Jon
Member since: May 3rd, 2006
Jon's Latest Comments
Blog Activity
| Blog | # of Comments |
|---|---|
| Joystiq | 2 Comments |
Featured Stories
Schilling says he could lose $50 million of his own money in 38 Studios implosion [update: Chafee responds]
Posted on May 29th 2012 10:00AM

Developer: PS3, 360 graphics "impossible to tell apart"
May 3rd 2006 1:46PM (Joystiq)Hold on, are you saying the framebuffer on PS3 has to go through main memory? That last step GPU->RAM has me confused. All of the compositing and whatnot should happen in graphics memory right? Both the 360 and PS3 should behave this way. From my understanding, the benefit of the cache on the 360 is that geometry and textures generated procedurally don't have to go back to main memory until they are needed; they can just sit in the cache until the GPU is ready. My argument is that the PS3 can do the same thing by using the local store on an SPE to hold procedurally generated content so that it never needs to touch main memory either.
My theoretical diagram would look like this:
Disc-->RAM-->Cache-->CPU-->SPE-->SPE LS-->GPU-->G RAM-->Output
I've never written a game using modern graphics hardware though, so I could be mistaken about how this all works.
Developer: PS3, 360 graphics "impossible to tell apart"
May 3rd 2006 1:00PM (Joystiq)I follow what you're saying about the 360. However, is it not true that the PS3 can pull a similar trick by running procedural synthesis on an SPE or two and dma'ing the results straight over to the GPU? I don't know much about the 360 hardware at the bus level, so maybe it is possible that the GPU can access the CPU cache faster on the 360, but I don't see an obvious reason why the 360 should be significantly faster at this.