Gold leaf as thermal compund

[DJRWolf]

So I had the thought about weather ultra-thin gold leaf could be used as a thermal compound. Since 24k gold is malleable so as long as the surface of the IHS and cooler are mostly smooth it should provide good heat transfer since it should mold it's self to small imperfections. Cost wise it is not bad since that same malleable trait also lets you get ultra thin sheets so not much gold is needed. A quick check of Amazon and I found a 30-pack of 1.2" square gold leaf sheets for $12.95. Even if you build 2 computers per year that is still 15 years worth of gold leaf sheets for you. You can use normal thermal paste when doing test fittings and only use the gold leaf once you confirm everything is good to go for final mounting. Here are some quick numbers for heat transfer. The numbers for normal and liquid metal I found on EKWB's site. One major advantage I can see for gold leaf over liquid metal is for laptop's and tablet/smartphone since it is solid until just over 1000 °C and will not leak out like liquid metal would.

 

• Normal thermal compound: 8.5 W/mK
• Liquid Metal: 20 to 40 W/mK (depending on compound)
• Aluminium: 237 W/mK
• Gold: 318 W/mK
• Copper: 401 W/mK

 

Would be nice to see LTT do a review since it is just for thermals and they can use older hardware (4th to 7th gen Intel or Bulldozer AMD) as long as the only thing that changes is the thermal compound to see how much of a difference it makes.

[Corrupt_Liberty]

It has been done.  It's not great.  Stick with a known good TIM and be happy.

[Skiiwee29]

10 minutes ago, Corrupt_Liberty said:

It has been done.  It's not great.  Stick with a known good TIM and be happy.

if its been done, best to list some sources to back up the claims. 

[BigBlackRat]

https://www.reddit.com/r/hardware/comments/939glu/gold_leaf_as_tim_tested/

[DJRWolf]

4 hours ago, Genwyn said:

Gold is soft but not soft enough, and gold leaf is way too thin to fill in the gaps. Consider the actual amount of material used in conventional thermal paste to cover an IHS vs how much actual gold is in a gold leaf sheet the same size. A blob of gold the size of the pea sized dot of thermal compound is about a gram and that’s substantially more expensive.

Thermal compound is only used to fill in air gaps between the IHS and the heat sink. The blog post I linked to for EKWB talks in detail about it and how you can even go without it if you spend the time to get both the heat sink/water block and the CPU lapped to a perfectly smooth finish as that is the best since as the bullet points in my original post show thermal paste is not a good thermal conductor but a perfectly smoothed surface will conduct it better. It is just that takes a lot of time and training to get it right. Linus even talked about it in a recent video (link skips to that part).

3 hours ago, BigBlackRat said:

https://www.reddit.com/r/hardware/comments/939glu/gold_leaf_as_tim_tested/

Most of those tests the gold leaf was not a smooth application but crumpled up thereby adding air pockets so I do not consider that a test series that would display what gold could do if it is applied correctly. But from what that post was saying it looks like applying it correctly will take a lot of skill or just finding out what method works well. May even be only useful for CPU's with larger IHS's like a Threadripper so you don't have to cut it.

[ryao]

9 hours ago, DJRWolf said:

So I had the thought about weather ultra-thin gold leaf could be used as a thermal compound. Since 24k gold is malleable so as long as the surface of the IHS and cooler are mostly smooth it should provide good heat transfer since it should mold it's self to small imperfections. Cost wise it is not bad since that same malleable trait also lets you get ultra thin sheets so not much gold is needed. A quick check of Amazon and I found a 30-pack of 1.2" square gold leaf sheets for $12.95. Even if you build 2 computers per year that is still 15 years worth of gold leaf sheets for you. You can use normal thermal paste when doing test fittings and only use the gold leaf once you confirm everything is good to go for final mounting. Here are some quick numbers for heat transfer. The numbers for normal and liquid metal I found on EKWB's site. One major advantage I can see for gold leaf over liquid metal is for laptop's and tablet/smartphone since it is solid until just over 1000 °C and will not leak out like liquid metal would.

