Cooling Heatsink theory

[Phryq]

Sorry for making so many new threads.

Anyhow, I'm thinking about how to make the best possible cooling solution, either 100% or 90% passive.

It seems maximum surface area of heatsink is the bottleneck (heatpipes are uber-effective at moving heat away from the CPU, so the only issue is heat-sink to air, or case-to-air if your case is your heat-sink)...

Are certain heatsink designs more conductive to airflow (passive or active) than others?

What are the most effective designs, or specific heatsinks, either for passive or with a quiet fan?

What about covering your case in aluminum hair?

something like this?

a list of various designs

Should I just go for as many fins as possible, and forget about fancy patterns?

Mostly I'm thinking... what should the lid of my case look like? How flexible is aluminum? Maybe with the hairs, I can just bend them to my liking? Change my computer's hairstyle on a whim.

Maybe the ideal would be covering the case in aluminum wool... but that would be ugly, itchy, and hard to clean. Would it be effective?

Which, if any, of these designs would impede airflow?

 

[Diablo]

I think you may find this video interesting by Dmitry at Hardware Canucks:


I'd say it makes a difference whether you're looking to cool something like a HTPC, without a GPU.
Otherwise, it's going to be quite hard to keep it passively cooled keeping it a decent size.

The NSG S0 is probably the only contender I'm currently aware of is capable of doing so that's on a manufacturer level.​

As you mentioned, the problem with passive cooling isn't getting the heat away from the component, but rather getting the heat away from the heatsinks. The NSG S0 solves that problem by having huge heatsinks compensating for the lack of heat dissipated heat.

 

[Phryq]

What I'm thinking is I may have a quiet fan *outside* the case, however I'd like to make it as passive as possible.

I've seen that video, as well as the NSG S0, both are really interesting. The NSG has huge radiators... and I think it works because it has lots of surface area, rather than because it can simply absord heat forever. I watched a video about it, I think Linus, who said he could feel the heat aura around it... and so I was thinking there was so much surface area, it moved heat from the metal into the air efficiently even with little airflow. If that's the case, then I'm right that basically more surface area=better, and I should just find the design that maximizes surface area.

The important question is, how well does the NSG S0 work after the entire case has been saturated with heat (how well does it move the heat into the air).

 

[|||]

So heat sinks work on the basis of air moving in, along the hot surface having energy transferred to it (needs a delta temperature), and leaving the structure. This can be done through passive natural convection, taking advantage of the natural buoyancy of hot air rising over cooler air, or with force air flow. Forced air flow can do this through much denser fin arrays, as they usual draw air in a certain area and force it in a specific direction (through the heat sink). Natural convection needs a less dense array, as air will want to rise out of it, but there needs to be air able to enter it as well, and if it is prevented from mixing in this way, it will just be stagnant air and useless.

For the fin types, for force air, you know the direction of air, so you want to maximize the surface area in that direction of flow, which usually will be a fluted fin design. Fluted fins are also good for natural convection if the fin are oriented vertically along a vertical surface where cool air can enter from the side of the array, but the array will be less dense than the forced air array. If the heat sink is oriented along a horizontal surface or the fins might be oriented in a way that doesn't go with the natural flow, a pin fin structure would be best to mitigate the issues of orientation. For instance, while there is a most common orientation for a motherboard, with the motherboard vertical, processor on top and PCI-e add-in cards out the back, there are plenty of cases that will orient it differently.

There are ideal fin spacings and can be calculated using heat transfer formulas and are assuming ideal conditions. CAE models can design much more complicated structures for complex conditions.

 

[Phryq]

Ok, I was thinking something like this...

So if I'm putting the heat-sink on the 'roof' of my case, I'm guessing 'pins' is ideal, as I don't know which direction air will be moving... or I should entirely give up the idea of 100% passive, and figure out which way the fan will push air, then consider the heat sink-based on that.

At least I have something new to sink about.

 

[|||]

Yeah, you'll want pin fin at the top, otherwise you'll end up in a situation like this where the air coming in from the side where a solid fin is and can't penetrate the array to absorb more energy.

