1. Yes, the heat must be transfered to the air, but a radiator has more surface area than a heatsink. Ex. I have two heatsinks, both are 120mm square, made of aluminium with an equal number of fins that are equally thick, however one is 1" thick, the other is 4" thick. The 4" one can dissipate the same amount of heat with less CFM from the fan.
I have a PA120.3, it is optimized for use with low speed fans, and it uses three of them. Since the fans are in parallel each additional fan adds to the total air flowing through the radiator. Heatsinks with multiple fans arrange them in series, so each fan adds to the pressure.
Check out the graphs on this page
'pumps' with 'fans'
'system curve' with 'airflow resistance'
'head' with 'overall pressure'
'flow rate' with 'CFM'
The serial graph is an example of the heatsink, when you add three fans the max pressure triples, but the max CFM doesn't go up, only the point where the curves meet is further along the CFM line.
A radiator behaves differently though. Since each fan has it's own bit of radiator to cool, both
the 'fan' and 'airflow resistance' curves stretch to three times their length.
Effectively, this means that three times more air goes through the radiator, and it goes over more hot material.
2. The CoolIT Domino ALC has three problems:
a. Thin 120mm radiator
b. Unknown pump, likely weak
c. Unknown waterblock, likely inefficient.
It's hard to defend this system, as it does seem to be really
bad. The radiator is a thin single 120mm model, the pump seems unimpressive, the tubing is thin AND corrugated (which hampers flow), and the waterblock is probably poorly designed. But I can't find any actual inside pics of the waterblock, or any specs on the pump, so I may be wrong.
EDIT: Most watercooling kits are like ThermalTake coolers, flashy but loud and hot.
3. I know that air cooling is less expensive than WCing, I wanted to point out that $500 Cad could WC the CPU and GPU, not just the GPU. It'll do a better job too.
4. See 1. and 7.
5. and 8. This is a pretty big risk, so I'll address it.
A. Pump stops, CPU block heats up, computer starts to thermal throttle, then auto shuts down due to over heating. Modern CPUs (5 years and newer) have max temperatures that will shutdown the computer if they are reached.
B. I once had a fluid leak because I didn't tighten a fitting properly, and I knew that I hadn't, but left it anyways. Using a Feser One additive kept my system from failing for at least a month before I noticed the leak. (it was a small drip) eventually my sound card's copper PCI contacts corroded. There was a small pool of liquid on the back of the card, and it never worked again. (though for some reason the manufacturer didn't put in 5.1 digital, only 5.1 analogue so it needed to be replaced anyways) The rest of the system was fine though.
Feser One coolant can be sprayed
on a running computer without causing a short. My leak lasted a while, so there was some kind of corrosion over a long period.
There is a bigger risk than air cooling, but most pumps for WCing are pretty reliable. My pump has been running for 2-3 years without any trouble.
7. I'll answer this first, then 6. My hottest core, core0 has delta of 30c with Prime95 at 100% load across all 4 cores. The other cores are 2c - 4c cooler.
I hadn't even heard of Linpack 64, so I'll have to try it out later.
I don't know what the actual dBa of my system is. As a comparison I have a 640GB version of this
drive completely enclosed in a 1/4" aluminum box, sitting on foam, which competes with my main system as the most aggravating part of my computer as far as volume.
6. Okay, first x-bit labs, their delta at full load with 2000RPM fans is 46c, which is hotter than my 30c, and it's unacceptably loud. I never put any of my fans above 1000 RPM. Ever. I can't live with them at that level.
At 800RPM the achieve a 55c delta, with (according to the Antec power thingy) a 255w processor. That gives a c/w rating of 0.216. Watercooling could have the fans spin even more slowly, as 255w in my system would give about 46c.
Also note that the CPU's picture showed a copper colour, indicating lapping was done, and the thermal goop wasn't specified. These are important differences that must be accounted for in any testbed comparison. (Speaking of which, I use arctic silver 5 on my non-lapped CPU. Don't know what Dan uses, though his CPU isn't lapped.)
Lapped CPUs run cooler than non-lapped ones, further invalidating the comparison.
Also note that FrostyTech does not
use a reference fan like SPCR. Every cooler is tested as the manufacturer intended, so the results are useful as a comparison of the package. The results are not comparable as far as which is the best -heatsink-. The moment the fans change everything else changes too.