Oh so you completely ignored the bit where I say I couldn't notice any obvious differences either?I also tried both demos and could not notice any difference at all which is why I debated the below

**** this, I'm out.
Oh so you completely ignored the bit where I say I couldn't notice any obvious differences either?I also tried both demos and could not notice any difference at all which is why I debated the below
LIAR!!!![]()
![]()
I see Dead People.
rp2000
Indeed, but then again, Criterion said some odd stuff.
Firstly they said the rumour that PS3 was weaker in terms of GPU was wrong, but then say it needs the CPU which can do video work to "bring it back up", so then the GPU is weaker... Seems odd.
There is nothing in it.
Absolutely, graphically they can do the same. However GPU wise the Xbox360 is more powerful which they said wasn't true, but then basically confirmed it was. I found it pretty amusing.
The GPU is definitey weaker. I'm dubious that the cell can be used to offload some processing as that would invlove interruppting the GPUs processing o extract data and reinsert data while maintaing synchonisation. The only things I can think of that would work would be post-processing effects, but these don't take up much GPU time anyway.
The PS3 rendering architecture is designed to be used like this:
The PPE functions as a controller for the entire system.
The SPEs function as the heavy lifters for game logic, physics, and dynamic vertex work. They also are used for some pixel work.
The RSX functions as the main pixel painter, but it is also used for static geometry.
A PS3 engine will start off with the PPE spawning off tasks to the SPEs and acting as the central control point for the engine. Static geometry is put onto the RSX. PPEs start cranking through tasks in parallel, usually setup to double buffer the data they operate on. A SPE will have its code uploaded, then it starts a DMA fetch for its initial data into to one half its local memory, and then it starts ping ponging back and forth: work through one half the local memory while the second half is being DMAed in, then swap. Ideally you have it setup so that you are effectively hiding almost all of your data loading latencies with the double buffer setup and chaining SPEs together where you do animation, deformation, physics, transformation, lighting all going on in parallel.
The heavy vertex work being done, that data is then sent off to the RSX to be rasterized along with the resident static vertex data. So in effect the PS3's Cell RSX combo is one giant unified rendering system. Depending on the nature of your game, your division of labor between the RSX and Cell will be different. It is entirely possible to do all vertex work on Cell or none. And the same for pixel painting.
What the design of the PS3's rendering architecture brings to developers is the unification of your physics, collision, dynamics, and geometry. On systems like desktop pcs or the 360 you have a division between your geometric data and collision/physics data with each of them sitting in GPU and CPU space respectively. The latest Factor 5 interview talks in general about this design of the PS3 and the Lair engine.
The RSX does exactly what it was designed to do, have the pixel painting power to handle 1080p output. The extra vertex power is a nice addition. The RSX in isolation is not very interesting to talk about. Or is a comparison to standard desktop graphics cards or the Xenos.
SPE's can also handle decompression, culling (to reduce load on RSX) and lots of other pretty amazing stuff.