The proof is in the results. If the Cell was all that it would be used in considerably more applications than it currently is. In single threaded performance and efficiency the X86 Intel CPU is king, in price/performance applications ARM is the one to beat and for high performance computing there isn't much out there which can hold a candle to GPGPU/Xenon Phi architectures.

The majority of the major optimisations developers have made are for graphics processing which is irrelevant if you attach a very strong GCN AMD GPU to the console. Once you've done that then you're going to throw away the majority of your legacy code as it simply isn't needed anymore.

Cell is only better at certain things tho, sure it runs circles arround a traditional CPU in terms of raw flops but that doesn't really reflect all workloads. No one has really prioratised SIMD performance on x86 in a long time unless you count larrabee/Xenon Phi which were never meant to be used as a CPU successor but rather GPU/coprocessor to be used in conjunction with a traditional CPU.

The CELL's asymetric CPU design seems to be dead, tho I suppose including a GPU on the same die as the CPU could be seen as a successor to that design.  I don't see anyone using 2 different types of CPU cores in the same die tho, that design did seem to be a dead end or at least one that IBM didn't think was worth pursuing.

So far in terms of heterogeneous compute GPU bassed design does seem to be winning the war but they are still a very far away from the point where you could do away with traditional CPU cores in a console let alone a PC. CPUs aren't replacing dedicated GPU any time soon ether, Intel and AMD have instead gone for the aproach of just including a GPU on the same die (with the option of an addon card) rather than try to do graphics on CPU cores. Maybe in 10 years we will see a new architecture that really does offer the ultimate balance of flexibility and performance that lets it supersede the current CPU+GPU paradime for games, but not this generation. 

If proof is in the results, why isn't the WiiU an X86 item? Why do you bring in Xenon Phi (>$1500 a piece) or high-end GPUs (>$150 currently) stuff into the discussion? None of that you will ever see in a console. A console has a $199 target, so discussing $2000 accelerators and $500 graphic cards is a moot point, consoles are never going to compete on high-end PC graphic levels. The simple fact is that a console maker will always target the $25 CPU and $25 GPU. A 3PPU/6SPU enhanced 32nm cell would have fit the bill in all respects for PS4, and would have had relatively low R&D costs attached to it. It is completely irrelevant whether the whole world embraces cell or not (again, see WiiU), the only relevance is that it fits the needs of a PS4.

Your point of (ab)using the SPUs for graphics processing is of course right, but as things stand now, both MS and Sony will discard 99% of their legacy code anyways, as they go AMD if latest rumours are to be believed.

1. Learning to ride a bike is hard, but once you do, it's an efficent form of transportation. My point still stands that any difficulty in 'learning' or 'using' cell is alrady mitigated by the 8 years of experience that developers have.

2. 'High end engines' don't get more powerful, they just demand more resources and processing power. Since you can't upgrade consoles with more memory or better GPU, your 'next gen x86' Engines will just get out-dated after a year or two and you have no way to use the newer ones. You should not lump PC life-cycles with Console ones. 

3. I don't know where you're getting that information: http://en.wikipedia.org/wiki/Advanced_Micro_Devices

"AMD also formed a strategic partnership with IBM, under which AMD gained silicon on insulator (SOI) manufacturing technology, and detailed advice on 90 nm implementation. AMD announced that the partnership would extend to 2011 for 32 nm and 22 nm fabrication-related technologies."

22nm is next step on ITRS

4. Theres no indication that R&D for an updated Cell woudl cost Sony '1.5b' if that is what you're implying, considering IBM already made an updated one just 4 years ago.

Engines get more complex as the hadware and games get more complex, building an engine that offers all the features that developers need and that take advantage of all the hardware features will be far more complex to build next gen. Which is why most next gen engines are now being built on high end PC because these engines now take multiple years to create by themselves. All engines get constantly upgraded but the core platform will be ported accoss, UE4, Cryengine 3, Luminous, Forstbyte 2 etc etc is going to be what next gen games are built on, and they will be optomised for whatever hardware is in the final hardware. I don't see what changing PC hardware has to do with being able to port across the engines that everyone is designing today to get ready for next gen console hardware. 

AMD APUs are all moving to 28nm next year, they are sticking to 32nm on high end FX CPUs which is not what will be in PS4. 

I'm not saying it would cost them $1.5b I am saying that it won't be as cheap as you seem to think it would be. They are not going to use a server CPU in the PS4.

You know, I think some of you guys are making this too complicated. If you simply look at it from a business perspective, then you'd understand that there is no reason to revisit the Cell and spend as much on R&D since you can already get something that's proven to work rather than having to go back and tweak all the internal workings of the Cell again. You'd need to make quiet a few changes in the Cell to make it relevant for modern GPUs, why bother with that when somebody else already has a solution? It's not like Sony has that much money ya know?

The better things are the things that slow processors down. I mean, not like typing a word document is gonna kill your processor or browsing a webpage.  Only thing it's more poor in that would matter is A.I.  but they had to sacrifice branching to get the parallel processing working.  Asymetric design will never be dead, GPUs are an example of this (Not even talking about APUs).  And like I said GPUs and CPUs are converging, never said it would be anytime soon.

Why isn't the Wii U X86? Quite likely because they want backwards compatibility with the Wii.

Why did I bring up other architectures? As an example of course. What's the point in including a CPU which is obsoleted by the GPU? The majority of effort this generation has been to use the Cell processor for graphics rendering because graphics rendering requires the fewest lines of code and that was required in order to bring the console versions to parity with the Xbox 360. If you have a next generation GPU you don't need the Cell as it is far more efficient/effective as a GPU than the Cell processor is and the Cell processor is utter crap for general purpose loads which would be better suited to run on an X86 CPU OR power PC architecture.

All of the C coded for the engines used today will recompile just fine for X86 because afterall developers have been supporting the PC throughout this entire generation of consoles so I can't anticipate them having problems.

I tried to make this point earlier. A custom designed Cell processor that developers dislike and will cost Sony lots of money to research versus a cheaper, slightly modified off the shelf CPU/APU + GPU that virtually all developers understand, has numerous tools ready to use and will do a comparable job.

Unless they plan to use a version of the Cell in all their hardware (unlikely considering the original Cell only made it into the PS3), then from a business perspective, the Cell processor is redundant.

Source?

My computer knowledge of GPUs and rendering tell me that's bull.