This whole argument is rather pointless-- a lot of mindless fanboyism on both sides if you ask me. Or all 3 sides, if you are counting Nintendo.
Now, I am not excited about any of the 3 consoles. I won't buy them, I won't play them, that's that. So here's my take on it from a purely technical standpoint. Both consoles use what is basically close to the same approach for their hardware last I checked on its finalization.
CPUs: The XBox 360 uses 3 PowerPC cores at ~3.2 GHz in one processor. These are NOT G5 chips as was speculated, they are not G4s, they are a separate architecture. It's a markedly inefficient one, too. It gets very little work done per clock tick, and therefore, the 3.2 GHz clockrate does not indicate it will perform on par with say, a G5, Athlon 64 or Pentium 4 at such a clockrate. The PS3 will use one of these same cores, working in tandem with a series of vector processors.
Now, the inherent problem with the lofty and much-touted 'Cell Processor' approach is that this isn't a big set of little processors forming a supercomputer-- it's a bunch of vector units. These are simple processors whose very premise of existance is to work on large sets of data doing the same thing again and again-- they were meant to be fed enormous sets of values to apply an instruction across. How useful is this given the context of how dynamically a game's code works? Not very. How useful is it for things like audio decompression, video decompression and playback? Pretty useful, but these aren't terribly stressful tasks and don't burden the main CPU during in-game operation much at all.
Another issue is that each vector unit has very little in the way of access to RAM, and to get instructions to them, a full interrupt must occur which stalls the main CPU to keep the vector processors fed and working. This means the CPU is constantly stopping whatever it's doing to get the vector units new work, which, if the work is even remotely dynamic (as I've said it typically is in a game environment), is work better done on the main CPU rather than wasting time giving it to other slower CPUs not well suited to the type of work being done.
The next fallacy of reason here is the very concept of multiprocessing in gaming. Game engines, quite simply, don't scale well to multiple processors. Tasks such as the software aspects of rendering, AI and physics rely on prior calculations' results being readily accesible in rapid succession, which does not lend itself well to multiple threads of operations that must be held tightly in sync. While it can be useful to throw sound processing, AI etc. to their own core, primary game performance won't skyrocket for it. And more dangerously, multiple weak CPUs are being used instead of one stronger CPU, and since the programmers of the industry at this time are not ready to work with a high level of parallelism, the real gains we'll see for these games will be rather minimal in the realm of raw processing capability.
In fact, according to developers for the XBox 360 and PS3, the CPUs, operating as best they can in a typical gameplay situation, generally hover about the performance level of a Pentium 4 1.4 GHz, which is roughly twice the speed of the XBox' CPU (compared to other processors around such as the Athlon 64, XP and PPC Gx series, this is a staggering low performance level next to the PCs these consoles are touted to 'replace'.) John Carmack, famed lead programmer at iD software and working in tandem with Raven Software in bringing Quake 4 to the XBox 360, has made it clear he is markedly unimpressed with the processing capabilities of the new consoles.
GPUs: Both manufacturers are using GPUs based on ATi and nVidia's newest architectures on the PC side of things-- nVidia's core for the GeForce 7xxx series (e.g. 7800GTX, GT) and ATi's R520 core for the X1xxx series (e.g. X1800XT, X1600..) These are very, very powerful GPUs. However, performance in modern games is bound heavily by CPU performance as well, and the bare minimum for decent operation in modern games as of late has been set roughly, as a bare minimum, at an Athlon 64 3000+ or Pentium 4 3.4+ GHz CPU. It should come as no news to anyone with even a casual interest in PC hardware as it pertains to game performance that the best video card in the business will churn out mediocre framerates if it is saddled with a slow CPU.
We also have the issue that the XBox 360's unified memory achitecture would seem to require the CPU, GPU and RAM occupy the same bus for RAM access. This would mean not only mean a great deal of excess RAM traffic as 3 CPU cores and a GPU attempt to access their own bits of one memory pool (used in lieu of separate ones for cost cutting measures, presumably), but also bottlenecking, as the CPUs will use a 128-bit memory access bus, while GPUs from ATi and nVidia BOTH are designed with a spec of a 256-bit wide memory access bus. This will impede graphics performance-- memory bandwidth from GPU to/from VRAM can be just as important as GPU horsepower and GPU/VRAM clockrate, arguably moreso.
Both consoles have very significant shortcomings and neither one takes any sort of clear lead over the other. The XBox 360's CPU selections, while gimmicky and clearly a response to Sony's latest lie-laced marketing gimmick (It's an army of processors! Run!) show more grounding in reality with 3 of these low-efficiency CPUs rather than ONE of the same chip with an array of vector processors poorly suited for their task and placed in a multitasking environment that gets in the way more than facilitates smooth interprocess communication. The PS3 will likely have higher graphics performance due to the possibly higher VRAM bandwidth, and if it is any indication of the scene on the PC graphics side of things, unless ATi can afford to use their most top-end silicon in the XBox 360 it will be outperformed by nVidia's offerings on the PS3 at the same price point.
It's a very mixed bag. As for the issue at hand, the Killzone video? This is pre-rendered footage. It was released before the PS3 specifications were set in stone,
it was released before the PS3's GPU was even built, according to nVidia, its manufacturer, and it shows a large number of inconsistencies with actual first-person shooter gameplay, including but not limited to the pace not matching any sort of true shooter and rather a railed, pre-sequenced and scripted run through of a level, a shoddily placed gun model added as an afterthought, a total lack of logical aiming (the player shoots at the space between 2 soldiers above him and kills both of them, moving right to left without he, his aim, or his gun moving a bit), actions by characters in the game unbefitting of anything but prescripted actors (e.g. a soldier who gets shot is aided instantly by the other soldiers around him, acting nothing like any sort of squad or real AI), the list goes on and on. The scene showed here is also physically incapable of being rendered on the GPUs built by ATi and nVidia for this generation of PCs and consoles, including ludicriously high polygon counts, 100% smooth and consistent framerates that could never be sustained at such an exact level without game slowdown or frame dropping, lighting and shading effects that have no place in any existing graphical API that yet exists.
Arguments like this so early in the game are simply illogical, especially when marketing hype on both sides has exceeded the point of 'stretching the truth' and entered the territory of 'outright lying'. If I may quote Penny Arcade, they make my next point for me quite nicely with an example of proposed Xbox 360 visuals next to the real deal, the first image being what EA called the 'target render' which is near photorealistic, next to an actual in-game shot. Look forward to a similar scenario for Killzone 2.
We won't know anything about either of these consoles in actuality until both companies stop spinning the story at 9000rpm.
