cell processor usable for games

[Lacostiade]

I think u all heard about the cell processor, but most of u don't even know that it's usable for games, as IBM says.Each different configuration of the cell processor uses a different type of interface.The XDR,which allows it to run at 3.2 GHZ ( it is used in ps3) and the FlexyOL interface, which runs at 6.4 GHZ.
If gaming is to be considered, the one which will be used is the one with the XDR, becasue i think the other one will be damn expensive, and becasue the XDR is better for gaming ( that's why it's put in the ps3).

and plz, i want serious comments.

 

[cafuddled]

I really dont know if you are making a statment or asking a question but it should be simple logic the one that will be the best is simply going to be the one with twice the clock speed (Even though gaming does not need that fast a proccessor yet it always helps to widen that bottle neck).

Really your question does not really mean much. If you think that the XDR interface will be better simply because it is in the PS3 you are misguided it is most likely be because it is cheap and good enough to top Microsoft's 360 CPU....

Faster the better always has been always will (with some exceptions but dont get smart ).

 

[TabrisDarkPeace]

XDR has a lower pin count per equal throughput compared to alternatives.

It is similar to PCI Express in many respects (Read up on why PCI Express is better than PCI, and I am talking about more than just clock speed and bus width here).

It would permit cheap to manufacture memory, with a 256 bit interface (possible / cost-effective on most PCBs) it would provide 'worlds' of throughput.

As for Cell being good gaming, most people already knew that ages ago, whatever does the math (including physics), AI / logic, and moves the data (eg: textures, audio, etc) around the 'best' way will be the 'best' for gaming.

Only problem is coding well for a given platform, and compiling and testing said software (games) once coded. That might be the downfall area.

Lack of decent software studio / development tools / advertising said tools / training & documents of said tools / etc is a very common factor in architectures failing, and often overlooked. (Just look at the entire history of computing).

The next 'major' issue (for x86/x64) is going to be the ever increasing space that advanced Out of Order (OoO) execution units require to keep scaling at the same rate. (Eventually they'll require the same amount of space as L1/L2 caches do now, and then the multi-core designs will not scale so nicely.... unless they trade OoO complexity for sheer number of cores or better performance in other areas. eg: FPU, or number of FPUs)......

..... which is exactly what IBM Cell is doing ..... and to some extent what Intel are doing with IA-64.

Even with constant de shrinks x86/x64 is heading into a brickwall where it can no longer scale performance like it used to..... or rather it will not scale performance anywhere near aswell as other designs..... (Looking at IBM Cell, Sun Microsystems UltraSPARCs, and Intel IA-64 here).

What I want to know is: What do AMD have planned past 2012 ?, as x64 will hit a wall then.