That is far too long for a normal development cycle of a video game, especially a AAA title. Just sit back and think how long that really is. Think about a game like Grand Theft Auto which is a HUGE game technology wise, they don't take quite that long between IV and V and they've been developing other games in the meantime.
PD need to get their act together, they can't keep leaving 5 year gaps between the main games, people are going to lose interest.
The thing about Rockstar is that it's not just one game studio. In 2002, after 2K had bought up lots of different teams, they gave them all their own little "Rockstar" monikers. DMA Design, the makers of the original GTA (and Lemmings!) became Rockstar North. They have, along with the other studios, helped out with other games, like Red Dead Redemption (Rockstar San Diego / Angel Studios) and Max Payne 3 (Rockstar Vancouver / Barking Dog) - as did all the other Rockstars, for their own specialities (e.g. Rockstar Lincoln / Tarantula Studios is primarily QA and localisation now). As such, I suspect their main focus has been on GTA (IV, "Episodes" and V). GTA IV came out in 2008; GTA V is due in 2013.
The problem is that it's already two years since GT5 was released, and we won't see the PS4 for at least another year (maybe more), and we all know the likelihood of a launch release for GT6. So either it'll be 3 years for PS3, or closer to 5 for PS4. Ideally it'd be both, but maybe they'll do something in the middle for a half-baked version. Then again, part of the reason for GT5 being so disappointing might be because they pulled a GT3 again (if they'd jumped from GT1 to GT3, that is), and have been developing GT6 for PS4 alongside GT5 for PS3, and struggling to "downscale" the GT6 / PS4 ideas to fit on PS3 (karts are the only concrete example). Maybe.
And the dev kits being sent out now are still not finalized. So for all this time what have they been developing GT6 on?
...
I think I remember reading that PD's development platform is almost entirely x86. It's quicker and easier to prototype
experimental things on general, high-powered hardware and then attempt to pare it down and streamline it for a console than to try to prototype straight to the console (at least initially). I think Sony now even provide virtualised emulations for testing certain aspects of Cell programming, and there are certainly "Cell simulators" available unofficially. It's feasible they can do it with the next gen, which is actually a great pre-development (of the console hardware) feedback tool for Sony to issue to game developers to play around with before getting any real hardware delivered.
In that sense, it makes little difference what the target hardware is, since all you do is change the "conversion" process. Of course, with experience the conversion will be quicker and you could even write directly to the console hardware's requirements first, without having to prototype on a more general purpose machine. Which is why getting an idea of the target hardware in advance is such an advantage.
So, in the case of the AMD A10s, Sony might be saying: "imagine this, with more clock speed here, and larger bus width there, plus AMD's newest graphics architecture you got to play with a few months ago, etc." (assuming they are going AMD all the way) - it won't take long for developers to grasp what will be possible and cater their initial work to suit, so as to make inroads on development before getting the real hardware. That said, proper optimisation can only occur on the real hardware, and that's where PD historically make their money.
