In some other posts last week, I compared the thermal resistance of what I "thought" were three different Swiftech water-cooling kits -- with that of the SI-120.
Under clearly understood and accurate assessments of Processor Model and the particular model's "Maximum [Thermal] Power," three or more independent reviews of the ThermalRight SI-120 prove that it's thermal resistance approaches 0.14 C/W in a fan-speed range of 2,000 to 2,900 rpm. Assume that there are differences in fan design and quality, but there is pretty much a linear relationship between fan speed and CFM of throughput for a range of 120mm fans of different manufacture. This of course would include two slightly different ranges for 120x25mm and 120x38mm fans respsectively, but -- yes -- a linear relationship.
If you check Sidewinder Computers (online) and look at there water-cooling kits, either "THE" best or one of the very best in the Swiftech line -- selling at Sidewinder for $259 -- is the Swiftech H20-APEX "EXTREME DUTY" Kit with the "Storm" waterblock.
At the bottom of the kit's description page, there is a table of the kit's performance at three different fan-voltage settings (and there is a linear relationship between fan-voltage and rpm's). This was the table I thought referred to three different Swiftech kits, but -- no matter.
The table shows a range of thermal resistances for the Swiftech kit with extremes of 0.125 C/W and 0.157 C/W.
By comparison, the thermal resistance of the XP-120 at the "high" threshold of some fan rpms is 0.167 C/W, and the thermal resistance of the SI-120 is 0.14 C/W. That is, running the fan on the SI-120 to some minimum level between 2,000 and 2,900 rpms will achieve a thermal resistance of that value.
The fan on the Swiftech kit showing a thermal resistance of 0.137 C/W is running at about 58% of the rpms exhibited by the fan setting at 12 V -- that is, a 7V fan setting exhibits a thermal resistance for the Swiftech Extreme Duty kit of 0.137 C/W -- very close to the rated thermal resistance of the ThermalRight SI-120. Check again, but I do not recall that ad showing the actual fan speed, although it does show the noise levels for the different fan volting.
The review at OverClockers.Com by Joe Citarella August 27, 2005, showed the 0.14 C/W value for the SI-120 for a 120mm fan (12V) running at 2,900rpm. Fan speeds as low as 1,800 rpm showed a thermal resistance of 0.15 C/W, and speeds of just over 950 rpm exhibited a thermal resistance of 0.18 C/W -- still better than the Zalman CNPS-7700-Cu with the fan on its "high" setting.
There are, however, other factors you will notice when comparing water-cooling to heatpipe cooling. The idle CPU temperature for water-cooling may be lower than that for a heatpipe cooler. On the other hand, thermal resistance, for any particular processor model of a certain "Maximum [thermal] Power," will completely determine the spread or difference between IDLE CPU temperatures and LOAD CPU temperatures. In other words, thermal resistance determines the minimum and maximum of your computer CPU's temperature variation over the entire range of computer usage.
But I believe there are other things that can be done to a computer, its case, ventilation and other features which will close the IDLE temperature "gap" between heatsink-heatpipe and water-cooling systems.
Even so, it is the variability of temperature, as much as anything else, that we would like to reduce -- starting at the lowest possible idle temperature.