[QUOTE="Steppy_76"][QUOTE="superferret2029"][QUOTE="Steppy_76"][QUOTE="superferret2029"][QUOTE="Steppy_76"][QUOTE="superferret2029"][QUOTE="bungie93"]superferret2029
The 8800(a DX 10 card) in a unified shader card, and doesn't share much of its architecture with the 7800(a fixed funciton shader DX9 card) where did you get THAT from?
I never said that bandwidth isn't important. It is one of MANY factors that comprise processor performance. Using it by itself is worthless, and doesn't take into account how much bandwidth an application is actually USING. The folding@home application for the PS3 IS something that actually WILL use the bandwidth, games are NOT.
Lastly, you are correct in that that is ALSO part of the cell's scalability. You are incorrect in assuming it has anything to do with performance in the PS3. The PS3's version of cell will ALWAYS be one PPU, with 7 SPU's(unless you think the cell inside the PS3 can spontaneously fab itself some more spu's). What that is talking about is that the cell can be tailored to other applicaitons. For instance, if you were to use on inside of a TV to control it's UI you could use a 1 PPU, 1 SPU version of the cell. It also refers to being able to use more than a single cell as they do in servers(ie the lastest one from IBM uses something like 9000 cell processors).
The Xenos is NOT based on an existing PC based GPU. The RSX IS. The newest cards from Nvidia have moved to unified shaders...cards with GPU's AFTER the 7800 the RSX is based on. The Xenos is a unified shader card, and is much closer to a DX10 card than the RSX is. How you think comparing GPU's on video cards(and also restricting them to similar pricepoints) has ANYTHING to due with this argument is beyond me. Just stop dude, you don't know what you are talking about.
if you knew what you are talking about, you would know that there are many different variations of the 8800 not all of them DX 10 compatible. look it up on google. you are boring me with your lack of knowledge.for example there the regular 8800 then the 8800GT then the 8800GTX etc. really look it up and their specs ie. which is DX 10 compatible before you post
And if you actually took into account certain things you would realize that it would take maybe billions to build a new fabrication plant just to feed the 'unique' needs of the ATI card in the 360. Man, its obviously based on some ATI existing arcitecture. they always base it on existing arcitecture. think about it. if you were ATI/AMD what would you do: spend billions to build a new fab plant or basing it on existing architecture. SO just like the RSX the ATI xenos is based on existing architecture. OLD architecture if you will since it was based on an ancient architecture because it went into R&D and development before PS3.
By the way you seem clueless as to when bandwith is utilized. first, check the meaning of bandwith. Looks, an architectures bandwith is the single most important factor when it comes to speed. each processor or component has a bandwith and the motherboard has a bandwith a well....ie. system bus. this means how fast it is able to transfer data from say Gate A to Gate B or data moving from the RAM to the CPU or FPU. Or the GPU and CPU or even within the CPU. every application i mean EVERY application )ie. software uses bandwith) Now the question is which badwith does it use??? Folding@home uses the Cell (CPU) predominantly since it is not as graphics centric as say MGS 4 for example. But the fact remains that they ALL use bandwith.Foldin@home would mostly use the cells bandwith to transfer information say from the cpu to the fpu to the cache then to the controller for example. meanwhile MGS for would do that butnot as intensely as folding@home since it will mostly tax the gpu and the gpus physics capabilities. meaning MGS 4 would mostly use the bandwith within the GPU. irrespective of the application, they will all use the RAMs bandwith. so now, do you realize what the term bandwith means? since MGS 4 employs a more immersive experience, both sound, graphics, and processor wise, it would overall use more bandwith than folding@home which is not meant to wow a user so it mostly uses the FPU for geometric calculations.
Now as for the SPUs argument, i will not bother to address that. go to the IEEE website, or wikipedia or IBM website to understand what i wrote earlier. you type a lot but you forget to lookup what youre typing first IMO
By the way, the Xenos is not based on existing cards. you are right. its based on ancient cards circa 2007. and just because a card utilized unified shader capabilities doesnt make is close to being DX10 capable. DX10 by the way, is mainly just instruction sets written (ie. APIs) to interface between newer gpu capabilities and the sotware programers programs. just because a card has DX10 apporval stamped on it doesnt mean its a beast. it just means that it has the cardware most compatible with those instruction sets. DX is an API just like opengl is an API. meaning its all software based, read it up
Log in to comment