What's the Difference in Conversions done with ACR vs. DNG Converter?

Question

As an experiment, on my PC workstation I just converted a set of 40 Canon CR2 files with both the free DNG Converter 6.5 and Camera Raw 6.5 (both 32 bit and 64 bit).  With Camera Raw, I dragged the files to Photoshop, then used the [Save Files] button and chose .dng format.  I chose "Medium" thumbnails in both cases.
The DNG Converter and Camera Raw 32 bit took almost twice as long as Camera Raw 64 bit (56 seconds vs. 33 seconds).
The .dng files produced by Camera Raw ended up larger than those by the DNG Converter, about 9% larger on average.
A file comparison shows that the differences are all near the start of the file, and that the latter 80% of the data is identical in all three cases.
The DNG converter files are more different from the Camera Raw files than the two different sets of Camera Raw files are from eachother.
This leads me to ask these questions:
1.  Why are the 32 bit DNG converter app and Camera Raw 32 bit plug-in so much slower than the Camera Raw 64 bit plug-in?  I have an 8 core machine, and in general don't see a 2:1 difference between equivalent 32 and 64 bit code.  It appeared by watching Task Manager that all of them were using more than one thread to do the work.
2.  Are there any plans to release a (much faster) 64 bit DNG Converter for Windows x64 users?
3.  Is there a practical / usability difference between the .dng files generated by Camera Raw vs. those generated by the DNG converter?
-Noel

Answer 1

ssprengel wrote:
You can see how many CPU cores are being used by watching the performance tab with core performance separated, as long as nothing else much is going on.
Not really.  Particular threads do not normally maintain a long-term affinity for particular cores, so the usage gets spread around as Windows does its multitasking stuff and what you normally see is just activity across the board, as seen in the screen grab below (that was DNG Converter running).  Only inasmuch as the overall percentage is higher than 1/n (where n is the number of cores) can we infer a certain amount of multithreading.
And yes to both of you - every setting that I could find is perfectly consistent across the board, including compression.  And as I mentioned before, the embedded previews are coming out exactly identical, both in size and color.
-Noel