 

[Phryq]

I see, thanks. So on top should be fins... but if I mount the thing sideways, will I get better convection? I'd rather not because it would feel 'tippy' to me, but I'm willing to do what I must.

https://www.qats.com/cms/2010/09/08...nvection-cooling-and-how-you-can-optimize-it/

https://www.micforg.co.jp/en/naturalcone.html

http://www.wakefield-vette.com/products/natural-convection.aspx

Some good reads about convection, but that last link says painting improves heat radiation... everything I've read has said the opposite.

 

[|||]

You want pins on top. The animation is to show that air can enter between the fins in the areas where there are cut outs...you want cut outs in more than one direction to allow as much air to easily flow past the exposed surfaces to convect away as much heat as quickly as possible.

A heat sink on the side will perform better than one on the top due to natural convection operating along that direction.

There are three modes of thermal transfer: conduction (through solids), convection (through fluids, liquid or gas), and radiation (through free space). They are talking about radiation there...it helps a little, but not too terribly much. And you don't want to paint something to get the emissivity up to improve radiation, you want to dye it, like with anodizing aluminum, or put on a coating of something that is both highly thermally conductive and like a black body for high radiation. Typical paint will insulate the heat sink and hinder its performance.

 

[Phryq]

Ok... I'm guessing in practice it's best to go anodized. I see a lot of anodized aluminium but haven't heard of any dyed or coated aluminium.

Will I get much better cooling then by placing the case vertically rather than horizontally? And would this be the best design for side-fins?

Something like this design, but make the fins a bit thinner. Will wobly or differently shaped fins impeding airflow mean that they're actually less effective, even though they have more surface area?

 

[|||]

Any complexity to shaping the fins will be hard to tell. You'd need to do a CAE analysis to determine that or just do a bunch of empirical testing to find out.

Anodizing is just a process of rusting aluminum...the resulting aluminum oxide is harder than the base material and unlike iron, the oxide isn't permeable, so it won't continue to oxidize after the outer later has. Anodization just creates a bit thicker later than would naturally occur, so it is a more durable finish, and any color added is a dye.

As for orientation, the best performing would put the largest surfaces in a vertical orientation with the longest axis parallel to the horizontal plane. You want the most surface area with the shortest height.

 

[Phryq]

I'm googling some CAE analysis, It looks like heat doesn't move through the aluminum as well as I was imagining. So longer-fins will have a diminishing return. And this shows how important it is to have heat pipes contacting each part of the 'lid' or heatsink

I'm thinking liquid copper would be best for attaching the heeat-pipes to the lid, but I'd rather do something safer for the CPU heatsink?

I'll see if I can do a CAE analysis on my entire concept. Do you recomend any CAE software? I'm trying https://www.simscale.com/ but it doesn't seem like it will work for a full system build.

It would be amazing if a piece of software could take into account materials / natural convection but I'm guessing that doesn't exist.

 

[Phryq]

So I was thinking to put this cooler on the ASRock Micro STX.

Which is rated to cool 135W, so really good. I would make it basically caseless. Maybe a frame with some dust-mesh to keep it clean, and a 'travel box' to put it in my backpack.

It would look like this,

But sadly, the heat-sink wouldn't line up well with the CPU, and wouldn't clear the ram. Something like this, that had space for RAM and could mount onto a delidded CPU would be perfect (it seems GPU coolers are closer to what I want... but I'm guessing big mounting problems. I could put a vapor-chamber between this and the CPU, but that would increase the total height. On the other hand, since the heatsink is sitting a bit higher, you could put a 140mm fan under the heatsink, where the MXM card would normally go, right?


Also, this comes with a heatsink to go under the GPU motherboard. I wonder if that would be useful for this situation, or if it's just a waste of space.

 

[Phryq]

So here's an idea based on the design above,

It's the ASRock MicroSTX, a large heatsink (I forgot to draw heatpipes going through the heatsink fins) attached to the CPU by a vapor chamber (I'm thinking a delidded CPU). Under the heatsink fins is a 140mm fan, in the area an MXM card would usually go (but I don't need one).

I could put sides / faceplates on this, or maybe I'll get better airflow by just making a frame, and wrapping it in a dust-mesh. Then put it in a 'travel case' / box for travelling.

I think the above design would let me overclock quietly, no? I don't plan to overclock, but want to run the fan on lowest voltage possible (and maybe occasionally overclock if needed )

 

[owliwar]

So here's an idea based on the design above,

It's the ASRock MicroSTX, a large heatsink (I forgot to draw heatpipes going through the heatsink fins) attached to the CPU by a vapor chamber (I'm thinking a delidded CPU). Under the heatsink fins is a 140mm fan, in the area an MXM card would usually go (but I don't need one).

