I understand. I'm just saying that the margins we all used to care about are no longer so fine. Suppose you have a heatload of 250W. You have a 120.2 rad with quietish fans. The rad may not do a great job of cooling the water resulting in a highish delta between water and ambient.
Water is amazingly good at removing heat. It take a massive amount, relatively speaking, of energy to raise its temps by one degree. The rads are also very good at exchanging that heat to air with the shear amount of water that passing through them, mass flow rate. So we see air/water deltas well below 10C. More like 4C at idle. Trouble is getting the heat out of the cpu core and into the water, there just so much energy you can transfer over a limited surface area. And the blocks are now at the point of diminishing returns, massively improved re-designs only see a degree or two improvement. And they simply can't get any better. Apart from the concave/convex IHS to deal with, between the water and the core is the block itself with its proportion thermal resistance TR. Then there's the IHS to solder to core encapsulation again with its TR, then to where-ever the DTS sensor is located on the core. All this adds up to an, err, interface bottleneck for the want of a better description.
Rads, cooling can be odd things to deal with, chucking more rads in a system doesn't always give the logical result. Doubling the rads may have little effect if the system isn't heat saturated. More pump flow makes next to no difference once you're past a certain point. Even increasing temps from the pumps heat load. And there's the mass flow rate of the other coolant, the air. The rads are very sensitive to the cfm of the fans, its the one thing that we can really improve on.
Does it matter?
I contend that it doesn't. Blocks are incredibly efficient and the chips can stand very high temps compared to five years ago. So what if the water-air delta is very high? So what if the resulting cpu temp is 60C? All I'm saying is that a 120.2 will handle enough heatload to stop a modern chip from overheating.
Up to a point no I guess it doesn't. With GPU's it makes little OC difference, sub 80C they could care less. My early C2D cpus chips had a tendency to get flaky with higher temps. The last few wolf/yorks I've had were less bothered by higher temps. After all the typical Tjmax on these things is 100C, the designers are happy for the IC to get plenty hot before they even worry about throttling. Remember though, the Engineers rule of thumb says for IC's
-10C = ~+2% frequency increase, OR x2 life at same freq, same voltage
By then I've sold it anyway, so meh. I have to say though these Ci7's are a different beast. They can take much higher temps and remain stable, but the heat they dump is amazing once the volts get going.
I have a Noctua on my CPU with a single fan and an accelero on my 4870 with a single 120mm fan.
CPU idle is 35deg and 4870 idles at 40 deg.
Under load CPU hits around 57-60 deg, not sure on GFX card though.
Those are good temps really. Water is nice, but don't expect it to change much, at least not on the overclock. Like I said a PA120.2 with reasonable flow, any DDC. And reasonable fans, medium speed Yates, with dissipate 250W all day and keep good temps with those two. If you go to silent fans temps be worse than you're getting now.
