DIY: Cooling Head for Intel® NUC

[omnium]

see the extreme good results of a 'dirty/crude' Intel NUC Skull Canyon cooling mod Intel Nuc Skull Canyon 6I7KYK cooling mod with the beloved Noctua NH-L9i..



..this solution is crude and dirty but effective...I love it and it encourages me to try such an easy way with the Macho 90...and I'm very excited about what's coming...

Datenschutzhinweis für Youtube

An dieser Stelle möchten wir Ihnen ein Youtube-Video zeigen. Ihre Daten zu schützen, liegt uns aber am Herzen: Youtube setzt durch das Einbinden und Abspielen Cookies auf ihrem Rechner, mit welchen sie eventuell getracked werden können. Wenn Sie dies zulassen möchten, klicken Sie einfach auf den Play-Button. Das Video wird anschließend geladen und danach abgespielt.

Youtube Videos ab jetzt direkt anzeigen

Video anzeigen

 

[mamatteo]

Staying below 55c with this setup and intel stress test

 

[omnium]

please say something about the noise while stresstest compared to the stock NUC...lower temps mean better score...which score did you get with the benchmark...i get 1224 Marks with the stock NUC.

 

[mamatteo]

This time noctua fan with 5v only... can’t even hear it at all( what I wanted)

Actually I am getting only 1098 marks... why is that?




- - - Updated - - -

That’s with the stock fan.... 1090 and much higher Max temp of 85C!

btw I have undervolted it.. maybe that’s why?

 

[omnium]

Actually I am getting only 1098 marks... why is that?

sorry, I don't know why...you have undervolted the CPU? I did it again just now and I 'only' get 1157 Mark..so what? let's stop 'Schwanzvergleich' You have a brillant result with the Noctua and you can't hear anything...you've reached your goal I'm not there yet.
.
did the plug of the Noctu fan fit to the motherboard or where did you get the power from?

 

[mamatteo]

Actually I don’t even care about this Benchmark wasn’t my plan at all to compare results...
Having so much fun building a small high perf box

BTW: an intel guy answered your question... no standalone cpu heatsink from them


Quickly soldered a connection with pwm and sense.... works wonderful

here some images via abload:

 

[omnium]

Actually I don’t even care about this Benchmark wasn’t my plan at all to compare results...

yes, you are right...i asked you for the Benchresult...sorry...pure curiosity...because of that statement:

▲ Intel Extreme Tuning utility software test score of 1276 points, limited by the heat sink...

Having so much fun building a small high perf box
.
Quickly soldered a connection with pwm and sense.... works wonderful

BTW: an intel guy answered your question... no standalone cpu heatsink from them

Okay, I could have guessed.

 

[thorosoe]

What I would find interesting would be a reliable and clean way to mount third party coolers reliably. Although I am still waiting for my NUC8I7BEH and don't know how it will handle my typical use in terms of noise, I very much want to have an upgrade path for cooling that thing in an acceptably silent way (most likely big air with very low revolutions). Any progress on that front? Keeping up some kind of airflow in the lower part of the case would be very desirable to keep RAM and M2 SSD at acceptable temperatures.

 

[omnium]

The idea is to make a copperblock (size of the NUC CPU-die) which flush with a universal/special lid, made of aluminium. Then it will/should be easy in the future, to test/use other/new/custom made/selfmade heatsinks/cooling heads on this NUC test base.

my fanless solution will be universal...but i am still not very happy with the RAM/MB temperatures...while heavy load. There seems to be only one way: the cooler the cpu stays, the cooler the RAM and motherboard stay...motherboard and ram should not get hotter than 60° so the CPU must not get hotter!

I very much want to have an upgrade path for cooling that thing in an acceptably silent way (most likely big air with very low revolutions). Any progress on that front? Keeping up some kind of airflow in the lower part of the case would be very desirable to keep RAM and M2 SSD at acceptable temperatures.

go mamatteos path!

 

[thorosoe]

Did you try to increase airflow around the RAM and SSD? Maybe some kind of chimney-solution would help: build a (mostly) airtight enclosure around the lower part of your big passive cooler (with some room to the sides). The ascending warm air should draw at least some air from the lower part of that case (there already is room for that between the mainboard and the case) and that in turn would draw in cooler air into the case itself for the mainboard components.

