Overclock.net banner

Building a custom passive PC case, need feedback and ideas

4K views 4 replies 5 participants last post by  Serifus 
#1 ·
This will be a pretty long post, but I'd appreciate your patience in helping me build my first passively cooled PC case. I am aware that there's a new Calyos case that can passively cool even a 1080ti, and that it will outperform even the best DIY PC cases. Still, I'd like to build my own passive PC case so it stands out in the crowd while being very efficient.

Up until two weeks ago i've been thinking of buying a HD-Plex H5 case, but after reading up on heatpipes and looking at the H5 design carefully, I've realised that case is far, far from being the optimal passive cooling solution you can build. So there came an idea of building my own case.

My main goal is low temperatures and zero fans. I want the case to fix an ATX board, along with an SSD rack for at least 3-4 SSDs and an ATX power supply. The ATX board will be placed horizontally, while the GPU will also be horizontal with the use of a riser cable. It must passively cool a CPU and GPU of at least 95W each, with a potential of some overclocking as well. Maybe with potential of adding another GPU in the future. The size of the case I'll build does not matter, it can be as big as it needs to be, as long as it provides optimal cooling solution.

I've been reading up on heat pipes, vapor chambers and cool pipes for the past few days. I've browsed a few websites that offer large heatsinks, trying to figure out what will do the job the best for me. However I'm not afraid to admit that my thermodynamics knowledge is more than lacking. I have a few ideas of how my PC case should be made like, but most of those ideas are based on guesswork more than anything else.

I'll try to present the ideas I've thought of. I'd appreciate constructive criticism on my ideas and advice on how to make the case even more efficient at passive cooling. I'll be making a lot of comparisons to H5 case and how, to the best of my knowledge, it could be potentially improved upon.

I) How the HD-Plex H5 case works.

The H5 is using eight copper heat pipes for the CPU, as well as optional 8 for the GPU. Each heat pipe is 6mm wide and about 300mm long. The CPU makes contact with a ~600 gram, ~40mm high copper block. The bottom part is ~30mm high and square-shaped the same size as the IHS. The upper part is about ~10mm tall and has bigger surface to accomodate the 8 heat pipes (about 60mm x 60mm). It has indentations for the 8 heat pipes. After inserting the heat pipes, they protrude at about half of their height. To keep them in place, an aluminum heatsink is placed that also has similar indentations for the heat pipes. The heatsink is screwed together with the copper base. Thermal paste is used at each contact of metals.
The heat pipes are routed horizontally to the aluminum heatsink panel of the case, then bend 90 degrees to the right or left and make contact horizontally with the case via half-tube indentations similar to those on the copper base. Each of the two side panel heatsinks weight about 3kg and has eighteen 26mm long vertical fins. Each fin is about 5.5mm thick at its base and narrows along the way, reaching about 4.5mm at the end. Fins are spread about 12mm apart. Each fin runs vertically from the bottom to the top of the heatsink. The heatsink base is 9mm, and it has 3mm cuts for the heat pipes. So there's at least 6mm of aluminum through which the heat has to travel from the heat pipes to the fins.

II) Improvement ideas over the H5 design.

a) Using longer and wider heat pipes. According to this article, longer and wider heat pipes can transfer more heat. The biggest heat pipes I've found are 12mm wide and 600mm long. They are very expensive compared to what H5 uses, but it's a price I'm willing to pay if the performance will improve.

b) Running the heat pipes vertically instead of horizontally. According to the above article, it can greatly improve how fast the coolant can travel back to the area where it absorbs heat, therefore greatly improving heat transfer. It looks to be especially impressive for wider heat pipes, and has less effect on smaller ones.

c) With the above two points in mind, the heat sink panel must be taller to accomodate the longer heat pipes and the base of the heat sink must be thicker if I decide to use 12mm pipes; this brings an added benefit of more surface area on the fins.

d) More contact between heat pipes and the base of the heat sink. In H5 case, only about 50% of heat pipes' surface made contact with the heat sink base. I am unsure what's the best solution here. Should I find a workshop that can drill very precise holes of the same diameter as the heat pipe (or +0.1mm larger than the heat pipe)? Is this the best solution here? I've read of an "embedding" technique of the heat pipes, but I have no idea how that is executed and if any workshops would be willing to do that (and how difficult that process is). This would require the flattening of the heatpipes in the area of contact with the heatsink (which considering 600mm long pipes, would require to flatten about 75% of their length). Flattening the heat pipes reduces their cooling capabilities, so I'm unsure if I want to do that. I think the solution of drilling precise holes for the heat pipes to fit tightly would be the best?

