Why would IBM cancel development on Cell CPU's anyway? Last time I checked IBM provide processors for multiple tasks and last time I checked there is some jobs the Cell is fantastic at, well ahead of other CPU types. I personally happen to think there is no way in hell we will see the PS4 sporting one but I think there is a lot of room in the video processing, server markets for the Cell among others.

I don't understand the aguments of the Cell being a stream processor or not, just reading your replies guys it seems quite clear technically it can be a stream processor, it's just not that effective as other solutions. Is this any surprise that one processor is less effective than another that is geared specifically for the task. I'll not get involved as I don't fancy joining you all in googling links to one up each other.

exactly...the classification of what a processor is is made by the team who design's it and create's it, it's not that hard to understand.

as for the PS4 sporting the Cell who know's? except Sony at this point their not saying, but If anything it should be fun learning what Sony does use in their next playstation.

To Clarify, the Cell's SPUs can act as very powerful and flexible stream processors. This due to very high internal bandwidth and very high speed dedicated memory available to each SPU.

For example when anti-aliasing is done on the SPU or when processing 7.1 surround audio, small chuncks of data is processed at very high speed, just a small part at a time. Regarding graphics and audio processing the Cell is most excellent for stream processing.

Executable code is usually of course much smaller, each SPE can act as independently operating systems on a chip as well. The Cell is very flexible and powerful with regard to what they can do with SPU optimised functions. It's more powerful than top PC processors and far more flexible than GPUs.

I think the point people were arguing over MikeB is that current top GPU's on the market are built with stream processing as a primary function and hence are leagues ahead for the task.  This isn't a slight against the Cell in the PS3 though as it has to be versatile in order to be flexible as you say.  Graphics wise it would be more efficient to not use the Cell if you had the dedicated GPU hardware availble, as it often isn't on the PS3 the Cell is brilliant at improvising any shortfall while still having plenty of power to spare.  The Cell can never fully take over the functions of a GPU until the number of SPU's are increased dramatically which unfortunately creates hideous bottlenecks elsewhere surely?

The Cell is only more powerful than top PC processors for certain tasks as you well know, there is a reason it hasn't made any progress in the desktop market.  In bespoke hardware its a wonderful solution but the PC market is embedded with way too much legacy software for the Cell to ever become a viable alternative.  The Cell is just a different type of CPU that is unbeatable when used for the right applications.

@ Slowmo

The Cell is good at performing any kind of CPU functionality, as long as it's the code is well structured for the Cell. Written in very clean ways. Basically anything that runs on the PPE (PowerPC) can run on the Cell's SPUs if properly rewritten. The US Air Force Research Laboratory gives some insight why the are using PS3s:

"Before it won the research award in 2008, the information directorate's advanced computing architectures team considered alternative configurations and the possibility of a hybrid system, but found multicore Xeon servers slower and more expensive than PS3s, and GPGPUs to be slower in some important types of calculations. "

http://www.informationweek.com/news/software/linux/showArticle.jhtml?articleID=221900487

It was clear that too much redesign would need to be performed with regard to legacy game engines to make optimal use of the Cell's SPUs to fully take over all the dedicated tasks of a GPU. So that's why Sony in 2002 partnered with NVidia to design the RSX GPU for the PS3, instead of using a multi-Cell design. That's a year before Microsoft partnered with ATi for the 360 GPU, despite that console releasing a year earlier than the PS3.

AA is not a high compute, low bandwidth task, where cell excels. Quite contrary, AA performance is directly related to chip-memory bandwidth. That's why Xenon has it's high bandwidth eDRAM chip, and higher number of AA samples in most of multiplatforms. As always, you don't know what are you talking about. The problem is that a lot of people believe the things you say, because it suits what they want to believe.

@ Kynes

Multi-platform titles are not fully or not equally optimised for all supported game systems. For this the XBox 360 and PS3 are far too different.

Top PS3 exclusives usually have better anti-aliasing than XBox 360 exclusives.

A good example would be comparing top high profile, top high budget exclusives like Halo 3 and Uncharted 2. Uncharted 2 has anti-aliasing and Halo 3 does not, in addition Uncharted 2 provides far more detailed and varied environments, steadier framerate, better lighting, better physics, better audio and overall more seamless experience.

And of course Halo 3 isn't even rendering in high definition!

The EDRam on the Xenos daughter chip could have been an important advantage equalling things out in terms of graphics potential to some extend (not with regard to game complexity though and Blu-Ray/default harddrive would still be a PS3 advantage with regard to streaming). But sadly the amount of EDRam is too small for high definition gaming to prevent technical sacrifices due to tiling issues.

Please explain how cell is good doing AA resolve, with technical reasons, not with void words.

In the Halo3-Uncharted2 comparison, you're comparing a 2 years old game, with a rendering method that can't use AA, to a new game. You can't do AA with the HDR high exposure-low exposure lightning method used in Halo3, and using Halo3 to compare AA methods is a crappy way to try to say PS3 has an advantage doing AA. Halo 3 isn't rendering in 720p because of the same reason, the two exposure buffer method used. It's not a console limitation, it's an engine limitation. As always, trying to mislead.

Please don't derail the question and answer exactly to what is asked. Why do you say Cell is good doing AA resolve?

Maybe using a bigger, bolded font you see the question asked.

Flexibility, there are different methods and approaches for doing AA (and the need for this also depends on rendering resolution, art direction and other lighting/color effects as well).

I better link to another discussion, so you won't be upset with me.

http://forum.beyond3d.com/showthread.php?p=1360606

Do you know what I confirm thanks to that thread? That you don't understand what they are talking there. Please read it again and tell me where they say a SPU is better than dedicated hardware doing AA resolve.

Flexibility doesn't equals to high performance. One of the limitations of RSX is that nVidia (unlike ATI since DX9 cards) had fixed AA resolve patterns in it's chips, so you can't create new AA methods/patterns. Doing AA resolve in a SPU is slower than doing it in the dedicated hardware of RSX. The only advantage is that you can chose better patterns, higher number of samples... in non performance critical situations. But AA in SPU is much slower.