It's pretty apparent that the HR-03 stands out from the crowd. The most important temperature is at load, and the HR-03 paired with a fan manages to keep the graphics processor 25°C cooler than the stock cooler did. That is incredible, especially when you consider that the Radeon X1900 XTX stock cooler actually does a pretty decent job of keeping the core cool (although it's a bit noisy). It's also amazing when you consider that it cools down the core almost 10°C more than the former VGA cooling champ, the KuFormula V1 Plus, running at full fan speed!
Of secondary note is the HR-03's passive (fanless) performance. The numbers here aren't particularly impressive until you realize it's doing basically the same job as the stock cooler, but completely silently, without any fan at all!
All the same, I'm not sure I would want to run the HR-03 fanless on an Radeon X1900 XTX. It runs a bit hotter than my liking, but certainly seems within the card's tolerance. And I doubt gaming sessions will get the GPU as hot as ATITool will.
Noise was measured with the side of the case off and a microphone held next to the video card; this was done with no VGA cooling fan, and then again with the different coolers. The resulting delta numbers aren't a very useful metric on their own, but they can serve as a useful yardstick. Note in particular that all of the coolers that measured 3 dB(A) over ambient case noise were almost completely silent, while the stock cooler - at 15 dB(A) over ambient case noise - was really, really loud.
The HR-03 is a strange bird when it comes to noise. We can't really make any definitive observations because it's as loud as the fan you choose to put on it. If you don't use a fan it is completely silent; with the particular fan we chose, the noise generated was in the middle of the pack, but the cooling results were spectacular. I imagine a quiet 120 mm fan that pushes a lot of air might produce even less noise and better thermal results.