News Modder builds a totally fanless 'retro-futuristic' mini PC — attached to a USB4-powered bank of SSDs

[Admin]

TheJiral on Reddit showcases a custom fanless PC.

Modder builds a totally fanless 'retro-futuristic' mini PC — attached to a USB4-powered bank of SSDs : Read more

 

[bit_user]

TheJiral even toggled an overclock option on the ASRock motherboard that enabled a 45 Watt TDP "performance" mode, though this did result in the setup eventually overheating when stress testing CPU cores

Note that the modder used the two fat heat pipes from the original solution to spread the heat into the base of the heatsink. They're being used sort of like a vapor chamber, which I was going to suggest.

I think thermal performance could be improved if they filed and sanded the heat pipes flat and used a good heatsink compound, rather than using a thermal pad between them and the heatsink.

 

[8086]

Aesthetically, his design looks a lot like the 1980s CRAY X1.

Source: https://cray-history.net/cray-history-front/fom-home/fom-xmp/

 

[bit_user]

Aesthetically, his design looks a lot like the 1980s CRAY X1.

Source: https://cray-history.net/cray-history-front/fom-home/fom-xmp/

Yeah, but I wonder if that's just happenstance?

If you follow the purchase link for the heatsink used, the bottom of the page shows lots of other heatsinks of a similar shape & style. I believe these are designed for LED lighting applications. The modder either knew this, or just searched for a large heatsink and happened to find/use one of these. Perhaps the Reddit thread covers this topic. I didn't read the whole thing.

BTW, I wonder how much the cooling would be improved if the system/board were oriented vertically and a heatsink were used that could better take advantage of such convection. There are some server heatsinks you can find that have a vapor chamber directly integrated into the base.

Another thought I have is how hot the RAM & SSD are getting.

 

[TheJiral]

The design is certainly "form follows function". I started with looking for easiliy available heat sink options. For compact heatsinks with higher convection based heat dissipation the options are more limited than some might think. LED heat sinks are the best suited option IMHO. They are not only designed to work in horizontal orientation and for high power dissipation and feature compact horizontal dimensions, but they also come with good surface finish as they are intended to be used where they can be seen and where looks often matter. I designed the case to integrate both, the board and the heat sink.

The looks that remind of the Cray or maybe works from Dieter Rams were a happy accident but part of the reason I stuck with this design and also with this colour scheme.

 

[TheJiral]

Note that the modder used the two fat heat pipes from the original solution to spread the heat into the base of the heatsink. They're being used sort of like a vapor chamber, which I was going to suggest.

I think thermal performance could be improved if they filed and sanded the heat pipes flat and used a good heatsink compound, rather than using a thermal pad between them and the heatsink.

Indeed. Using a good heat sink compound on a sanded flat heat pipe would probably be more efficient in the heat transfer but also a lot less forgiving regarding tolerances. I really did not want to sand the heat pipe as I have no experience with how much one can get away with, without damaging the integrity of the heat pipes internals. More importantly though, I expect the rate limiting step not at that interface of heat pipe to heat sink but rather at the heat transfer from heat sink to air.

My experience with it so far seems to confirm that, as under long lasting stress tests, the heat sink is probably reaching an equilibrium between 60 and 70°C. That is pretty much where I want the max Temp to be.

 

[TheJiral]

One small correction to the article. The original ASRock Industrial Box 4x4 7640U is not a fanless MiniPC but a regular fan based mini PC. I do have a spare radial fan now indeed

Because the question was asked in the article. The white case was printed with my custom built Voron 2.4 in Formfutura ABSpro which is a flame redartant ABS/PC blend. Rather challenging to print actually but it has good heat resistancy and is self extinguishing. I designed it in the open source program FreeCAD.

 

[bit_user]

Thanks for stopping by!
: )

Indeed. Using a good heat sink compound on a sanded flat heat pipe would probably be more efficient in the heat transfer but also a lot less forgiving regarding tolerances. I really did not want to sand the heat pipe as I have no experience with how much one can get away with, without damaging the integrity of the heat pipes internals.

I'd need to see how much material is being removed, but I'd probably risk it with also some judicious use of pliers and a file to work down the high spots.

More importantly though, I expect the rate limiting step not at that interface of heat pipe to heat sink but rather at the heat transfer from heat sink to air.