 

• Normal thermal compound: 8.5 W/mK
• Liquid Metal: 20 to 40 W/mK (depending on compound)
• Aluminium: 237 W/mK
• Gold: 318 W/mK
• Copper: 401 W/mK

 

Would be nice to see LTT do a review since it is just for thermals and they can use older hardware (4th to 7th gen Intel or Bulldozer AMD) as long as the only thing that changes is the thermal compound to see how much of a difference it makes.

If you did a really good job of getting the surfaces to be smooth, you would not need thermal compound. Here is a link to a discussion of how others get extremely smooth surfaces:

 

https://www.finishing.com/36/23.shtml

 

If you pull it off, you would get better results than you would from any TIM. The thermal conductivity of zero space between two surfaces should be infinite. There is nothing better than that.

[ryao]

Another idea would be to use solder as your thermal compound. Intel and AMD use solder as the TIM to attach the IHS to dies in certain models, so using it as your TIM is likely an option. You might not be able to remove your heat sink from your CPU afterward though.

 

This suggests that a mix of gold and tin are used to solder dies:

 

https://www.electronics-cooling.com/2006/08/thermal-conductivity-of-solders/

[DJRWolf]

20 hours ago, ryao said:

If you did a really good job of getting the surfaces to be smooth, you would not need thermal compound. Here is a link to a discussion of how others get extremely smooth surfaces:

 

https://www.finishing.com/36/23.shtml

 

If you pull it off, you would get better results than you would from any TIM. The thermal conductivity of zero space between two surfaces should be infinite. There is nothing better than that.

With gold leaf you only need to be mostly smooth but with lapping them to the point you don't need TIM that takes a lot of work. The EKWB blog I posted in my original post cover the topic. Also a link here so you don't have to scroll for it.

 

It comes down how much work you want to do. What takes more work: Using gold leaf as a TIM or lapping the CPU and cooler.

20 hours ago, ryao said:

Another idea would be to use solder as your thermal compound. Intel and AMD use solder as the TIM to attach the IHS to dies in certain models, so using it as your TIM is likely an option. You might not be able to remove your heat sink from your CPU afterward though.

 

This suggests that a mix of gold and tin are used to solder dies:

 

https://www.electronics-cooling.com/2006/08/thermal-conductivity-of-solders/

I don't see how that would work for a desktop. You would need to have melted solder to stay liquid long enough to mount the CPU to the cooler and get it installed to the motherboard. The problem is how do you make sure the CPU and cooler are all lined up so you can mount the weight of the cooler to the motherboard if you are using an air cooler. Also you have to make sure you get the right solder as some will melt at 90°C.

 

I would think applying gold leaf to the bottom of a cooler before you mount it would be easier then getting solder to co-operate.

[libranskeptic]

The killer argument which seems omitted is that gold's non corrosiveness makes it a permanent TIM. No hardening degradation requiring ongoing maintenance.

[RejZoR]

I've tried aluminium foil as TIM and it didn't work well.

[AngryBeaver]

The difference in temp is pretty small anyways. The problem we are running into isn't as much a problem with the TIM or transfer material... it is the amount of contact being made between the DIE and IHS, then the IHS with the block. We are moving to smaller and smaller manufacturing processes which means the heat is being condensed into a smaller and smaller area.

 

Now with AMD moving to chiplets instead of a monolithic design like intel is still using it does help create more contact points to remove the heat, but those chiplets still have relatively small contact area.

 

So even if you had a super high rate of conducting the contact point is the problem.  I mean which do you think will transfer heat better

 

XXXXXXXXX

==========

       ^^^

 

OR

 

XXXXXXXXX

==========

^^^^^^^^^^^^

 

X = block 

= IHS

^ die

 

So the answer is the second even if you had much less effective TIM... it would still produce much lower DIE temps since the heat can transfer away much more quickly. I won't get into thermal dynamics on this, but there is still a limit of how fast heat can transfer in even perfect conditions.