For a quick and dirty trial run a duct tape ring around the top of the case and your passive cooler might do. Stick it for one or two centimeters of its width on the case so a few centimeters of free duct tape form a makeshift chimney. In case you are using the big tower cooler, a second round of tape might be needed to get enough hight. It needs to enclose warm air for it to ascent end suck up air from below.

 

[omnium]

Thanks for your suggestions. I will experiment in this direction...although I'm convinced that it won't do any good. I need to keep the CPU temperature as low as possible everything else has a low/no effect.

 

[thorosoe]

Sure, the main load on the CPU will be most important. But the hope is that you will find a combination, that allows for that few extra percent of performance, that will help making your system futureproof

 

[qwaarjet]

How does the NUC8i5 compare to Skull Canyon for those who have it? Pretty similar/even but at a lower power draw? Great work, I'm wanting to do something similar once mine comes in.

 

[omnium]

Thank you! You are welcome. Sorry I don't know, if the Skull Canyon is better/in line with the NUC8i5 which is relatively quiet, draws low power and is very snappy...for me.

 

[qwaarjet]

Here is my XTU benchmark with stock TIM and fan on "cooling" profile and an undervolt of -.09
https://i.imgur.com/bBYytHy.png

Good number but very bad temps. I will be replacing thermal compound and thermal pads this week and see if it makes a difference. Omnium, I look forward to your finished build. I hope to get some good ideas so cool this little monster down. Thanks again!

 

[omnium]

I will be replacing thermal compound and thermal pads this week and see if it makes a difference.

I'm curious, but I guess the difference is marginal. The short-term (turbo) 90°C is absolutely no problem...interesting is the temperature at long-term full load.

Source: Dan's Data: Thermal transfer compound comparison​

 

[thorosoe]

Funny test. Very entertaining read, thank you. Mine is not as entertaining, but there are tests that show some difference (that seems to be consistant with other tests): Best Thermal Paste for CPUS and GPUs - 85 Products Tested, Compared

One thing to note is that it is pretty much useless to draw detailed conclusions from a total swap of thermal compound and cooling pads from the factory default to some custom solution. Just the pads being a bit thinner and allowing the CPU to have better contact with the cooler has a way bigger impact on cooling than the thermal compound used. So the only conclusion can be: Yepp, this works better or it doesn't. The why is up in the sky with so many variables ... Still should be worth a try. Please share what you used, especially the thickness of the thermal pads!

 

[omnium]

Q: should I or should I not?

How do you think the MB and RAM temperatures will change, after I cut off the heatpipes along the red line and than put the heatsink (Thermalright Macho 90/Arctic Alpine) on the heatspreader? Who dares to give a prognosis about a possible temperature change?

I think the heat radiation of the heatpipes with the little heatsink onto the MB with RAM etc., is not negligible...but it will probably only take a little longer for the CPU and MB/RAM temperatures to approach, as there is no additional airflow...shit! I have to try that, but admit that it's hard for me not to have a way back to the original state afterwards. But first I'll make a try without the black cover (if I can remove it somehow) of the cooling fins and provide convection...maybe that is sufficient and i should not worry about ~75°C MB temperature after a long time of full load without any throttling...read this review..>100°C for the MB with throttling..

Please share what you used, especially the thickness of the thermal pads!

I measured it with a caliper: 0.5mm is good.

@mamatteo
What about your project? You're making progress?

 

[Dr.House3]

What about custom water cooling, by using custom waterblock and an external pump and radiator ?
So you only have to drill 2 holes in the case for the pipes.

That would be the best solution, but also the most expensive one.

 

[omnium]

I'm sure a project like this with liquid cooling will come, but not from me..my options are limited.

Datenschutzhinweis für Youtube

An dieser Stelle möchten wir Ihnen ein Youtube-Video zeigen. Ihre Daten zu schützen, liegt uns aber am Herzen: Youtube setzt durch das Einbinden und Abspielen Cookies auf ihrem Rechner, mit welchen sie eventuell getracked werden können. Wenn Sie dies zulassen möchten, klicken Sie einfach auf den Play-Button. Das Video wird anschließend geladen und danach abgespielt.

Youtube Videos ab jetzt direkt anzeigen

Video anzeigen

This project is a one-off project, but I'm trying to make sure everyone can recreate it.

 

[omnium]

#Milestone: Cirrus7/Mr. Engemann has a heart for tinkerer

Mr. Engemann, the boss of Cirrus7, has a heart for tinkerer and has send me a NUC8 heatspreader which Intel does not offer as a spare part. Now there is a way back and I can continue with my project in a very relaxed way

Thank you Cirrus7/Mr. Engemann!!

