I've decided to compare Microsoft's DISM v10.0.10586 against the open-source multi-platform WimLib v1.9.1.
I created a nice batch script which saves the start time and end time into a log file. I used a 3.08GB WIM File which contains Win7 wSP1 x86.
The WIM file was created with WimLib 1.9.1 with XPRESS Compression.
Prior to beginning the tests, I checked the file with DISM and with 7-zip and made sure it is clean with no errors.
I deployed the image on an old Toshiba R600 laptop with 300GB Hitach 2.5" HDD. It's slow enough to detect any differences, as you'll see below.
I booted the laptop from a cheap SanDisk Ultra Flair USB3.0 32GB running WinPE 10 b10586 x64. Do note this laptop only has USB2.0 ports, so speed was capped by the USB interface. I used the same script, with the only difference being the executable used to deploy the image to the internal HDD.
I ran the script 4 times for each. Here are the results:
As you can see, DISM actually deployed the image faster in all three runs.
Same laptop, same source image, same source drive, same target drive, same environment and ambient temperature. It's a clear win, which in this case, DISM was faster anywhere from 5 seconds to 50 seconds. It may not sound like much, but if you need to deploy many images every day, it adds up.
By the way, for capturing the WIM image, I would still recommend WimLib, as it reliably creates a clean image which can then be edited using 7-zip (adding/removing files). DISM usually created an image which could be read, but not edited ("dirty"). It required an extra "optimization" step to export the image to a nice clean one.