e) Different thickness of the aluminum base of the heat sink. The H5 case's base is 9mm thick, but only 3mm are used for housing the heat pipes. So 6mm of aluminum base separates the heat pipe from the fins. To my general understanding, a thinner base would allow the heat to travel faster to the fins, which are responsible for dissipating the heat into the air. Am I wrong in thinking that? My idea is there to be only about 1mm of base thickness between the fins area and the heatpipe. This way the heat would get to the fins faster. Please correct me if my way of thinking is incorrect.

f) Make the base located between the CPU and heatpipes as thin as possible. Copper has several dozens times worse thermal conductivity than heat pipes. Therefore, the H5 design is not optimal, as the copper base is about 40mm tall. That's a lot of distance before the highly conductive heat pipes receive the heat from the CPU. I suspect they made that decision because some motherboards have oversized VRM heatsinks, so the wider part of the copper base had to be elevated in order not to conflict with other objects on the motherboard (?). I'd need to consider finding a motherboard that has a lot of room around the CPU area, in order to make the copper base as low to the ground as possible.

The idea is to make the heat from the CPU reach the heatpipes over as little distance of copper as possible. I'm not sure what would be the best way of doing the contact between copper base and heat pipes. I could do it like with H5 case (half-pipe shaped holes in both copper base and the upper piece that will rest above the heat pipes), or I could drill pipe-shaped holes in a one-piece copper base. I don't think flattening the CPU ends of the heatpipes will make much sense. One, this means I can fit less pipes across the limited width of the copper base. Two, flattening the pipes has an impact on cooling capabilities. So overall, I don't think trying to embed the heat pipes into the copper base is not a good idea.

I recently had a completely different idea of handling the copper base problem; instead of putting a copper base on the CPU, I could install a vapor chamber which can have flat dimensions of 50~60mm x 50~60mm. There should be enough space around the CPU for that. A vapor chamber has many benefits over a copper base. It has about 10-12 times better thermal conductivity than copper, and is capable of evenly spreading the heat over its surface. However, the vapor chamber is flat and thin. It's not possible to drill mounting holes for the heatpipes on the vapor chamber. So it will be necessary to flatten the ends of the heat pipes to a certain degree and place as many of them on the vapor chamber as possible. The more a heat pipe is flattened, the bigger negative impact on its performance. I'd need to consider the optimal flattening. Finally, secure the heat pipes in place with a top-mounted heatsink (aluminum with fins on top? or copper with fins on top? or copper with a flat top, so I could place a bigger heatsink on top of it?).

g) More vertical fins. The H5 case has less than 20 fins. Placing more fins will increase their surface area. However, I don't know what's the ideal spacing between the fins. Is there a point after which placing the fins more densely will bring no additional benefits, or even make the heat dissipation less efficient? Or should I just go with as many fins as I can find on the market?

h) Longer fins. The fins on H5 case are less than 30mm long. I found some heatsinks that offer fins up to 100mm long. What would be the best length? Is there a point where longer fins no longer bring any benefits, or even make heat dissipation worse? Or should I just go with the longest fins I can find on the market?

i) Thickness of the fins. In H5 case, the fins are 4.5-5.5mm thick. To my general understanding, thinner fins would make more sense because the heat can reach the air over shorter distance; this also allows for more fins to be packed across the radiator's length. Should I just go for thin fins?

j) Connecting the fins' ends with the base of the heat sink. The ends of the fins are usually very cold. If the base of the heat sink is only a little bit thicker than the heat pipes running though it, then the heat from the pipes has a short distance to travel before reaching the non-fin surface of the heat sink base. This creates the idea of connecting the flat base with the end of the fins through the sides of the heat sink. This could only potentially work well while connecting through the sides. Going through the bottom or top part would be a very bad idea. It would block the hot air from going up naturally, and it would block fresh air from entering between the fins from the bottom.

The connection between the base and the ends of the fins could be done with highly-conductive vapor chambers. However, the vapor chambers aren't infinite in length; the longest I could find were about 500mm long. This would mean I have to consider the fins length as well. It could be beneficial to make the fins (just an example) 20-40mm shorter to allow additional 20-40mm of vapor chamber to make contact with more fins.

You could place a lot of these vapor chambers next to each other, making it possible to transfer heat to a large portion of the fins' ends. However, won't it have a negative impact on airflow? This will cover the ends of the fins, which would mean the fresh air can enter only from below the fins and not along the side of the heat sink.

k) Place a heatsink with fins and very thin base inside the case as well. The additional heatsink would be screwed together with the outer heatsink on each wall that dissipates heat. This would require the case to have no top cover (not a big deal for me). This wouldn't interfere with PC components - the additional fins would be above all the components. The inner fins wouldn't pose a big barrier for passively cooling the motherboard components; there's still plenty of space between the fins so the warm air can rise from the motherboard.

l) (This one is probably incredibly stupid idea, but here goes anyway)
Routing one or two of the heatpipes through the fins. Is that a good idea? This would mean one or two heat pipes will run horizontally along the base of the heat sink (similar to H5 heat sink, run along half-pipe shaped incendations), make two 90 degree turns and then finally go through the fins.

