I explored DXO 5.1 some more this weekend, with a problem image I took a while back using a Canon 10-22 EF-S on my 40D. I was mostly interested in the quality of lens corrections, noise removal, and "capture" sharpening, vs. my standard Lightroom/PS CS3 + PTlens & NeatImage workflow. I made little effort to compare the color balance between the two processing paths, mostly because I've been using a custom camera profile in Lightroom that I developed with one of the popular scripts, and I don't know how to implement a similar profile with DXO. Now that I've decided DXO is really useful, I'll have to figure out how to implement a color profile.
The big surprise was that DXO was much faster. I had assumed that my far greater familiarity with Lightroom/CS3 would make that the quicker option, but it didn't. The reason is that I tend to fuss around with things like the CA correction and frame straightening tools in Lightroom, dialing them in precisely by looking at several different locations in the frame at 1:1. Also, on my Vista PC at least, it takes a long time for an edit copy of an image to open in CS3 after I select that step in Lightroom. DXO chugs a good long time in the final processing step, but the automated nature of the lens corrections more than made up for that with my test.
The quality of the final image was also better with DXO. I was using PTlens to correct the geometric lens distortions, and although it did a great job, DXO was better. The shot I used had a lot of tree branches silhouetted against a bright sky in the corners. Even though I fussed around with the CA correction very carefully in Lightroom, and then applied PTlens under CS3, the branches looked very natural in the DXO version, but had a disgreeable cyan cast in the Lightroom/CS3 one. Details also looked much cleaner in the corners of the DXO version.
Keep in mind that I tried this on one image, and spent about ten minutes using each path. I'm sure the high volume pros would never spend that much time with any processing path. Also, I was trying to correct an image from a Canon 10-22. I love that lens, but I'll be the first to admit that it doesn't approach most primes in terms of CA and distortion. Overall, I'd say the DXO version was clearly better, but that might be hard to detect when printed. At least by non-pixel peeping types.
I'll try to find another image and take more careful notes on processing times, and then upload the results here. At this point though, I can safely say DXO is back in my list of preffered tools.