CPU Thermal compound, Bismuth-Lead alloys.

[I_Saldana]

I was looking at a LinusTechTips video from some time ago about someone who made his own thermal compound.. This made me think about the following.. bare with me. (this is not about the "liquid metal" product)

I know there are solders that contain bismuth with the sole purouse to help desolder crusty old older joints. The bismuth-lead alloy gotten while desoldering melts around 90 degrees celcius making it easier.
There are bismuth-tin alloys for use with reflow soldering that have relative low melting points but still above the safe temps for cpu's. the caution indicated with those is not to use it when fixing lead based pcb's since spillage could cause low melting point joints on the board.

Have any of you experimented with using a thin film of bismuth-lead alloy (which melts at around 90°) as a thermal link between a cooler and a CPU lid?

The film could be placed between the 2 components, then the computer started without a cpu fan.. monitor the temperature while it goes up to near 100 degrees and the alloy melts. Turn off the computer. Let it cool down a bit.
Restart the computer again >>with<< the fan.
If everything goes well the lid should never reach the melting temperature of the alloy again.

Problems would include expansion and shrinkinge of the alloy which would make it flake off in some circumstances, leaking when reaching melting temperature, the alloy not being conductive enough. more I'm sure.

Any thoughts on the subject are appreciated, the results of any tests conducted would be appreciated as well.

Regards.

[wraper]

Thermal conductivity probably will be worse than of a good thermal paste. Thermal conductivity of pure bismuth is 10 W/mK, which is comparable to a good thermal paste. Lead 35W/mK. And in alloys thermal conductivity usually is worse than of the worst of metals it contains.

[I_Saldana]

Thermal conductivity probably will be worse than of a good thermal paste. Thermal conductivity of pure bismuth is 10 W/mK, which is comparable to a good thermal paste. Lead 35W/mK. And in alloys thermal conductivity usually is worse than of the worst of metals it contains.

Thank you for your feedback. I'm still curious about actual tests though. There are a whole lot of exceptions and even mysteries to that rule from superconductivity to magnetism. Even an impurity could skew the balance either way.

edit: this made it looks like it's somewhere in the middle between both ( blue line on page 5 )

https://www.researchgate.net/publication/222197018_Thermal_conductivity_of_liquid_lead-bismuth_alloys_possible_coolants_for_fourth_generation_spallation_nuclear_reactors

[debreuil]

I've been looking into bismuth as a desoldering additive (works great btw [1]), and noticed Bi/In alloys are often used to control for thermal expansion. It seems you can dial it in over a wide range of properties [2]. I like the idea of using it as a thermal transfer bond - maybe even allow it to go to a liquid state towards the higher range (assuming you can mechanically prevent it from drooping all over your chip of course). I'd think the thermal conductivity would be higher in a liquid state, and the heat sink bottom could be patterned for maximum surface area. Well worth trying imo!

[1]
[2] https://www.indium.com/blog/properties-and-uses-of-low-temperature-solder-alloys-part-3.php

[I_Saldana]

Bi/In alloys are often used to control for thermal expansion. It seems you can dial it in over a wide range of properties
Maybe even allow it to go to a liquid state towards the higher range (assuming you can mechanically prevent it from drooping all over your chip of course).
 I'd think the thermal conductivity would be higher in a liquid state, and the heat sink bottom could be patterned for maximum surface area. Well worth trying imo!

Thank you,
I'm glad to hear regulating expansion is documented.
Irm afraid the liquid temperature range would be a tad too high for a cpu to run at comfortably. I was hoping for the 90 ish cap to be reached while not using a fan allowing the compound to flow and stick, then letting it cool.
After cooling I would hope it doesn't go above 65, doing it's job. The solid state, linking the copper and the lid was my hope for making a good thermal bond.

If something could get added to the/an alloy to make it fluid at 65 it would be the territory of the "liquid metal" product I would think. Applied sparingly enough I'm sure it wouldn't leak very much. idk...
The "liquid metal is advertised as having a 70 + W/mk conductivity"

Still I would like to see if what comes down to basically "soldering the sink to the lid" gives as a result.

[debreuil]

With Indium or even Cadmium you can get melt temps as low as 47C (chipquik is 58C, using afaik In, Bi, Sn and Pb). But your right, simple may be best here. I do like the idea of having a non-flat surface on the heat sink, even just scuffed up should improve heat transfer and strength a lot.

Here are some low melting alloy ratios (in the chart half way down):
https://www.rotometals.com/low-melt-fusible-alloys/

[I_Saldana]

Thank you , that is a very interesting chart !

[ejeffrey]

Thermal conductivity probably will be worse than of a good thermal paste. Thermal conductivity of pure bismuth is 10 W/mK, which is comparable to a good thermal paste. Lead 35W/mK. And in alloys thermal conductivity usually is worse than of the worst of metals it contains.

You also probably won't get the layer as thin as thermal paste. 

There also just isn't really much room for improvement.  Thermal paste was a bigger deal back in the early Athlon days which had exposed silicon and very small contact area.  CPUs now have integrated heat spreaders (the best performance ones use indium), and thus much more available contact area to the heat sink.  Heat transport is really limited by the heat spreader and the heat sink itself.  When properly applied, even moderate quality thermal compound will have negligible temperature drop on modern CPUs.

[I_Saldana]

There also just isn't really much room for improvement.  Thermal paste was a bigger deal back in the early Athlon days which had exposed silicon and very small contact area.

Not to mention delidding is what people are focussing on now, which is basically not aplicable with what I suggest.
I am not really implying what I propose will beat everything out there. I am just curious how "soldering a lid to the sink" will perform.

We also don't know if deliding is something that will be doable in the future. I think it's worth investigating and was also wondering if anyone here experimented with it.

I didn't get here from a cooler perspective, I came here from the bismuth / lead solder mix melting at 90+ degrees perspective.
Arou you not curious about the results?

[wraper]

From soldering perspective, it's much safer to solder unpowered chip by external heating. When not powered, semiconductors can survive much higher temperatures. Internally there is a thermal paste between the die and heatspreader or they are soldered with indium on in AMD and some Intel CPUs.

[I_Saldana]

95 is below or just Tj Max for most cpu's. It would only need to get that high and throttle for a small amount of time.

I grant you that it will be hard to get the baseplate to that temp and get the alloy in a liquid state. using a hot air gun on the fins from above might be needed to go over the edge.
I don't even know how one would check if a liquid state was achieved.

And sure, actually soldering the thing on at temperatures higher than Tj Max but off would also be an option. Reflow oven to not go far above temp.. The traditional order of mounting the cpu then the cooler on the mb wouldn't work.. you'd have to detach the frame and let it cook in the oven too.. I mean it's not impossible.

[ejeffrey]

If the die is throttling at 95C the heat sink is going to be a lot less than that even with the fan stopped..  I don't think soldering using the operational heat is really going to work.  It's much better and safer to do it with a heat gun or oven while powered off.

[I_Saldana]

And that would require some seriously precise positioning, disassemblming the socket mechanism making somewhat impossible for joe average to install.

Unless there's a viable option that melts at a couple of degress less and is fully solid at 70..

I think that settles it , for now. Thnx for playing.