I've been experimenting a bit with this, and there might be a diminishing returns effect on large tables, or maybe tables of a certain structure.

I tried to run the query over a table that is 20GB in size, and I can achieve 100mb/sec DB read speeds when I try to get the index fragmentation on this DMV in "detailed" mode. This slows down to around 2MB/sec when I switch to "limited" mode for this table.

When I do the same to my 330GB table, I am restricted to the very slow disk read speeds, regardless of the mode I use.

I'll see if I can test loading a subset of the data from my 330GB to another table, and see if a reduced subset subsequently increases the I/O speed.