After installing the X8500H, I configured the receiver for 7.1.6 format, designating my three pairs of height speakers as Front Height, Top Middle, and Rear Height respectively. Then I performed an Audyssey calibration to adjust the volume levels and apply EQ.
Of the three upmixers, only DSU utilizes all three height pairs. Both Neural:X and Auro-3D are limited to 7.1.4 processing, ignoring the Top Middle speakers. (Auro-3D can be configured for 5.1.6, but I did not test that this time through.) On the other hand, although it fills more speakers with sound, DSU only actually decodes height information as a single stereo signal, using the entire left side of the room (all speakers) as one channel and the entire right side of the room as the other, with no distinction between front and rear. (Note that this only applies to DSU upmixing, not to native Dolby Atmos soundtracks, which have fully discrete sound from every speaker.) In their favor, both Neural:X and Auro-3D have four distinct upmixed height channels, front and rear on both the left and right.