| captain carot said: First, we're in the same boat about multiple cores, for games as well as apps. I was refering to the stuff where CELL actually seems to shine. Like cloth simulation and other stuff that can be parallelized that good. If not neccessary you wouldn't do that on the CPU. |
Jaguar can do things like cloth simulation, in-fact if you were to do cloth simulation at FP64 levels of precision, Jaguar would be almost twice as fast as Cell. - Doesn't mean you should though when you have a GPU which is 1,000x better than a CPU at said task anyway.
| captain carot said: In other cases, as far as i've always understood CELL, the architecture can be a massive hindrance. Even IBM's improved PowerXCell didn't work out as planned in the end. |
Even PowerXCell had a ton of caveats that were a hindrance.
IBM could never keep pace with Intel and AMD, but that was to be expected as Intel and AMD often had a process and R&D advantage anyway.
| captain carot said: But PS3's CELL actually depends on dividing the workload on as much SPE's as good as possible, which wasn't that easy and sometimes did not work at all. So leading to games with good multithreading, they should definitely benefit from multicore CPU's if they are to use all SPE's. That doesn't give CELL any advantage over 'real' multicore CPU's htough, at least in my understanding. |
The main issue was... Software was just not heavily multi-threaded at the time of the 7th console generations launch, developers were struggling to use two CPU cores, let alone having to deal with essentially half a dozen of them.
It took many years for that to change, even on the PC games didn't really start using 4+ CPU cores fully until the end of the 7th console generation.
But you are right, Cell doesn't have any advantage over a real modern multi-core CPU though, PC CPU's haven't stopped progressing since Cell came along in 2007.
www.youtube.com/@Pemalite
