Antimemetic: Fully Passive GTX 1080, SFF, Scratch-built

[Phryq]

I could probably improve the cooling capacity by adding a third contact plate to the heatsink, which would likely need extended heat pipes. Really, though, the gain would be negligible compared to spreading the heat to the rest of the aluminum. I mean look, I'm only making contact with like 1/8 of the heat sink. Increasing cooling capacity is not a heat sink size problem; it's a heat spreading problem.

Why not use a thin vapor chamber?

Spread the heat all over the fins. I think you'd get waaay more cooling.

 

[Kmpkt]

Any chance you've looked at Pyrolytic Graphite sheets as a conductor to your enclosure wall? Apparently conducts up to 5x as well as copper. Looks pretty cool and I'd love to see someone try it:

https://www.digikey.ca/products/en?v=10&keywords=pgs&WT.z_slp_buy=panasonic_pgs

 

[|||]

I'd be curious about pyrolytic graphite, as well... it conducts really well in-plane, but is horrible in the z-axis.

 

[ricochet]

Any chance you've looked at Pyrolytic Graphite sheets as a conductor to your enclosure wall? Apparently conducts up to 5x as well as copper. Looks pretty cool and I'd love to see someone try it:

https://www.digikey.ca/products/en?v=10&keywords=pgs&WT.z_slp_buy=panasonic_pgs

https://en.wikipedia.org/wiki/Pyrolytic_carbon#/media/File:Diamagnetic_graphite_levitation.jpg

 

[Phryq]

I guess graphite would be lighter / easier to cut/shape than a vaper chamber? But I guess the vapor chamber is more conductive?

I wonder why it doesn't conduct vertically.

 

[jeshikat]

Graphite sheets are very good at transferring heat within the sheet, but they perform poorly when transferring heat in the Z-axis, as in from the sheet to another material attached to it.

They have potential, but currently they aren't of much use for heatsinking.

 

[Phryq]

Graphite sheets are very good at transferring heat within the sheet, but they perform poorly when transferring heat in the Z-axis, as in from the sheet to another material attached to it.

They have potential, but currently they aren't of much use for heatsinking.

Ok, I see. So they can transfer vertically, it's just that the heat is held within the graphite. So I guess also graphite to air thermal transfer is low?

So which material transfers best between materials? Anodized aluminium? I guess this would be the most important quality of the 'fin' material.

 

[jeshikat]

Silver has a very high coefficent of heat transfer

But copper is the best cost-effective material and aluminum a good second if copper is too expensive.

 

[|||]

I wonder why it doesn't conduct vertically.

Pyrolytic graphite is like a 2D diamond. Graphite is a single atom sheet of carbon atoms and, at a macro level, it is many of the sheets layered on one another. The bonds between the carbon atoms transfer heat extremely well (diamond is one of the best heat conductors), but getting the energy between the layers is difficult.

 

[Phryq]

Here's a list of emissivity coefficients

https://analytical-testing.ie/blog/emissivity-coefficients/

It looks like paint scores very high. Maybe this is why some people paint their heatsinks. But then maybe heat transfer from the aluminium into the pain is low?

Anodizing brings the aluminium to 0.77, much higher but not as high as some materials.

I guess if we want to be very specific, we would want emissivity between 50C and 80C?

If the heat-transfer from the CPU to the heatsink/case is very efficient, then I'm guessing the case could reach these temps?

What about covering the fins in a thin layer of diamond thermal paste?

Black Silicone Paint 0.93
Carbon, not oxidized 0.81
Carbon pressed filled surface 0.98
Glass smooth 0.92 – 0.94
Lampblack paint 0.96
Oil paints, all colors 0.92 – 0.96
Plastics 0.90 – 0.97
Polypropylene 0.97
Rubber, natural hard 0.91 (also add some protection? Rubber computer!)

I guess we need something with both a high emissivity coefficient and thermally conductive between materials. I'm thinking diamond thermal paste could do that, but don't know what else. Buy a huge bucket of diamond thermal paste and cost the entire case?

 

[Lando_Komando]

Any updates on this thing? I really like the redesign.


From an engineering standpoint i REALLY like the practicality of using off the shelf components for your heat sinks and the heat pipe configuration. No need to reinvent the wheel and raise the cost to an astronomical level if you just need something that works, and it seems that what you have "works"


BUT.......

I noticed that on Overclock.net forums you made the following statement..

The entire panel heats up nicely, but the difference between the hottest and the coldest part of the panel is significant. That means I'm probably not spreading the heat as well as I could be, and not using the panel to its fullest.​

I do not know your background, but if you are doing some real calculations to build the optimal heat sink then you should read about Spreading Resistance and how to calculate the Nusselt number for different fin geometries.

