Does Cell Have Any Other Advantages Over XCPU Other Than FLOPS?

[ihamoitc2005]

CELL not just for consumers but also military and research

I dont think anyone is saying that Sony doesnt know how theyre going to use it, only that theyre not sure if it will provide a tangible benefit to the mass market consumer...other than FLOPS.

A Sony exec saying the cell is needed going forward is like bill gates saying vista is needed for PCs to move forward. Don't waste your time looking.

Consumers do not know or care about flops so idea of CELL being high flops for marketing is silly. CELL is high flops because many future applications are high flops and CELL has very high programmable flops density. Applications are many from research to military are very profitable with very high margin. Currently many use custom design chip for these purposes due to low flops density of "off-the-shelf" chips but CELL allows cheap scalable high programmable flops density alternative to custom high flops design.

Good example is expensive chip in F/A-22 jet fighter design by Hughes called CIP. CIP has very poor performance but is very expensive due to custom high programmable flops density design. Main purpose of CIP is scalable high performance multi-purpose flops for current and unknown future flops applications for navigation, targeting etc. CELL is similar but very cheap, low energy, and higher performance.

 

[Vince]

Where is it clear that people are demanding more functionality? What exactly are they demanding from their TVs other than higher resolution? Memory card slots?

Uh, it's clear in their adoption of products of increasing speed, convinence and convergence. EV-DO/3G, DVR's, the Ipod & iTunes, WiMAX, the PSP. People want pervasive computing and this extended functionality which you get from devices which aren't just single function.

You totally miss the point, not surprisingly, and invoke the friends -- who are ultimately the standard and the only necessary bar of what everyone wants -- argument. The point is that by utilizing a processor like Cell you get all this additional ability, utility, as added value, it's basically free and transparent to the end-user. They don't need to care about what type of algorithms are running, but if they can buy a TV that has the flexibility enabled by a Cell processor and get the types of functionality mentioned earlier or later for close to the cost of the competition; that's a strong selling-point.

Look at the PSP, at the additional functionality they keep adding to it... possible because of their inclusion of such flexible processors, like the VME, over standard DSPs.

Of all the things you mention here, features, options, ease of use- NONE of those are only made possible by the use of a cell processor. These things are products of good design first and foremost, and technology second.

Ok, so how many DSPs are necessary to replicate the functionality of a Cell Processor? The POINT is that we're coming to the age in which processors like Cell are going to start replacing all of your DSPs and do it by having more functionality, easier use, and doing it in a single chip solution.

Discrete ASICS have worked, will work, and are still working. Dedicated chips will always be able to do the same required functionality for a lower transistor count, power, and cost. If it makes more sense to Sony to use Cells in everything they make, thats peculiar to Sony becuase of their investment in it - but does not guarentee any competitive advantage whatsoever. Just because its good for Sony doesnt necessarily make it good for the consumer.

Vacuum Tubes worked, will work and are still working. But that doesn't mean you would use them now that you can integrate (Notice that general trend across the entire computing industry for it's entire history? Yeah, it's not going to stop) half a billion transistors on a single die.

And on your comments of "dedicated chips," it's related to my comment to Aaron on implications of NFLT in computing; DSPs are only useful for their specific and static program landscape; they are useless outside of it. NFLT dictates that you look to the niche and back-propogate for the strategy which is optimal if you want to be sucessful. Cell is dynamic and yeilds a balance between the preformance of a DSP and the general computation of a CPU, targeting the digital media computational landscape. They looked at the near future needs and designed an architecture around it: HD stream manipulation; multiple HD stream decoding; HD recording, compression, archiving, searching; DVD archiving and HD upconversion; Video-over-IP to client devices; augumented reality, visual-based GUIs; Multiple Tasking, IM, SMS, video email; Seemless device connectivity; etc.

You will not find another system, especially a static DSP based on which can come close to this type of functionality, across such a broad spectrum, so effortlessly.

 

[nelg]

And on your comments of "dedicated chips," it's related to my comment to Aaron on implications of NFLT in computing; DSPs are only useful for their specific and static program landscape; they are useless outside of it. NFLT dictates that you look to the niche and back-propogate for the strategy which is optimal if you want to be sucessful. Cell is dynamic and yeilds a balance between the preformance of a DSP and the general computation of a CPU, targeting the digital media computational landscape. They looked at the near future needs and designed an architecture around it: HD stream manipulation; multiple HD stream decoding; HD recording, compression, archiving, searching; DVD archiving and HD upconversion; Video-over-IP to client devices; augumented reality, visual-based GUIs; Multiple Tasking, IM, SMS, video email; Seemless device connectivity; etc..

Yes and I look forward to when cell is in my cell phone.

 

[patsu]

I have a few questions as a consumer. Can PS3 replace/interface with my settop box without hardware modifications ?

* Is it possible to record > 1 HD and non-HD channels while watching another "live" one (and pause, rewind, fast-forward like TiVo) ? Will the hard disk be the bottleneck under this scenario ? Is this a solvable problem ?

* On top of it, assuming if cable operators provide flat-rate VoIP (with excellent adaptive acoustic echo cancellation and noise suppression) as planned, what happens if there's an incoming call in the middle of the above scenario ?

