I was browsing the various low melting point metals, and found a rather interesting fusible alloy called Cerrolow 136. It's melting point is only 57°C so the hot side of a peltier would easily be able to melt it, it's a lot less aggressive than Field's Metal or Wood's Metal (i.e. it shouldn't eat your blocks and radiator from the inside), and is Cadmium-free (the similar Cerrolow 117 melts at 10°C lower, but contains Cadmium).
The idea is to use a double-loop system: The 'cold' side is filled with Ethylene glycol, and runs sub-ambient (or below zero), and consists of a pump, the blocks for the CPU and GPU, and a compact heat exchanger attached the the peltier's cold side. The 'hot' loop would be filled with Cerrolow, and would just be a pump and a fanless radiator.
The main problem is that Cerrolow is eutetic, so it shrinks when cooling before expanding again. This means that as it cools in the loop, liquid will flow in around the cooled and contracted 'core', then as it cools further it will expand again and cause problems. The only solution I can think of is to keep the loop permanently hot, or use some sort of expanding concertina tubing to handle the expansion (and a really beefily sealed heat exchanger and radiator). I'd have to contact HiTech to find out what degree of initial shrinkage it exhibits (the expansion is to the same density as it's liquid state). If it also shrinks prior to melting again, the problem would probably be minimal.
The other problems would be pumping (could use a MHD pump, but I'm wary running a current through something that connects to the internals of a computer, even if it does form a closed circuit), preventing refreezing, startup (the hot loop may have to be pre-heated if heat in the cold loop builds up too fast. May not be a problem), and heat exchange between the hot and cold loops. My best guess is to use either a high-powered peltier (maybe even 1kW, but probably closer to 500W) and sandwich it between two standard heatsinks, or to use a physically larger peltier (or several peltiers) and fab a custom pair of blocks.
The idea behind using purely convective radiator cooling is that the larger temperature differential will make it relatively efficient, and to prevent accidental cooling back below the melting point. Plus it'd be quiet.
This is purely theoretical for the moment. Cerrolow is likely to be a little expensive (it's commercial use is to allow gunsmiths to take accurate barrel impressions, hence it's eutectic properties), through less so than some of the more exotic non-conductive oils. The only prices I've found peg it at between $20 and $200 per pound. The system might also be rather inefficient due to the peltier requirement. Running it straight to the blocks would mean a period where it would not be sinking heat before it melted causing a nasty temperature spike, and it could never cool anything below it's melting point. On the other hand, the 'hot' loop at least would be highly efficient, both due to the metal's excellent specific heat capacity. Though to get the most out of it would require a very powerful peltier, which might counter any efficiency increase. Less a viable cooling option, and more a "because it's cool" option. Imagine looking into an external reservoir and seeing quicksilver flowing inside.
Does anyone have any ideas on the solidifying problem, or if it even is a problem? I'm not quite sure that once the pump stops it'll cool and solidify as one, or if the residual heat (and thus cooling effect from the radiator) will cause a potentially damaging 'wave' of solidification to run down both arms of the loop to the heat exchange block. Any other obviously unfeasible components that I've overlooked?
The idea is to use a double-loop system: The 'cold' side is filled with Ethylene glycol, and runs sub-ambient (or below zero), and consists of a pump, the blocks for the CPU and GPU, and a compact heat exchanger attached the the peltier's cold side. The 'hot' loop would be filled with Cerrolow, and would just be a pump and a fanless radiator.
The main problem is that Cerrolow is eutetic, so it shrinks when cooling before expanding again. This means that as it cools in the loop, liquid will flow in around the cooled and contracted 'core', then as it cools further it will expand again and cause problems. The only solution I can think of is to keep the loop permanently hot, or use some sort of expanding concertina tubing to handle the expansion (and a really beefily sealed heat exchanger and radiator). I'd have to contact HiTech to find out what degree of initial shrinkage it exhibits (the expansion is to the same density as it's liquid state). If it also shrinks prior to melting again, the problem would probably be minimal.
The other problems would be pumping (could use a MHD pump, but I'm wary running a current through something that connects to the internals of a computer, even if it does form a closed circuit), preventing refreezing, startup (the hot loop may have to be pre-heated if heat in the cold loop builds up too fast. May not be a problem), and heat exchange between the hot and cold loops. My best guess is to use either a high-powered peltier (maybe even 1kW, but probably closer to 500W) and sandwich it between two standard heatsinks, or to use a physically larger peltier (or several peltiers) and fab a custom pair of blocks.
The idea behind using purely convective radiator cooling is that the larger temperature differential will make it relatively efficient, and to prevent accidental cooling back below the melting point. Plus it'd be quiet.
This is purely theoretical for the moment. Cerrolow is likely to be a little expensive (it's commercial use is to allow gunsmiths to take accurate barrel impressions, hence it's eutectic properties), through less so than some of the more exotic non-conductive oils. The only prices I've found peg it at between $20 and $200 per pound. The system might also be rather inefficient due to the peltier requirement. Running it straight to the blocks would mean a period where it would not be sinking heat before it melted causing a nasty temperature spike, and it could never cool anything below it's melting point. On the other hand, the 'hot' loop at least would be highly efficient, both due to the metal's excellent specific heat capacity. Though to get the most out of it would require a very powerful peltier, which might counter any efficiency increase. Less a viable cooling option, and more a "because it's cool" option. Imagine looking into an external reservoir and seeing quicksilver flowing inside.
Does anyone have any ideas on the solidifying problem, or if it even is a problem? I'm not quite sure that once the pump stops it'll cool and solidify as one, or if the residual heat (and thus cooling effect from the radiator) will cause a potentially damaging 'wave' of solidification to run down both arms of the loop to the heat exchange block. Any other obviously unfeasible components that I've overlooked?