crazy idea

[Guest]

Not that this is a serious idea, but I was just wondering. Instead of using a thermal grease or something to interface the heatsink to the CPU, could one use some sort of metal? The CPU is insulated, right, so perhaps mercury or melted and resolidified gallium? The thermal conductivity should be better, I would imagine. I'm just curious what anyone thinks of this, or if it's just not feasible.

 

[jg38141]

well, i believe the thermal grease is a zinc compound- so it really has quite a bit of metal in it and zinc- though I'd have to break out the chem book to be sure- may have better conductivity than mercury or gallium. Of course a thermal compound based on copper- or better- silver, that would be great.

 

[Guest]

yeah, i was just figuring that a pure metal like Hg would probably conduct better than a mixed compound. and copper/silver would have to be premolded, so there would be inevitable air spaces, while mercury (or gallium at CPU operating temps) would be liquid and fill all holes. of course, as long as you're going to all the trouble, you may as well just get a bunch of that non-conductive flouride solution they (the ubiquitous they) used to keep the old crays cold.

 

[jg38141]

Good points- I'd like to see the gallium idea used, Mercury of course would piss off parents and environmentalists, but in an enclosed sytem, a gallium solution would be cool. No puns entended. And "they" are reading your every post so be careful ;=)

 

[jeffg007]

"Of course a thermal compound based on copper- or better- silver, that would be great."

click my link
<A HREF="http://www.outsideloop.com/Merchant2/merchant.mv?Screen=PROD&Store_Code=olcomp&Product_Code=ARC-SMALL&Category_Code=grease" target="_new">Arctic Silver Thermal Compound</A>

Jeff

 

[Guest]

from that link regarding the silver based thermal grease: "Superior thermal conductivity. 4.65 to 5.15 W/m*K depending on the version and particular batch. "

Heat sink compound (metal oxide loaded grease) .4 W/m*K (link: http://www.ai.mit.edu/people/tk/tks/tcon.html note the different unit of measurement, which i've converted)

so this silver grease does outperform a tradition thermal grease by a significant margin, but it's still greatly limited by the grease. Pure gallium, on the other hand, has a thermal conductivity of 29 W/mK (pure silver 430) (link: http://www.webelements.com/webelements/elements/text/Ga/heat.html) Mercury is only 8 W/mK, so it's not really feasible given environmental concerns. However, i imagine with a large copper heatsink/fan interfaced straight to the CPU through gallium, you could probably keep the CPU temp near ambient.

 

[jg38141]

I'm gonna have to get some of that silver compound. As far as galium goes, it's a great and I'd imagine, fairly inexpensive idea. Especially when compaired to current refrigeration based heat sinks. If I were you, I'd make one just for jollies, and then patent the idea. How is galliuum for reacting with other elements? Is it pretty stable? Make one and I'll be customer #1. Seriously though, the idea really isn't that crazy.

 

[Guest]

from a little poking around, gallium bought in bulk (100g or more) is about $1.70 a gram, you'd probably only need a couple of grams to get a thin layer (that wouldn't drip onto the MB) on a socket A processor.

However, gallium is listed as "causing burns in direct contact with human skin", so i imagine it probably oxidizes with air without much hesitation. but that doesn't make sense; i remember melting gallium by hand in my high school chemsitry class. perhaps its very midlyly corrosive, especially at higher temps and over prolonged times. any ideas on how to keep it from corroding? would it react with the CPU itself? i don't really have the spare resources or time to be testing it out myself, and besides, i'd rather keep the idea open.

it should be easy to just manufacture heatsinks with a thin layer of gallium on them; you'd just clip it onto the CPU, and when the cpu got warm the metal would melt and fill all the airspaces. and if the heatsink ever shifted and airspaces caused the CPU to heat up, the metal would just remelt and refill the new holes.

 

[jg38141]

I'll see if I can't get some and start messing with it. The hardest part will be making sure it stays between the chip and heat sink and doesn't melt off onto the board. It souldn't react with the chip I wouldn't think, so maybe a small metal cover that fits around the actual chip could trap the gallium against the chip and heat sink while also increasing heat dissapation by surrounding the chip with a cunductive metal cover. this would also, depending of the tightness of the seal, help block any air from reacting with the gallium. I'll need to save a little bit to get a flip chip-probly a duron- to try this with, if anyone gets a chance to try it sooner, let me know, I'm very curious. BTW- where does one aquire gallium?

 

[Guest]

one can order it from a chemical supply company. of course, since you're probably not going to want to get a huge ingot, it's going to be a little more expensive than the $1.70/oz i quoted earlier. no more, i would hope, something like $25 for 10 oz in granules. i'm very interested to see if this actually works.