Possible TIM?

[Jmalvich]

Would it be possible to use a sheet or 3 of gold leaf as a replacement for regular old thermal compound?

 

I want to try, but I also don't want to kill any hardware.

 

Please shun if this question has already come up.

[givingtnt]

9 minutes ago, Jmalvich said:

Would it be possible to use a sheet or 3 of gold leaf as a replacement for regular old thermal compound?

 

I want to try, but I also don't want to kill any hardware.

 

Please shun if this question has already come up.

It would be ok, however they might not be thick enough for this..
gold leafs are insanely thin, way too thin to make any significant difference vs dry.

[Samfisher]

Unless its liquid then no.  The point of having a paste is that it can be mushed and make contact with all the microscopic grooves and holes on the surface of the cooler and the IHS of the CPU thus having a solid connection between the 2.

[Bombastinator]

Carbon actually has a higher thermal conductivity than gold so you would be better off with one of those pieces of carbon cloth I see sometimes used.  As stated gold leaf is craaaazy thin.  You would need a lot of gold, more than a sheet of gold leaf, so really expensive, you would have to crank it down to the point that the gold fills the micro structures of the plates, which might be more torque than they can handle, and it still wouldn’t be any better than one of those carbon cloths which I understand are merely “OK”.

[Guest]

You lap the Cpu IHS plate and cooler when using gold leaf. And you have to be sure there are no crinkles or creases, dents or blemishes. Thermal benefit is not so great. Reason is that copper and silver actually conduct heat better. You would rather use Silver Leaflets. 

 

And I'm sorry Bombastinator, Carbon is far worse at thermal conductivity than gold.

[Bombastinator]

32 minutes ago, ShrimpBrime said:

You lap the Cpu IHS plate and cooler when using gold leaf. And you have to be sure there are no crinkles or creases, dents or blemishes. Thermal benefit is not so great. Reason is that copper and silver actually conduct heat better. You would rather use Silver Leaflets. 

 

And I'm sorry Bombastinator, Carbon is far worse at thermal conductivity than gold.

?
Ah you mean graphite.  Yes.  Graphite is below gold.  Not all carbon is graphite though.  Hex link carbon has multiple times the thermal conductivity of gold.  Gold is 300 some and graphite is 260 some.  whereas hex-link carbon is 1000-2000

 

[Guest]

45 minutes ago, Bombastinator said:

?
Ah you mean graphite.  Yes.  Graphite is below gold.  Not all carbon is graphite though.  Hex link carbon has multiple times the thermal conductivity of gold.  Gold is 300 some and graphite is 260 some.  whereas hex-link carbon is 1000-2000

 

Well I know MX-4 is carbon based, very decent paste. 

 

I have a silver coin lapped both faces. was a Morgan silver dollar actually. 

When de-lidding, I often times use the coin in place of the IHS plate. 

Bummers about the thermal pastes and the metals, the Wm.k difference is very great.

Example, MX-4 only 8.5 Wm.k, while our copper plates and blocks are 385 Wm.k

Thermal pads a tad higher at 6 Wm.k I think it is, or 6.5.... something.

 

The issue now with cooling is actually transistor density much more so than anything else. Getting the heat from the transistor which is instantaneously generated from idle to full load when you run Cinebench and even low idle minor loads generate "spikes" in temperature. 

There is absolutely no way around the temp spikes. From general testing, we can reduce the temp spike range IE: 40c down to 20c using a few techniques. A couple I've tried, sadly nothing on 7nm, but the same rules apply from 12nm. 

 

There was a thread at TPU, something about keeping a higher temperature gradient with the cooling device. Some claiming this also helps reduce spike temps, which is true to a certain degree while there's always a difference between idle and load, typically very greatly in temps. So running a high temp gradient on 12 and 7nm chips becomes rather difficult while each core may consume less than a single watt at idle or parked.