Help me understand something here...

Allright Xbox 360/PS3 specs released, and from what people are saying the processing power in either will kick most PC processors in the rear.

My question is, how did IBM/whoever made these processors just whip these out? I mean surelly they didn't start designing these processors right after the Xbox came out... they could have only had maybe 2 years to come up with them.

If they can make such powerful processors for a dang console, why can't AMD/Intel make similar processors for PC's? The gaming systems may be popular but a PC CPU has MUCH more use and MUCH more sale potential.

If I was AMD/Intel, i'd be going WTF how did they do that?

yes the consoles will be faster in procesing, but no way in Gpu gamming. the pc will have a big adavantage, at least in power of Gpu.
how ?
the new PS3 will have 2 TFlops of power, but 1.8TFlops will be from Nvidia next gen g70 gpu.
that mean, that pc in june will have G70 with Sli, one year ahead PS3. The Xbox will have Unified Shader suport, that thing a pc will wont have 1 year from now..

but if we look only at TFlop power, than PC will rule all next gen consoles..(at a to much high price)

for cpu comparision, there are no comparision, the consoles will own big time a pc, till at least 2007 and more...

You also have to have in mind that in consoles the CPU also takes the part of the GPU and also works pretty different. On PC the CPU doesn't do any rendering at all. On the console it does EVERYTHING!!!

look at how many processors there are in the PS3. It's easy to make a processor that only has to do 1 or 2 things, extremly fast. It's not so easy to make a general purpose processor that has to do all sorts of things, and make it as fast as a specialized processor.

That, and the x86 architecture is so old and dated, that it handcuffs how fast you can make things happen with CPU speed. Move to a better architecture and you would see much more speed improvements.

Know Nuttin said:

That, and the x86 architecture is so old and dated, that it handcuffs how fast you can make things happen with CPU speed. Move to a better architecture and you would see much more speed improvements.

Click to expand...

This is a bit of a fallacy. Yes the x86 instruction set is not very good. But, modern x86 processors aren't really x86 processors. They decode the CISC x86 instructions into more RISC-like micro-ops to be executed by the back end. There is a performance hit from the decoding, but it's not so bad. Multiple decoders and/or trace caches are used to reduce this overhead.

Gnufsh said:

This is a bit of a fallacy. Yes the x86 instruction set is not very good. But, modern x86 processors aren't really x86 processors. They decode the CISC x86 instructions into more RISC-like micro-ops to be executed by the back end. There is a performance hit from the decoding, but it's not so bad. Multiple decoders and/or trace caches are used to reduce this overhead.

Click to expand...

wait, so let me get this straight. Modern Processors are not really x86 processors after all? but instead type of processor that in a sense, emulates the x86 code? why dont we just drop the decoders and come up with a new instruction set?

fAlCoNNiAn said:

wait, so let me get this straight. Modern Processors are not really x86 processors after all? but instead type of processor that in a sense, emulates the x86 code?

Click to expand...

In a way, but they are still designed and optimized for x86 code, they just break many instructions into multiple smaller ones. The PPC-970 uses the same scheme with PPC instructions (PPC is much closer to a RISC instruction set than x86 already).

why dont we just drop the decoders and come up with a new instruction set?

Click to expand...

The primary reason that x86 has refused to die is backwards compatibility. The 8086 was introduced in 1976, so that's almost 30 years of x86 software. There is a huge base of software that would take forever to port to a new architecture.

The transmeta crusoe actually does some of this in software instead of hardware:
http://arstechnica.com/articles/paedia/cpu/crusoe.ars

It has a VLIW (Very Long Instruction Word: http://www.research.ibm.com/vliw/basic.html a similar concept is used in Intel's Itanium) core and uses what they call "code morphing software" to decode instructions and do things like instruction re-ordering, branch prediction and register renaming.