What cool query! I was an overcloker in the recent pass (prehistory in computer time) with a computer done by me!! I was able to obtain 1300 for a AMD 1000 on a nice Asus using a special air conduct (an external direct pipe to processor with two fans on sides which obligated to re-site the source power in the case, etc, etc.) designed by me and helped by my young brother -who did many contributions- on a Volcano system, when then the record was near 1500 for the AMD 1000. I had problems with memory modules -one crash |:(-. I would import expensive specialized overcloker tested memory modules for continuing forcing it to 1400 or so, and then i decided to stop to play. I know noting about modern state of art and records :-)
So my question is, since metals transfer heat much better that water does, would a solution containing metal ions (probably gold or copper nitrate, or something similar) be better and conducting heat?
First about heat transfer mechanisms. There are three: radiative, contact, and convective.
Solid cooper has an excellent thermal conductivity because is a solid metal. All solid metals are good contact conductors, because the electronic cloud of their atomic structure. Mike is correctly right here. If you use a solution if ions of Cu, your main thermal mechanisms will be convective which is dependant of size and movility of ions and is not so efficient as via contact. I have no data but i would wait no real improvance here by use of ions (i suspect).
Try to calculate by yourself! Take the molar conductivities Lambda for chemicals that you want use, compute the total conductivity, and compare with water
Units = W/ (K m)
Lambda water = 0.6062
Lambda air = 0.0262.
Lambda (solution of I) = Lambda-per-I * (concentration-of-I)
+ Lambda-per-water * (concentration-water)
Units of concentration you amy use are dependant of units used for Lambda-per-specie: molar, massic, etc.
Remember to compute for all ions on solution, that is if there is Cu2+ you may sum also for the corresponding anion generated in the neutral solution.
If Lambda-per-I for your ion I is bigger that for water, then increasing concentration of I you are increasing total conductivity of your solution, but there is a limit: ionic solubility. Beyond that limit, you obtain salt precipitation as Mike said.
I suspect that water is directly contacting with the Coooper waterblock. Right? Then remember that introducing ions in the solution your are increasing electric conductivity in the solution and this would make damage in the computer. This is reason that some water-coolings guys prefer to use deionized water in the circuit. That is, water with no ions (well, still there is a few there). But then one obtains corrosion... Last i know about these topics is the use of "water wetter". Search in google about it.
Options if watter wetter does not work for you needs? A lot of
Re-desing water circuit. But this is also limited by surface of the chip, that is it is.
Substitute passive dissipation in the external radiator by forced one external via pumps air. This is cars enginering!
More radical, use mineral oil and a completely different desing as here
I do not know improvements of that and may offer problems with oil.
In the past some people try to use a refrigerator for cooling instead of a tank as above!
But remember that after of all dissipated air is the mixture of static requirements X dinamic requirements. Yes, Lambda water is 23 times Lambda air but if you increase 23 times the flow of air... This review
claims water-colling performance for the new Volcano 7+.
I suspect that improvement in future coolers will be by the use of some special material. Chemists have desiged special nanomaterials (e.g. Carbon nanotubes) with better thermal conductivity (x3) than Cu or any other metal available in nature. It is now expensive, but probably we will a new generation of dissipators without Cu in future.
Apparently, this industry
is ready to comercialize an improved graphite/Cu material (x2) for cooling electronics.