I never was a fan of watercooling, simply because there are too many things that can go wrong and either cause whatever you're using a water block on to overheat and fry, either because of pump failure, air bubble in the line or just having to much heat built up for the radiator to cool, which runs the risk of W/B cracking water getting heated to the point that the resovior or tubing explodes and turns your case into a fish tank and possibly, if you're near by when it starts spraying water about, can give you that chic freddy kruger melted face look.
Yes, i'll admit that those situations may be rare, but still a risk, and short of using extremely high end W/C componets with a large resovoir cooling performance isn't really that stellar, especially considering the cost compared to a normal HSF cooler. Unless you throw some TEC's under the waterblocks, but that ups the cost drastically, as well as power consumption and potential for the previously mentioned scenarios to occur.
I don't mind fan noise personally, since when i'm gaming i can't hear any ambient noise despite havin a lian-li PC-g70 full tower with 5, 120mm 110cfm silverstone fans which are 54db at full speed, and am running crossfire 4870x2's with software enabled fan control cranking them up to....well jet engine sound levels. That kind of cooling config has served me well in my overclocking ventures being able to get more than half a dozen barton core socket a mobile chips to
between 2.95ghz and 3.1ghz on air, ambient of 25c with idle temps of 31c load temps topping at around 45c, never went over 50c
939 socket 165x2 toledo core opty from 1.8ghz to 3.2ghz,
144 venus core from 1.8ghz to 2.98ghz, and
a 4400x2 toledo core from 2.2ghz to 3.67ghz (magic stepping, yes i have screenies ran 24/7 with a big typhoon for near 2 years till upgrading)those kept at an idle of around 33c load temps of 42c-48c
In all those instances the cooler used was a TT Bigtyphoon, with a 110cfm 120mm SS cooling it in place of the 54cfm stock fan.
and now my 9850 quad core phenom from 2.5ghz up to 3.4ghz without the ACC aided 750SB board using a duo-orb. Idles at 36c, load temps of 46-50c, the NB actually gets considerably hotter then the cpu thanks to the passive cooling "solution"
Even so, i designed and built a TEC case cooler, that is air cooled atm though am playing around with using W/C for the tecs as they reside in a case seperate from the PC hardware. Getting creative with airflow i can drop componet temps 20-25c, with uncreative airflow methods case temp drops by 10-15c depending on ambient. So i'm content with my little design over W/C, but i must pay respects to the horribly designed $400 TEC cooled case that came out a few years ago, because that was my true inspiration, remeding a horrid design.
Before getting the 4870x2's, i had bought a single cored 4870, and the thing idled at like 89c or something, load temps approached 100c at times and there was no software method of fan control....even if they were within the designed thermal envelope...i don't like hardware that toasty and, the thing was actually heating up my room.
Thankfully though Rivatuner upgraded, and Ati Expert tool as well came about giving access to fan control....and depressingly ATI decided to limit the "dynamic" fan speed to a base of 25%, even under heavy load of more then a couple hours i never saw the fan speed increase past 29%
Now with the single 4870 card in a 25c ambient (78F)....even increasing the fan speed to
50% drops the idle temp to about 35c, with load temps going into the mid 60's,
taking it up to 75% drops idle temp to 32/33c with load temps staying in the low high 40's/low 50's.
100% fan speed drops mine to 28/29c with load temps staying in the low to mid 40's.
Yes, 100% is loud, 75% isn't really that noticiable...and 50% doesn't sound any different from the "dynamic" 25% baseline unless you're within 2 feet of the back of the comp.
Now with the 4870x2, you have the problem of the second gpu having the heat of the first blown across it so there is always a 5-15c difference in gpu core temps in a 25c ambient room.
Baseline 25% dynamic temp is 86c/1 94c/2 idle, 94c/1 99c/2 load
50% takes it to 41c/1 52c/2 idle, 58c/1 67c/2 load
75% takes it to 37c/1 45c/2 idle, 49c/1 60c/2 load
100% takes it to 34c/1 39c/2 idle, 40c/1 49c/2 load
Long term heavy load, those temps will prolly rise by 5c or so, but that's still a far far far cry from the base line temps, and it doesn't require more than a free download and 90 seconds to obtain.
As long as load temps stay under 60-65c i'm a happy guy, and it doesn't increase the noise factor by very much to get that range, and you have the added benifit of being able to clock the card higher with more stability if so inclined.