Some initial impressions:
Test case 1: ISO3200 jpeg from 50D
Comparing Topaz DeNoise 4 with jpeg light noise preset against Neat Image on "remove half noise" setting. There is very little in it when pixel peeping. Sharpness is about the same, but I think Topaz has the slight edge. Neat Image has slightly more smoothing particularly of lower frequency noise at these settings, but even then Topaz removes more chroma noise. Topaz does seem a little better at suppressing bright spots. Zooming out to 50% view, neither have successfully removed very low frequency colour blotching.
Test case 2: ISO12800 jpeg from 50D
Normally I wouldn't consider this setting worth using, but is there noise reduction good enough to change my opinion? I tried all 3 settings of Topaz on this one, light, moderate and high. I kept Neat Image on half noise removal setting. First thing that strikes is Neat Image doesn't cope with bright pixels at all. There's red and white spots all over after processing. Topaz does tackle them. Other than that, in detail kept, NI "half" is about same as Topaz "light". NI has a finer noise quality which to me looks better than the coarser noise of Topaz in this setting. The higher settings progressively clears this up more, but smears the details more too. Here I'd give Topaz the advantage again since it deals with the bright specks better than NI. I tried varying the settings in NI but they were not able to get rid of the bright specks. If anything, they stand out more with the other noise removed!
On the user interface, the Topaz preview is pretty slow to calculate even on a small area. While it is calculating, you can't do anything. So if it pops up with an area and you want to move, you have to wait... and wait... switching to 100% view (default was 200%) and clicking maximise window was a bad idea in that case!
My initial conclusion is that for these high to extreme noise conditions, Topaz does seem to have a slight image quality edge over Neat Image, although not enough to make a switch if you already own NI. More of a concern for my operating style is the speed. It takes quite a long time to process previews and the final image. That to me is bad enough to be a deal breaker even if it could offer a revolutionary improvement in image quality.
Disclaimers should apply: this was a test on camera jpeg, raw results may vary significantly. I could have played about with settings some more and got better results with either.
_________________Canon DSLRs: 7D, 5D2, 1D, 600D, 450D full spectrum, 300D IR mod
Lenses: EF 35/2, 50/1.8, 85/1.8, 135/2+SF, 28-80 V, 70-300L, 100-400L, TS-E 24/3.5L, MP-E 65, EF-S 15-85 IS
3rd party: Zeiss 2/50 makro, Samyang 8mm fisheye, Sigma 150 macro, 120-300 f/2.8 OS, Celestron 1325/13
Tinies: Sony HX9V.