I could put sides / faceplates on this, or maybe I'll get better airflow by just making a frame, and wrapping it in a dust-mesh. Then put it in a 'travel case' / box for travelling.

I think the above design would let me overclock quietly, no? I don't plan to overclock, but want to run the fan on lowest voltage possible (and maybe occasionally overclock if needed )

I guess that would be doable.. this is nice, but then, will it be really that low profile? some low profile coolers may take only shy 55mm of height or less... but i like the design very much

 

[Phryq]

I guess that would be doable.. this is nice, but then, will it be really that low profile? some low profile coolers may take only shy 55mm of height or less... but i like the design very much

The Accelero S3 for example is 42mm, and it would have to sit above Ram (if you could cut out a spot for ram, you could make it even thinner, but I'm thinking it's not possible with my abilities). So maybe a total of 8cm tall... which is taller than I'd like. If you included the bottom heatsink, it would be even thicker.

So I'll need to find a 20mm-ish thick radiator that's around the same dimensions as the MicroSTX and has the heatsink in around the right place. Also, mounting the fan underneath might be tricky (though come to think of it, there might still be room for an MXM, if a gamer wanted to try this design).

It would be ultra cool if the frame could be made out of heatpipes, but I think it's not practical. They'd be too malleable.

Here's a slightly thinner radiator, but it would definately have to go over the ram, and looks painted.

It could sit on top of a vapor chamber, but how would it stay secure? I guess the frame would have to physically support it, or is there another way?

How possible would it be to make something custom like my drawings above? Bending heatpipes and individual fins?

 

[Phryq]

I'm not sure whether to simply use a GPU heatsink, or make my own... possibly by 3d printing fins onto the heatpipes, and supporting them with the frame. Like in the picture below (also cut holes in the faceplate). I would cover everything in a dust mesh. It would be best for the fan to blow up, blowing hot air away from the fins, yes? Also added some fins going down on the sides. I think this could handle any amount of CPU overclocking, no? And very tiny.

Here's someone talking about bonding the heatsink directly to the die.

And here's an interesting fin design

^^ More metal in the hot parts, and less metal further away.

 

[Phryq]

By the way, if the fins were made custom, they could go around the ram slots, and then it could be *really* low profile.

I've also thought, it would be amazing if instead of aluminium fins, they were all heat-pipes. Like thin heatpipes, or like NoFan's "IcePipes".

I've thought of just sticking one of these one too.

But I'd probably have to take it off every time I travelled. Then I'd worry about having to turn my computer on for the airport security, and it could overheat without a heatsink.

But imagine using that same principle ^^ with my design. It would be similar surface area, but super compact.

 

[Ceros_X]

What about using a Thin Itx board where the ram lays down flat on the board?

 

[Phryq]

Yes, thin ITX would be best for this. You could use 5cm thick cooler, and have a system 6cm thick... the problem is I can't find a Thin ITX with either 2 m.2 slots or supporting 64gb ram. I need one of those options for my throughput. But maybe something will come out soon...

Also, here's an interesting site that let's you do heatsink calculations. I'll look at the cost for them to make something like my design above. Seems they only do bulk orders... so I'll have to convince 200 others to adopt my design.

They also do custom heatpipes, including thin heatpipes. So you could make it entirely out of pipes (the fins are heatpipes). I'm sure that would give ultra cooling, even without a fan.

And another idea,

Make the entire lid a single vapor-chamber,

http://www.inheco.com/de/heating-cooling-technology/vapor-chambers.html

If this can move heat as well as a heat-pipe, then it's the ultimate. Simply put some pins above it. Make it sit just above the ram, and have another vapor chamber / copper going straight up from the CPU.

 

[Phryq]

It seems relidding is really useful for Kabylake...

So I'm thinking... Relid but replace the lid with a thin VAPOR-CHAMBER!!! Mwa ha ha!

You can then get a wider lid, meaning more surface area to mount heatpipes onto.

The ideal would be if the case were a single 3D vapor chamber, moving water straight from the die into the case fins. I know that will never happen, so I'm trying to figure out how close it's possible to come to that.

But think, if that were possible it would be like having a giant NoFan mounted direct die on an open air system... It would cool anything.