up...

[AlexTheGreatish]

6 minutes ago, baRRacudax said:

I'd love to see more DIY engineering stuff, but i can't really imagine more serious projects with the equipment i've seen in the LTT workshop.

I'm planing on doing a Laptop heatsink next week. Maybe i'll post something on this forum, if i don't mess it terribly up...

Oh there is more coming, we're just getting started haha - someday there will be a shop to take on Adam Savage's here

 

Defs post it to the forum, sounds interesting to me

[darknessblade]

14 hours ago, AlexTheGreatish said:

Oh there is more coming, we're just getting started haha - someday there will be a shop to take on Adam Savage's here

 

Defs post it to the forum, sounds interesting to me

why not use liquid nitrogen to cool down a pc sooo cold that the CPU will crack in half of the cold. (i think using C2Duo's are the perfect solution since they are cheap to buy, and you can do multiple runs for more fun)

 

or have a server PC with 4 CPU's run so hot that you can bake meat on the CPU's  cheap diy grill

[JTrevail]

On 7/26/2018 at 12:58 PM, AlexTheGreatish said:

Lol I wish, those things start at like 50K.  I'd much rather have a good manual mill and lathe, and a smaller CNC mill

One can dream. 50k is great for a new CNC mill though.

[JTrevail]

On 7/26/2018 at 2:10 PM, ipseity said:

'My previous life as a mech eng' ...

I'm programmed and ran HAAS mills in a prototyping shop for years. I have one more year at Algonquin for mechanical engineering technology, so I fully intend to put that experience to use as I design parts. It really helps to know HOW something can be easily manufactured.

 

A pedestal grinder, surface grinder, bandsaw, manual lathe, and a good CNC mill would be my dream shop.

I would drive a beater to achieve that dream someday.

[Mjohns2002]

You should just cut a chunk off a cars radiator and try it as a heat sink.... already has copper pipes, fins, and probably made out of aluminum. 

[Holski77]

These heat sink discussions have been on my mind since I saw the first episode where they tried to cast a heat sink. I was really hoping to see some new heat sink designs instead of trying to make a traditional heat sink design, that are constrained by a few production methods available like CNC and stamped ans casted metal. I have had some thoughts where you could print a heat sink with baffles opposing each other, so that pressure gradients created by the computer fans could create turbulence and encourage more heat transfer (not sure if this would be better than laminar). I made a ginormous model with this concept that is meant to be used with two 120mm fans, just like some of the other extreme coolers.

 

Because this heat sink is so huge, it is very expensive to print in metal. Way to expensive for LTT to afford.

Also, printing the heat sink in steal is cheaper than printing it in aluminum, at least of Shapeways, which changes things.

https://www.shapeways.com/product/EZZD8CUE9/dual120mm?optionId=69193260&li=shop-inventory

 

 

 

I also made a small single fan heat sink. This sits a 80mm fan on top, and has a duct below the fan that directs the air into the middle of the heat sink. Once in the center of the heat sink the only way for the air to escape is underneath the ducting for the fan, and above the base of the heat sink. To go this route the air must escape through a fractal of baffles. I made the baffles with a Fibonacci spiral, so that the baffles look like they are patterned like a sunflower seed.

This heat sink is much more affordable compared to the larger one above, and is slightly cooler to look at.

Most importantly I am sure if LTT could actually afford to order this print (and maybe have some money left over afterwards to finally order some decent computer hardware).

https://www.shapeways.com/product/M9ZDGCFNP/ducted-fan-cooler-with-fractal-baffles?optionId=69193814&li=shop-inventory

 

 

Let me know if you guys want me to model any other heat sink ideas!

I'd like to make one with fractal branches that grow like a tree.

What do you guys think?

Is there a better place to print the model than Shapeways?

 

OH BTW the heat block is meant to mount to LGA1155.

[AlexTheGreatish]

3 minutes ago, Holski77 said:

~snip~

The problem with 3D printed metal is that it has really bad thermal conductivity, and even worse if it is made from steel.  I think it would be a lot more likely that we'd create a block from a solid block of Al using a CNC, although using a CNC feels almost against the spirit of Sketchy Heatsinks.

[AlexTheGreatish]

6 minutes ago, Holski77 said:

I'd like to make one with fractal branches that grow like a tree.

What do you guys think?

I've thought of this too, I wonder if it would be possible by using a CNC machine since 3D printed metal doesn't really work.

[Razor Blade]

@AlexTheGreatish is this legit?

 

 

Would be nice to see a sketchy heat sink episode with a right skookum choocher... I know he has done a collab before.

 

[Holski77]

37 minutes ago, AlexTheGreatish said:

I've thought of this too, I wonder if it would be possible by using a CNC machine since 3D printed metal doesn't really work.

Alex, if you didn't have such a propensity to order extra water pumps then maybe you would have enough in the budget left over to at least try!

Not all 3d printed metal is the same BTW, increasing sintering temp can increase thermal conductivity or 3d printed parts.

Anyway, there is no other easy way to make these heat sink designs without 3d printing tech. You could still experiment with the design by testing the heat sink and then testing to see what the thermal conductivity values were. Then you could do simple calculations to guess how effective a solid heat sink was.

 

I guess you could CNC the small heat sink though. It would have to be two parts, but you could machine the baffles onto the heat block, only the ducting would be a separate parts that you could screw on.

[JTrevail]

A 3D printed heatsink would make a good episode for sketchy heatsinks, emphasis on the SKETCH.

 

If a design can only be 3D printed, it's not very useful IRL.

[passwordisTAC0]

On 7/26/2018 at 5:34 PM, AlexTheGreatish said:

Oh there is more coming, we're just getting started haha - someday there will be a shop to take on Adam Savage's here

 

Defs post it to the forum, sounds interesting to me

learning G code can get tricky. guid video coming?

[Nomecone]

Researched this today. And first post on a necro thread. Don't care.

 

Great first try with better tools. The struggle of every DIY of realizing I need more or different tools than what I already bought to finish this job the way I want. Then compound that with I have to finish this now with what I have because this is the time slot I have. 

 

I think if you kept going at attempts you could get to close to a H212 with H212 layout or if you did dual tower like Noctua/similar you could eventually get that level of performance DIY. 

 

To move beyond the performance of retail parts the design would have to have something new/different they don't have. Not necessarily novel. Could be something the manufacturers don't do because of cost, assembly cycle time, assembly difficulty, etc. So materials, methods, alternative designs. 

 

I don't work in a computer field but I do design engineering and everything I have every worked on has compromised peak performance for cost and ease of assembly. I always want peak performance and the joke I used to make after getting assigned a new project was sending a Rambo "Do we get to win this time" meme to close colleagues on the same project. The answer to that is No by the way.