Just to follow up on this, I had a chance to do some testing in GB3 (64 bit) today.
Mac Pro 4,1 flashed to 5,1.
Intel Xeon E5520 @ 2.26 GHz (memory runs at 1066)
2 processors, 8 cores
Results: each config. run 3 times, these are the best results from each
48 GB of ram, six identical sticks in slots 1,2,3 & 5,6,7
64 GB of ram, 16GB in slots 1 & 5, 8GB sticks in 2,3 & 6,7 (the config. at the top of this thread)
24 GB of ram, six identical 4GB sticks in slots 1,2,3 & 5,6,7
32 GB of ram, eight identical 4GB sticks filling all slots
Not sure what it would all mean in terms of real world performance, but just thought I would post what I found.