I currently use the XP-120. The XP-120 has a thermal resistance of 0.167, while the XP-90 is something closer to 0.18. By comparison, all of the Zalman 7x00-Cu coolers show a minimum TR value of between 0.19 and 0.24 when the fans are spinning at a realistically high enough speed.
I wrote to ThermalRight many months ago when the the XP-90"C" version was first introduced, and they suggested that the copper version was worth at most about 1 to 2C in reduction of load temperature in comparison to the original XP-90 under controlled conditions (room ambient, test-bed, etc.)
However, the latest development from ThermalRight is the successor to the XP-120, dubbed the SI-120. The SI and XP versions have the same basic design and the same basic footprint, "wingspan," and "basic" appearance. The SI version, however, has been designed to avoid incompatibility with some motherboards in current production or those which were being produced when the XP version was introduced (requiring the bending of capacitor wire-mounts for the ASUS P4C800-E mobo, for example).
The SI version weighs in at 400 grams, while the XP-120 weighed 375 grams. This is no accident, since the SI incorporates a nickel-plated copper heatsink-base that significantly improves performance, and you will find a recent review at OverClockers.com dated August 27, '05, proving that the SI version exhibits a thermal resistance of 0.14 with the 120mm fan in the test-bed spinning in the acceptable range of 2,000-plus rpm.
By comparison, three Swiftech water-cooling kits sold at Sidewindercomputers.com demonstrate thermal resistance values ranging from 0.125 through 0.15.
The recent review of the Zalman CNPS-9500 cooler tried to control all the coolers in the comparison test (including the ThermalRight XP120) to 25 CFM of fan throughput and something between 16 and 20-plus dBA in measured noise. In that test, the Zalman cooler showed a thermal resistance of 0.18 and the XP120's value was 0.20-plus, but "in the wash," they both "tied" in the evaluation.
In that comparison, one would have to make certain assumptions to judge the cooling capabilities of the Zalman, the first of which would be the assumption that thermal resistance increases linearly with higher CFM, which (obviously) would be accompanied by an increase in noise level. However, this assumption carries little certainty: there is an absolute limit of CFM beyond which an increase in CFM does not change the thermal resistance of the heatpipe cooler beyond its maximum. Since the review never reported "a maximum," we are left with about as much uncertainty as before that particular (American) review was published.
However, a review at Hartware.de -- a German tech-review web-site -- was more realistic in its comparison between the CNPS-9500 and the XP120, employing more realistic fan-speeds and CFM in its comparison between the two. It also acknowledged that the Zalman cooler trumped all others except the XP120. But the table of test results shows that at the same room ambient, the XP120 shows a load temperature of 1C lower than the CNPS-9500, when the "roughly equivalent" fan speeds are kicked up to a realistic level to test full cooling capability (as opposed to pure, 100% noiselessness.)
It should therefore be obvious that while Zalman hyped their CNPS-9500 to equal or exceed water-cooling using deceptive or meaningless graphs of temperature against fan-speed in their advertisements, we actually have concrete proof that ThermalRight's SI-120 really does perform within the "water-cooling range" of thermal resistance values. The comparison showing equivalency between the CNPS-9500 and the XP120 further substantiates this statement.
