In other words 'We shot outselves in the foot buy trying to change industry standards which in the end, the gamer loses overall; Developers spend more time and money trying to figure out how our system works when they could have vested that effort into creating innovation and ideas'- Ahh! My magnifying glass works wonders!
Wow, that's the biggest piece of Public Relations B.S i've read from Sony in a while. Of course they'd say that because they want developers to make it for them first so everything is a timed exclusive... however the fact of the matter is that PS3 is harder to program for, and that slows development times all the time. 360 is made as a developer friendly unit with easier to program code and hardware. You would ge the most of the 360 if it was exclusive to just that system and not the other wy around.
On the PS3 if you want any kind of decent performance out of the CPU, it means going to the SPE's. The way the SPE's work is that instead of a cache that is filled/flushed automatically (like you have in the 360 CPU or your average PC CPU) you have a local store. The local store is a bunch of memory that's as fast as cache, but its contents have to be explictly managed by the programmer.
Now the thing with local store is that it's very fast which lets the SPE's fly...but space is very limited. This means that if the game engine wants the SPE's to work on a set of data, that data has to be organized in such a way that it can be split up into small chunks so that it can be fed to the SPE's a few pieces at a time. Coonsequently, the game engine has to be designed to work this way: you can't just work with all your data as one huge intertwined bulk, like you could in the old single-threaded apps that ran on one core. You have to instead work with task-level parallelism, where you break things down into small tasks and assign those tasks to threads (in the case of Cell, SPE's) as they become available.
On the 360 or a PC running two cores, this isn't really necessary. You can still get a way with the "one thread for everything" approach for the most part. Or you could also take an approach where you run one major subsystem (for example rendering, or physics) on a seperate thread from the main game logic and get good performance, since you're working with 2 or 3 cores. However the thing is that doing your threading task-level (breaking things up into small groups) can also give good or better performance: it helps ensures that none of your cores or hardware threads are sitting idle at any point, and the partitioning of data into small chunks helps to keep things in cache where they're fast.
Now it's of course debatable whether doing things "the PS3 way" would actually lead to improved performance on the 360 (especially since what we're talking about here is mainly theory), but at the very least it would be hard to argue that you at least wouldn't be any worse of on the 360 side of things.
It's something actual deveopers have said in the past. I can explain to you the technical reasons why, if you'd like.Ontain
well, developers have said that starting on ps3 and porting to 360 is easier than viceversa but not that the 360's end result would be better than if it was the lead platform.
It's been suggested by some very smart people. But it's something that's pretty much impossible co conclusively prove or disprove, so I don't think any multiplat developer is going to take it heart or anything.
It's probably more likely that studios will consider using PS3 as their baseline to simply make sure that their 360 versions don't end up looking noticably better than the PS3 version (because of memory, GPU, not having time to really use Cell, whatever).
Log in to comment