Ars Technica has a post up on the challenges faced by PS3 developers due to the use of the new Cell processor from IBM. It sounds to me like they may well have jumped "a bridge too far" for a device they want to ship this year. To develop against the Cell without tools, the author says:

What IBM has in mind is what I would call a tiered approach to Cell. At Tier I, there's the "expert programmer" (IBM really means "expert programming team") who codes to the bare metal, manages memory alignment and traffic flow issues like NYC's finest, and just generally makes all the parts of the Cell scream in perfect harmony. This guy, if he exists, is going to be worth his weight in gold. No, scratch that. He'll be worth Marlon Brando's weight in diamond-studded platinum.

As to the "sweet spot - i.e., a compiler that mostly manages the vectorizing for you:

Finally, on Tier IV is the programmer who just wants to port his single-threaded x86 program to Cell in as painless a manner as possible. This person doesn't even care to know anything about "heterogeneous multiprocessing" or any of that fancy stuff. He just wants to see "Hello World" greet him on the screen. Ok, just kidding. IBM claims the following for this highest level of Octopiler hand-holding:

The compiler provides user-guided parallelization and compiler management of the underlying memories for code and data. When the user directives are applied in a thoughtful manner by a competent user, the compiler provides significant ease of use without significantly compromising performance.

Getting Tier IV to work where the money is at. It's also going to be quite painful for IBM to achieve their stated goal of "not significantly compromising performance." I think they, or someone else, will get there eventually. Meanwhile, the PS3 is still due out in 2006.

Sounds to me like they really, really have their work cut out for them. Keep in mind, if development of games for this system is hard, that's only going to drive up costs more - and on the end (the buying of games) where - based on the unit cost of the PS3 - they are going to need to sell a bunch of games to each PS3 buyer just to break even. Based on some back of the envelope calculations, a friend of mine was setting that number at 16 games per buyer. That's steep already.