GScube revisited....

[DeadmeatGA]

I wonder how many people here remembers the GScube announced in 2000? It wasn't clear exactly what Sony intended with this machine at the time, as nothing meaningful came out of it, but I now understand what it really was. It was basically a clustering technology demonstrator, and its success(it did run) has led Kutaragi to commit to CELL.

I just wish that Kutaragi ran physics simulations instead of CG graphics; then he would have realized that clustering was unsuitable for low-latency physics calculation required of gaming and he might have settled for a more normal architecture instead.

 

[Grall]

Yes, I too wish that Kutaragi, an engineer with decades of experience of programming, research, development etc, had listened to you, a layperson with no qualifications, knowledge or practical experience whatsoever.

Surely, Kutaragi is the biggest idiot walking in a pair of shoes.

It is good to know there's still people like you around, always ready with some sobering words of wisdom; psx3 (sic) is d00000000med! Just as d00000med as you said psx2 (sic) was going to be, and obviously is.

 

[PiNkY]

I just wish that Kutaragi ran physics simulations instead of CG graphics; then he would have realized that clustering was unsuitable for low-latency physics calculation required of gaming and he might have settled for a more normal architecture instead.

Don't know why i even bother but sampling object physics at 30 or 60 hz is certainly not a "too low latency requirement" for distrbution among multiple cpus...

 

[DeanoC]

Indeed physics and collision detection is very easy to split up for multiple CPUs.

Most middleware physics engines already have support, IIRC Havok already has hyperthreading support on PC which scales fairly linearly with number of CPUs.

 

[DeadmeatGA]

...

Indeed physics and collision detection is very easy to split up for multiple CPUs.

??? Physics and collision detection are the type of computations with heavy data dependency among ojects(Change the state of one object and every other object in that universe is affected) and are better served by going serial than parallel.

 

[archie4oz]

I wonder how many people here remembers the GScube announced in 2000

I think pretty much everybody around here remembers it quite well (there used to be lots of threads with the I-32 involved before you were posting here)... And "some of us" were actually quite intimate with the GScube (both versions)

I just wish that Kutaragi ran physics simulations instead of CG graphics; then he would have realized that clustering was unsuitable for low-latency physics calculation required of gaming and he might have settled for a more normal architecture instead.

1.) What makes you think it didn't?
2.) Low-latency physics? Is that opposed to "high-latency" graphics? :?

Physics and collision detection are the type of computations with heavy data dependency among ojects(Change the state of one object and every other object in that universe is affected) and are better served by going serial than parallel.

Horseshit...! You say the exact same thing about ray-tracing, yet its a problem that's also become amenable to parallelization.

 

[ShinHoshi]

From my experience on coding IA algorytms (I am just a student, nothing more) paralelizing things instead of doing them in a serial way but speed up things a lot.
It all depends what you (the developer) understand for paralelism and at which level bring it.

 

[Magnum PI]

...boring deadmeat anti-cell crusade...

 

[Panajev2001a]

I do not know why Deadmeat is trying to slam the GSCube down in a place where actual developers have worked on the GSCube ( Archie [leaving Fafalada and the others in the salt mines ] ).

 

[Tagrineth]

Re: ...

Indeed physics and collision detection is very easy to split up for multiple CPUs.

??? Physics and collision detection are the type of computations with heavy data dependency among ojects(Change the state of one object and every other object in that universe is affected) and are better served by going serial than parallel.

What's funny is that you just contradicted an actual game programmer. Who even has his OWN FORUM here.

 

[Kolgar]

Ridiculous. Mods? Time to take out the trash.

Kolgar

 

[Sonic]

The mods have been out for a few days. I had a busy schedule at work last week, this week is clear.

There really is no relevancy to this thread other than to spam anything PS3 related. Thread locked.