I do take your point and I hope that it does work for them, but from the dev reports I've been seeing developers don't like it. If you were the head of a dev team/company which would you prefer to develop on, a system that is easy to make a game for quickly & looks good, or a machine that needs to be done "the old fashioned way" and is therefore more expensive in man-hours and cash to complete? If you look at it that way they are behind the ball before you even start comparing the system specs/units shipped, etc
Just so you're aware, I am actually on a games development and design course at uni, so I think my whole perspective on the PS3 is coloured by the possibility of my ending up making games for it some day.
The thing you have to understand is that when I say it's "harder work", I mean the machine itself has to work harder. The games developers, according to my logic, will either not notice a change, or else will actually have LESS work and difficulty to contend with.
Developers fall into two categories: programmers, and designers
programmers are the hard code guys. They're the ones who wind up writing all the C++ code on which a game engine runs. Not just that, they're responsible for creating the tools that the designers (mappers, skinners, animators, modellers, etc) use to turn all that code into a game.
For the designers, there is no visible change. You're still using an SDK provided by the programmers. You move objects and entities around, put them together, and come out with your finished product. It's not your job to understand how the machine code works, it's your job to use the tools you're given to produce a finished, artistic product.
Now, I'm no programmer, but I rather imagine that from a coding perspective, it's easier to tell the system to run everything off the CPU, rather than having to program in the subroutines that tell the machine where to send each individual process. Let's say you have a game that uses bump mapping, HDR lighting, a Havok-based physics system and 64-bit vertex shaders (alongside the hundreds of other processes). As things stand, a fair old chunk of the processing done by the CPU is actually responding to the code that tells it where to send each of those processes. the HDR is sent off to one section of the graphics card, the vertex shading to another, the bump mapping to a third, and the physics goes off to physics processing. usually, that's handled by the CPU, but nVidia are releasing joing phyics/graphics cards now.
I'm no expert, but to me it sounds easier to run everything on central processing power, rather than having the CPU shuffle and sort each of those processes off to its designated subsection of the system first, THEN process it, then send it back to final processing to be displayed. The system has to do far more, in terms of calculations per second, but from a coding perspective, I don't think it would be more difficult to do things that way - precisely the opposite. It'd be simpler, faster, easier and cheaper, because you drastically cut down the number of code lines the programmers have to write - which leaves them with more time to include the SDK features that the designers are asking for.
As far as I can see, it's a win-win situation.
You're incorrect. If you really want an explanation, I cooked up some diagrams and such a while back when I was researching it, and have some great links, but it gets pretty hairy.
Quick summary: The PS3 still has a graphics card, and it's still a fairly traditional one, but the processor takes the basic approach of taking off the last 10 years of design to make it smaller, and then hitting copy-paste 7 times. If your code vectorizes *and* multithreads well, it'll fly. If it doesn't... figure on getting something like what my laptop can do (1GHz G4). The launch titles are always somewhat rushed, so they'll likely be doing something like: PPE (the non-wacky core, 98% identical to an Xbox360 core) runs the game, SPE 1 (wacky vector core 1) decodes video, SPE 2 decodes audio, SPEs 3-7 sit idle most of the time. The real question of the PS3's power is whether game developers will be able to take anywhere close to full advantage of the SPEs before the PS3 is totally obsolete.
To me, this still implies less work spent shuffling processes about and more time spent on the core, which to me sounds like easier programming.
Feel free to correct me if I'm very very wrong, however.