Jonathan is correct. Noise Ninja or Neat Image provide far better noise reduction. Why? Because they use noise profiles which are essentially documents of a camera's specific noise pattern, at different IS0s. The PS tool relies on generic noise reduction. I personally prefer Neat Image, not only because i prefer its layout, but because Neatimage updates its product far more frequently than Noise Ninja. Noise Ninja has been stuck at ver. 2.1.2 for a while now... I've used it myself, and have found that it doesn't compare to Neatimage 5.2 (I have both pro plugin versions). The Neatimage website has far better support for camera profiles, where you can select from multiple consumer submitted profiles, and read about the RAW converters and parameters which were used to generate the profiles. Still, both are far more effective than the PS tool.
As soon as you become acquainted with either program, you should move from profiles that others have created to making your own. Everyone has a slightly different workflow. For example, I've found that your RAW conversion should mirror the RAW conversion performed to generate the noise profiles you're using. Why? Because if you do something differently to your image than how the profile image was generated, the noise pattern will differ, and reduce the effectiveness of noise reduction. Noise reduction should be performed right after raw conversion (JPEG in-camera parameter settings should also mirror the JPEG noise profiles). If you use noise reduction in Adobe Camera raw, for example, and the noise profiles you downloaded didn't have ACR noise reduction applied, then there are different pixel values. I've found that you can get away with subtle differences, but larger differences between subject image and noise profile can actually introduce noise. For whatever ACR parameters you set, you should create noise profiles in exactly the same manner. I've found that if you do this, it's unnecessary to do initial noise reduction in ACR, because an exact profile will pick up both the luminance noise and the color noise... provided your noise profiles are very high quality.
