| NJ5 said: @TheBigFatJ: That's exactly what I think about the Cell even though I'm not a professional games developer (and I likely won't be since I'm in a different career path now). If the Cell had two or three PPEs, it would be no harder to program than the 360's CPU, and the SPEs would still be there to help with whatever tasks are better suited for DSP type work. However, the cost of doing this would probably have been prohibitive at this time. Maybe in the next gen. As it stands, the architecture of the Cell is probably a burden, since you have to force the SPEs to do things you'd rather use the PPE for. I imagine this would be especially true in a team with lots of developers, where each one would like to fight for a bit of PPE power, but will often have to settle for programming their components on the SPEs. |
This is the crux of the issue with the Cell -- you have precious little general purpose CPU time to work with compared to the 360's 3 cores. The cell was originally designed with 2 PPEs, I believe, but because the PS3 was prohibitively expensive the Cell was castrated/simplified for the final design. I'm sure Kutagari was disappointed, because this makes the Cell a much less balanced processor for game development, but he had to cut a lot of things he originally envisioned.
The SPEs have their uses. However, in game engines situations where you can rely on them heavily or boost performance that would otherwise be assigned to the PPE are *not* the common case. The timing issues that I touched on are part of the problem. You almost have to design a game around them to come up with situations that would use the floating point power available through the SPEs efficiently and consistently. With general purpose CPUs, it's much easier to keep them doing whatever you need. Still, coding multithreaded engines is no walk in the park and definitely requires design and planning as well.
@ TheBigFatJ
This has been talked about here:
http://www.vgchartz.com/forum/thread.php?id=11075#end
If you want to revive the thread I would be happy to discuss the points you've just made.
All I see is you regurgitating talking points, theoretical numbers, and making odd statements to the effect of "ratchet and clank is proof of the cell." Talk to me from the standpoint of a programmer (specifically, an engine developer) or computer engineer where we can talk about actual tradeoffs associated with the Cell, in specific, with regard to game development.
I don't undertand why you're so convinced of the Cell when you seem to have very little actual understanding of it. It's a neat processor for some purposes. It doesn't have the kind of general power that game developers would like.
Game development is running into a very serious problem right now. Games are expensive to develop and aren't necessarily selling enough to justify their development. Many of them aren't even coming close. Saying that a developer is lazy because he didn't design a game from scratch around the Cell is not an oversimplification, it's just foolish. Developers have to make a profit -- putting those kinds of resources into a game for a tiny installed base that almost certainly won't profit doesn't make business sense.
Consider that the size of the installed base of a console is a feature -- almost as if it were an investment toward good future third party games. If you doubt this, look at the example of the dreamcast.
I'm nit picking a little, perhaps, but any given HD TV owner is not much more likely to buy a PS3. I have a sweet display, for example, and I don't have a PS3. Perhaps HD TV owners are slightly more likely to buy a PS3 that SDTV owners, but I don't think it's going to be all that high at this point. Many more people buy HDTVs for football than PS3s here in the US, and many of those people never consider even upgrading their DVD player for HD content.
