Orginially posted by SonyFan:
I mean, everything I\'ve seen and heard about the G4 indicates that it blows away the Pentium III.. outpreforming a 1.5ghz chip while running at only 500mhz. Now I know that the PIII in the Xbox is going to be slightly modified to handle the load better, but it\'s still going to be dogging the system down when compaired to the other processors MS COULD have put in it.
So what I\'m getting at is that while the GeForce3 working with the G4 can produce those kind of graphics, how is the 733(I think)mhz PIII in the Xbox going to handle an even more advanced GPU? Most Xbox fans I talk to say that the PIII won\'t be bearing any significant workload like the EE in the PS2 does, and the NV2A is what will really be pusing the Xbox. What will the preformance loss be? If the card is a generation ahead of the GeForce3, but takes on twice the workload, wouldn\'t that represent a huge bottleneck in the system? Is that why MS and Nivida are making the Xbox GPU a full generation ahead of the GeForce3.. because the inadequacies of the rest of the hardware will cut the NV2A\'s preformance down to the level of what we\'re seeing in the GeForce3 tech demos?
You know this is actually a great question and it is one that I asked myself early on!
One thing that is great about the Xbox is that the PIII is for AI and small routines only! In a PC the CPU calculates the meshes for a model, in the Xbox the GPU will use do that work and do it several times faster than the CPU. I\'ll explain:
**If you have a monster 3d acclerator (GF2 GTS or higher
and you see slowdown its because the CPU is calculating
the meshes and T&L engine is saying "bring it faster"
but because of PC architecture you are slowed down not
only by bus constraints,memory bandwidth,CPU. With the
NV20 the memory bandwidth is solved with the memory
crossbar but the PC still has a bottleneck in AGP bus.
And as far the CPU the T&L engine is more than ample to
handle the mesh drawings which a texture mapped and
ready to go for animation! The bus problem is also
solved by AMD\'s \'Hypertransport\' technology.
So when the T&L is free to do the meshes then the models will be drawn much faster then you have more triangles per second. Hope this helps!