The study of electronics cooling is very new so there are not really any text books out there on it, but there is an excellent technical publication. Electronics Cooling. They have excellent articles on how to perform most calculations needed to design the optimal heat sink.
https://www.electronics-cooling.com...-for-arrays-of-vertical-parallel-flat-plates/

Here is an article from them on optimizing your vertical fins for maximum cooling. (From 2002)
https://www.electronics-cooling.com/1998/01/calculating-spreading-resistance-in-heat-sinks/
And another for estimating the spreading resistance. (From 1998)​

I have a pretty significant amount of material that I collected while enrolled in an Electronics Thermal Management Class. I took the class in Spring 2016 so it is all very current. If you or anyone else is interested in the material just let me know and Ill be glad to share.

As for spreading your heat I would suggest using a thermosyphon type system in conjunction with your heatsinks.

Some companies that might have some components of interest.

1. Calyos - If you watch Linus Tech Tips you know about these guys.

2. Quantacool - Less known, but they are in the United States. I think they are partnered up with researchers at Villa-nova (dont quote me on that.)

3. Microcooling - These guys make the evaporators/water blocks that Quantacool uses. Part of Wolverine Industries and located in Alabama (let's hope they get their engineers from Auburn or somewhere out of state..lol)​

 

[NYCesquire]

It's working just fine, actually Still plenty of improvement to make though. I'm not really getting full use of my gtx 1080 since it throttles quite a bit at full load. I have been thinking through some new redesigns...

Why not use a thin vapor chamber? Spread the heat all over the fins. I think you'd get waaay more cooling.

Can you find an appropriate one to purchase? I could not.

Any chance you've looked at Pyrolytic Graphite sheets as a conductor to your enclosure wall? Apparently conducts up to 5x as well as copper. Looks pretty cool and I'd love to see someone try it:

https://www.digikey.ca/products/en?v=10&keywords=pgs&WT.z_slp_buy=panasonic_pgs

I have not! That stuff looks very interesting...

I'd be curious about pyrolytic graphite, as well... it conducts really well in-plane, but is horrible in the z-axis.

So which would be better:

A) A 4"x3" block of copper sandwiched between the GPU and the heat sink:

--GPU:
----Copper block:
-------------Aluminum Heat sink:

B) The PG sheet sandwiched between the copper and the GPU like so:

--GPU:
----Copper block:
--------PG Sheet:
------------Aluminum Heat sink:

 

[NYCesquire]

Graphite sheets are very good at transferring heat within the sheet, but they perform poorly when transferring heat in the Z-axis, as in from the sheet to another material attached to it.

They have potential, but currently they aren't of much use for heatsinking.

See above. How much heat can they realistically spread? 180 watts worth? If the sheet is really thin, then do we really have to worry much about the z axis?

 

[|||]

The issue isn't how much heat it can spread; it's more of how much heat can you get into and out of the material. The thermal conductivity is over twice that of aluminum through the material, but less than 2% (yes, two percent) than that of aluminum into and out of the material.

 

[NYCesquire]

The issue isn't how much heat it can spread; it's more of how much heat can you get into and out of the material. The thermal conductivity is over twice that of aluminum through the material, but less than 2% (yes, two percent) than that of aluminum into and out of the material.

Bonkers. So what is it used for commercially? I'm looking at some of the application demonstrations and it seems to be just this.

 

[|||]

From https://www.electronics-cooling.com/2002/08/pyrolytic-graphite-thermal-performance-by-structure/

Application areas for pyrolytic graphite are manifold: The extremely smooth surface together with low thermal expansion coefficient makes it a good material for scientific mirrors. It consists only of carbon atoms, which makes it a good substrate for elemental analysis measurements. The lattice type lends itself to application for filters in x-ray and neutron diffractometry. For its thermal conductivity properties, pyrolytic graphite was formerly exploited only in aerospace applications, but lately it is also being considered for use for electronics cooling applications. However, compared to natural graphite, the more perfect lattice comes at a higher price.

I would imagine that to properly use the material for thermal dissipation, you need to orient the material up on its side. To think about it, realize that graphite is basically composed of a lot of graphene sheets stacked on one another. Heat transfers very well through a sheet, but horribly from layer to layer and other things touching the side of the layer. The get the high thermal transfer, you have to touch the edge of the material and it transfers to a heat sink on another edge of the material. So, you end up working with some pretty awkward constraints with the material.

 

[NYCesquire]

Fascinating. So would this sort of application not work?

Here, the heat source is on the same side of the PGS as the heat sink.

 

[|||]

Nope, you'd basically need a bunch of those sheet and stacked them into a block to transfer heat over a large surface area to benefit through all of the sheets on edge.

 

[aquelito]

Embedd an Amec Flat Coolpipe into your aluminium heatsink to improve heat transfer ?

 

[Phryq]

^^ Exactly what I've been thinking of doing for the past year.