That's a moot point for XSX - extra bandwidth isn't some sort of an 'add-on', it needs to be there to facilitate the faster GPU, regardless of the tradeoffs associated with it.And there's no realistic situation in which it won't be better than a single 256-bit bus.
Yea because it used debug-channel for CPU to read from GPU ram. It was an architectural flaw, but likely one that had more to do with timescale of RSX integration than much else. RSX still had full 25GB/s access to both memories with only minor differences in access latency(and again, that doubled possible bandwidth available to the chip).For example, IIRC Cell's read from GPU DDR3 was incredibly, unbelievably slow, at something like 1/1000 of percent of actual memory pool bandwidth.
Wii, GC, PSP to name a few more, if you don't want to count PS3 because CPU/GPU didn't get the same treatment.I can't think of any split memory pool console that would be as fast or handle being treated as one single pool as well as this.
Ultimately we haven't had to work with per-device split-address ranges since PS2 (or NDS, if we count handhelds).