 

[omnium]

#work with the original heatspreader

The heatpipes are very soft (apparently soldered) and can be 'rolled up'/removed carefully with a long nose pliers. There is no danger of bending the base copper plate. Before we go any further, you have to remove the solder residues with emery paper.

I will do the first experiments with an aluminium block, which will probably be replaced by a copper block later on. The thermal conductivity of copper is twice as good as of aluminium...I'm curious about the difference.

The alu-block is glued with Silverbead SG100X

The first experiments with the Arctic Alpine AM4 are very promising.

The Macho 90 should perform better.

 

[omnium]

#Stresstest with Alpine AM4 and Macho 90 (on aluminium block)

Stresstest with Arctic Alpine AM4

Stresstest with Thermalright Macho 90

As you can see, the difference between the Macho 90 and the Alpine AM4 became very small and I wonder about it. What do you think?
.
Bad news:the current experimental solution is not suitable...at a handbrake encoding job (with the Macho 90) the cpu temperature rises slowly and reaches nearly 100°C at the end..shit! Will a copper block instead of the aluminum block change anything?

 

[thorosoe]

A copper block might help. Aluminium has a thermal conductivity of around 210 W/(m*K) at the higher end of the temperature spectrum you are using it in. Copper has 380 W/(m*K). So that makes an increase in thermal conductivity of 81%. But thickness also plays a big role. Thermal conductivity is directly antiproportional to thickness. So half the thickness would double the thermal conductivity.

So:

Thermal conductivity of the material: directly proportional (2x higher conductivity = 2x more heat energy transported per time)
Thickness of the conductor: directly antiproportinal (2x thicker material = half the heat energy transported per time).

The only advantage a big block has, is thermal inertia. It just takes longer to heat it all up. But at a certain point that advantage is moot and being able to get that heat into the adjacent cooler is way more important.

So my advice: Go for a thinner copper block, ideally one that is wider at the top so you can use the full area of the cooler contact plate, as the amount of heat transported is also directly proportinal to the contact area.

 

[omnium]

thank you for your feetback..

So my advice: Go for a thinner copper block, ideally one that is wider at the top so you can use the full area of the cooler contact plate, as the amount of heat transported is also directly proportinal to the contact area.

ok..

Image from Logic Supply​

..I think about a modified copper block.

 

[thorosoe]

The thin part in the middle would limit the heat transport quite a bit. If you could avoid that, you should. Such a neck, that is only half as wide, would decrease the thermal transport by 50%!!! The one depicted up there would be even worse.

Think of it as a pipe that drains a lake (factually unlimited supply with constant pressure). Smaller pipe = less flow. This is exactly the same as your case with direct proportionality the area the medium passes through (at any given temperature of the CPU).

 

[thorosoe]

In the meantime I got my own NUC8i7BEH and think it runs atypically warm. (usually 52+ degrees on Windows idle, 57 in UEFI). I ordered some first grade cooling paste and thermal pads and will try to get it down a bit. Might also be intersting on your side to make the first part of your passive system - the stock cooling plate - recieve the hear more effectively.

 

[4400]

the project continues on twitter.

 

[Tzk]

Sad to see that omnium left HWLuxx. Just read his (?) last posts. I guess i'll follow on twitter now

 

[CSMR]

I've been trying to do this with a Noctua NH-L9i. I did something similar with a BRIX Pro before but this is a lot harder.

Steps so far:
- Remove the fan and heatsink. Easy, but I managed to displace a resistor so may have broken the motherboard (pic attached).

So have gone backwards.

TODO:
- Find a shop that can solder the resistor back. I hope the resistor/the motherboard can be salvaged.
- Get some emery paper as advised above to remove clean the area for the heatsink where the heatpipes were attached.
- Apply thermal adhesive to attach a 46mm*22mm*1mm copper strip to the area for the heatsink, and more adhesive to glue the Noctua to the copper strip.
- Splice the 4 noctua fan cables on to the original much small fan attachment.
- Glue the feet of the case onto the case to provide support for the motherboard. (They are usually screwed in to attachments that come from the top of the case, but those attachments won't survive.
- Find someone who can cut the top off the case. Hard as it's metal. Probably needs a dremel and a clamp.