==================================================================

Same considerations could be made for the second heat sink wall for the GPU. Third wall (front panel of the case) could be used for a 2nd GPU. The back wall would house the I/O part of the motherboard at the bottom. Not sure how I want to tackle the upper part of the back wall. I think the best idea would be to make it a giant heatsink with large fins as well, then connect the base of the back wall with the base of the walls responsible for cooling CPU and main GPU. This would allow the heat from the heat pipes travel to the back wall as well.

I'm not sure how high above the ground I want my case to be. Will it make any difference if I make the case 2cm above the ground, or 10+cm? Will more clearance between the case and the ground offer more air to enter the fin area? I'm also aware that the bottom plate on which the motherboard and other components will be screwed, needs holes for air to enter to cool the motherboard components.

The whole PC will be 100% fanless, including the PSU. I plan to buy some decent copper heatsinks for the mosfets as well as for the chipset. Are there any other components on the motherboard that would require better heatsinks in a no-fan case?

This became a very long post. Just noticed almost 3 hours have passed since I started writing it.

To summarize, I would like you to read through my ideas and comment on them. Are they good? Bad? Can they be improved even more? Is there anything else I forgot to mention? Anything else that can make the PC case more efficient at cooling?
 
See less See more
#2 ·
interesting stuff:
https://smallformfactor.net/forum/threads/fanless-ryzen-polaris-project.2138/
https://linustechtips.com/main/topic/573717-antimemetic-scratch-built-fully-passive-gtx-1080-sff/

a- seem strange that longer=better, usually you want to dump heat as soon as possible; unless some strange phenomena happens like oscillations end to end
a2- don't waste money on questionable performance improvement, 8mm are still good
b- correct, capillary effect take care of the liquid return but nothing wrong with an aid from gravity force
c-
d- the H5 side panel contact is very poor, better the 2-block style;
slightly bigger holes (+0.1 +0.05 mm) to accomodate thermal paste
if you will need some more pressure just slightly grind the surface of one block
d2- "embedding" technique : I think to the "direct die contact" of the newer cooler
it's no good for your use : the heatpipe can't be moved because plastically deformed into the groove
e- thinner base = fast transfer but less heat spread = less surface
doable but you have to spread your bundle of heatpipes and contact points
f- see point e but reversed
f2- I think you have to make a choice : use the side panel as heatsink or meke a "stack effect"-style case
if you spread the heat too much you won't have enough deltaT to make the 2nd cooler work (yes, they have a working range of T)
g- too musch fins = more restriction to airflow and you don't have a fan to push it...
h- there is a breaking point where longer fins don't bring benefits, what length ? you have to run a CFD analysis...
i- see point e , g and h
j- useless if you have already evenly spread the heat
plus every contact point call for a resistence = you can't add an infinite numbers
k- buy panels with already fins on both side, not worth the hassle
l- not stupid but not woth
m- clearence from the floor will help
it's the first think that I thought seeing the H5: << horizontal?! little to nothing air vent !? really!? >>
 
#3 ·
Super interested in your project. I have not built a passive, fanless PC like you're talking about, but plan to follow this thread. As a former engineer with a focus in fluid flow and thermodynamics I can get real nerdy on this stuff real quick. Looking forward to seeing what you do here.

I will say that I agree that vertical is better for heat pipes and heat sink fins, too. That H5 case fails in these areas, presumably because they're going for a standard home theater / flat-and-squat design constraint. If not, well, I'm not sure what they're thinking. An ideal design for the heat management is tall and narrow with generous air inlets on the bottom and equally generous exhaust vents on the top to promote natural convection through the case. Modern flat panel TVs do this pretty well if you check out where the lower and upper vents are. Get your internal heat sinks and heat pipes also designed around this idea as much as possible without blocking that convective airflow and you should have good results. The last benefit of this is that the air at the floor is usually noticeably cooler than just a foot or two above the floor, and you'd be drawing that cool air into the bottom vents (go ahead and measure it with a Thermapen or similar and see for yourself). I'm sad my PC case only has front vents for the main compartment and only one floor facing vent for the PSU. I'd prefer to draw from as low as possible.

Anyway, I'll keep checking back. Sounds pretty awesome.
 
#4 ·
Interesting project. I am eyeing the Calyos as well, but it limits the number of PCIe devices that one can use. I also wonder if completely fanless is necessarily. The new Noctua 200mm fan can run as low as 350rpm. At that speed, the fan practically is silent. Having two or three at the bottom blowing cold air upward should be good for the VRMs and other stuffs generating heat.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top