Originally posted by seven
Architecture that is 100% dedicated to 3d/2d processing:
Well, this one is easy. Xbox is roughly very close to a PC architecture. And anyone knows how good the x86 platform is for gaming and moving pictures, right? The PS2 however is designed from the core to the rest of the hardware components to run this specific type of applications. The high bandwidths, the small memory - it all adds up. You should read the article from Jon Hannibal at Ars, it\'s a very interesting read and explains why the PS2\'s design is more then excellent for this type of apps.
Graphics?
Anyone who uses the vector units to perform pixel shader effects (reflective/refractive bump-mapping, depth sprites, image convolution, etc.) is just wasting his time... performance would be unacceptably bad. Per-pixel effects belong on the processor that is responsible for handling pixels - this is where the PS2 fails most miserably... the GS supports fewer per-pixel effects than the 3dfx Voodoo 1.
For practical graphics/gaming applications, there is very little different between VU1 in the PS2 and the vertex shader on the NV2A, except for the fact that the vertex shader is vastly easier to program and runs much, much faster. And you\'d be silly if you thought that vertex shaders couldn\'t be parallelized -- in fact, the Xbox already has 2 of them.
VU1 has more temporary storage in its on-die RAM, some minimal integer computation capability, and the ability to branch -- aside from that, it really doesn\'t offer anything above and beyond what vertex shaders offer... and most of the branching/loading capability is used to do inane things like load vertices and matrices (things that are abstracted from the programmer in vertex shader assembly).
As for pixel shaders -- there are limitations in both competing pixel shader implementations (ATI\'s is slightly more generalized with a little more control given to the programmer for address computation texture stages, but DX8 makes it look much more generalized that it is); however, there is still quite a bit of opportunity for programmer-controlled pixel math, compared to virtually none on the PS2. Comparing pixel shaders with vector units is a very flawed concept.
Pixel shaders aren\'t going to become more and more complex -- they are going to become simpler, and capable of running unbelievably fast.
Sony, on the other hand, chose a difficult-to-understand architecture that is nigh-impossible to emulate without dedicated hardware and (unfortunately) has some rather significant flaws. There\'s nothing keeping Sony from fixing the flaws in the current GS; however, the costs of fabricating a chip with at least 16MB of embedded DRAM, at least 16 programmable pixel shader units (each with at least 2 texel fetch units), and running at at least 500MHz would probably eclipse the cost of NVIDIA\'s chip.
Connectivity:
Well, it\'s not necessery, but with the low price of the future upgrade of an BBA and HDD will ensure that they will be enough buyers who are interested to play online and other stuff. With connectivity I also ment the interfaces, that enables you to hook up any USB device, any thing really as long as the application allows it. This also give developers the freedom to implement things that people can make use of. This makes the PS2 have quite a potential if developers or Sony choose to make use of it.
add-ons = FAILURE !
Poll
Topic: Whats the system with the best hardware options? (Read 1350 times)