is it just me or, firstly did Guru3d not link the site they supposedly got the info from and, as is usual this is just an educated guess from a look at the silicon under microscope.  However in this case it seems miseducated, the line they've decided is the 9th row of SP's looks nothing like the others,  whatever it is its by no means identical to the other rows so is very unlikely to be another row of SP's, but some other part of the core logic.  1 through 8 look identical, 9 does not.
As for redundancy, theres nothing to say there is or isn't any.  Anandtech talked about redundancy in their article, I don't think they claimed they knew there was any.  Infact they hinted that you had to have redundancy on a core the size of Nvidia's latest as if you can't get a core working you lose a lot of the waifer each failed core.  While AMD have gone for a tiny core,  losing one doesn't lose you anywhere near as much cash as Nvidia cores failing.  The smaller you go the less need for redundancy,  the less redundancy the smaller the size, the smaller the size the cheaper they get and more per waifer etc.