It can be tough to go on intuition. What if your 0.5 mm thermal pad is only making good contact at a few spots?

Anyway, regarding thermal pads, which did you use? I've had a good experience with ARCTIC TP-3. It comes in 0.5, 1.0, and 1.5 mm, in case you find the surface variations are bigger than you expected.

https://www.amazon.com/ARCTIC-TP-3-Performance-Conduction-Installation/dp/B09V53W559
​

What heatsink compound did you use between the heat pipes and the die? The paste I've used for a similar direct-die application is ARCTIC MX-6.

 

[Notton]

That is an interesting heatsink choice.
Since the APU is <30W, I would have used a regular C-shaped heatsink, like a thermalright SI-100, without the fan and called it a day. The exposed base plate makes it easy to fashion a DIY bracket for the mini-PC's mounting holes.

But I like the look from the vertical fin heatsink better.
It reminds me of Alpha and Swiftech pin fin heatsinks of the Pentium 3/4 era, before heatpipes were invented.

 

[TheJiral]

That is an interesting heatsink choice.
Since the APU is <30W, I would have used a regular C-shaped heatsink, like a thermalright SI-100, without the fan and called it a day. The exposed base plate makes it easy to fashion a DIY bracket for the mini-PC's mounting holes.

But I like the look from the vertical fin heatsink better.
It reminds me of Alpha and Swiftech pin fin heatsinks of the Pentium 3/4 era, before heatpipes were invented.

I did not check how the base plate aligns with the location of the mobile CPU, but from a first look it would be probably off centre along one axis. That is not a problem technically of course, merely a design question. It would be interesting though how it would compare in terms of performance. Convection based cooling is helped by larger plate spacing which goes against what is needed for heat sinks intended exclusively for use with fans. Still, C-shaped heat sinks allow for vertical pass through, which would be a plus.

Creating some metal bracket is beyond my capabilities and I wouldn't want to use a plastic bracket for that job even if polycarbonate for example could probably handle the temperatures.

In the end I would have chosen the LED heat sink anyway I suppose, I just think it makes the whole thing look more interesting and unique.

 

[TheJiral]

It can be tough to go on intuition. What if your 0.5 mm thermal pad is only making good contact at a few spots?

Anyway, regarding thermal pads, which did you use? I've had a good experience with ARCTIC TP-3. It comes in 0.5, 1.0, and 1.5 mm, in case you find the surface variations are bigger than you expected.

​

https://www.amazon.com/ARCTIC-TP-3-Performance-Conduction-Installation/dp/B09V53W559​

​

What heatsink compound did you use between the heat pipes and the die? The paste I've used for a similar direct-die application is ARCTIC MX-6.

I chose "Iceberg Thermal DRIFTIce" which claims pretty good heat transfer (no idea if it is true) and is just the best value, by far, where I live. I think it was 3,33 EUR per piece. I did reseat the heat sink a few times, without using new pads. The pad did appear to be nicely compressed along the whole length. Initially I planned to replace it with fresh ones but as the thermal performance of the setup looked just fine as it was I kept it as it was. I can still change it when need arises.

The heatsink compound I used above the die was Noctua NT-H1, simply because I still had some lying around and I would expect it to be decent enough for the job.

www.iceberg-thermal.com/product/thermal-pad/driftice/
www.noctua.at/en/nt-h1-3-5g

 

[bit_user]

It reminds me of Alpha and Swiftech pin fin heatsinks of the Pentium 3/4 era, before heatpipes were invented.

Uh, heat pipes go back a lot longer than that! They just weren't used in PC cooling until sometime in the mid-2000's, though I haven't pinpointed exactly when.

https://en.wikipedia.org/wiki/Heat_pipe#History
​

 

[Notton]

Creating some metal bracket is beyond my capabilities and I wouldn't want to use a plastic bracket for that job even if polycarbonate for example could probably handle the temperatures.

It's a lot simpler than you might imagine. All you need is a slotted woodworking mending plate, washers in various sizes, and the appropriate screws. If the plate is sized perfectly, it doesn't need to be cut down to size.
The heatsink stays in place, even without the alignment dimple.

I did not check how the base plate aligns with the location of the mobile CPU, but from a first look it would be probably off centre along one axis.

I ran into the off-center problem as well. Some older C-shaped heatsinks have an off-center axis, like a Raijintek Pallas. However, I don't know if they are discontinued or easily available.