* Now with Skype and more VoIP services providing video conferencing, can I have a service that turns my background into blue screen (dynamically) so whoever I'm talking to is not distracted by whatever is happening behind me ? How possible is this ?

Do you consider the above needs "over the top" ?

 

[aaronspink]

Why do I get the feeling this is going to be a typical answer? Secondly, $50? We're talking about next generation products: Blu-Ray, HD-DVD, 1080i/p broadcast quality. Where did the $50, POS chinese knock-off player enter into this argument?

First to <$100 wins. Its simple, and has been repeated throughout time.

The facts are simple. For CE makers like Sony and Samsung who are targeting the lucritive buyers who actually have disposable money to spend (eg. not your shit-ass knock-off players from China), it's clear that people are demanding more functionality, more features and options, and a greater ease of use; and that this needs to be combined with the realization that processign requirements are increasing as HDTV is on the verge of exploding in adoption starting this Christmas and beyond.

You are making a lot of assumptions, care to back them up?

Discrete ASICs are not going to work, they will move into the same niche as the concept of a DSP array for every task. Logic is inexpensive and as capacity increases, computation is decreasing in cost to the point where someone like Sony can replace their discrete DSPs and ASICs with a Cell-esque IC which does all the functions of the DSPs, have added value (eg. look at PSP) and can be updated to support new standards, connect to their network fabric and offer all the little bells and whistles that people who actually spend money demand.

Ask TiVO about people who actually spend money. The simple fact is people want it and want it cheap, which is why TiVO is close to bankrupcy as all the sat and cable vendor integrate a cheap, less functional, bad interface version of TiVO into their basic set top boxes.

An really, for the CE space, people will be using ARMs with their media extensions which are cheap and do everything that will need to be done.

Aaron Spink
speaking for myself inc.

 

[aaronspink]

I have a few questions as a consumer. Can PS3 replace/interface with my settop box without hardware modifications ?

* Is it possible to record > 1 HD and non-HD channels while watching another "live" one (and pause, rewind, fast-forward like TiVo) ? Will the hard disk be the bottleneck under this scenario ? Is this a solvable problem ?

* On top of it, assuming if cable operators provide flat-rate VoIP (with excellent adaptive acoustic echo cancellation and noise suppression) as planned, what happens if there's an incoming call in the middle of the above scenario ?

* Now with Skype and more VoIP services providing video conferencing, can I have a service that turns my background into blue screen (dynamically) so whoever I'm talking to is not distracted by whatever is happening behind me ? How possible is this ?

Do you consider the above needs "over the top" ?

No they aren't over the top, and can and will all be done by asic with a little dedicated logic and an ARM processor. People like vince believe that only CELL can provide some mythical consumer functionality, mean while ARMs are selling into all the CE markets in the billions.

Aaron Spink
speaking for myself inc.

 

[Vince]

People like vince believe that only CELL can provide some mythical consumer functionality, mean while ARMs are selling into all the CE markets in the billions.

I never said that. I never implied that. My position is and has been that as process technology advances the influx of logic will allow for greater computational ability at cost and will result in the assimilation of tasks, which were formerly handled by discrete ASICs, to SoC-esque processors like Cell; giving both a functional and value influx into CE equiptment. How about you read what I actually stated, or just shut up, jackass.

Yes and I look forward to when cell is in my cell phone.

I specifically mentioned CE equiptment. You know, if people wouldn't artificially inflate comments in their mind from what's written to some imaginary view of what's stated, there would be alot less debate.

 

[aaronspink]

I never said that. I never implied that. My position is and has been that as process technology advances the influx of logic will allow for greater computational ability at cost and will result in the assimilation of tasks, which were formerly handled by discrete ASICs, to SoC-esque processors like Cell; giving both a functional and value influx into CE equiptment. How about you read what I actually stated, or just shut up, jackass.

But then, the future is already here. A lot of the ASIC out there are based around an ARM core with associated logic. Things like video scaling and mpeg decode are of course done with custom logic because if is simply more cost effective. These circuits have been designed for years, take a minor amount of area, and are quick and easy to integrate. There is little point having the processor actually do these things fully.

Remember, we are talking about chips that generally will have die sizes less than the die size of the PPE let along a PPE and one or more SPEs. While cell will be put to good use in the PS3, it is vast overkill for pretty much all CE devices. I seriously doubt you'll see sony using cutting edge process technology and mid to moderate die sizes for the CE space when they can utilize paid for process lines and get the same end user experience.

Aaron Spink
speaking for myself inc.

 

[one]

I seriously doubt you'll see sony using cutting edge process technology and mid to moderate die sizes for the CE space when they can utilize paid for process lines and get the same end user experience.

Here's the Sony roadmap for Cell. The PSX DVD recorder, which is a CE product, used 90nm EE+GS in 2003. That Mini-Cell will compete with ARM and likes, and normal Cell will compete with x86, both in the CE space. But who knows, ARM-based Cell is not impossible.

 

[Kryton]

Here's the Sony roadmap for Cell. The PSX DVD recorder, which is a CE product, used 90nm EE+GS in 2003. That Mini-Cell will compete with ARM and likes, and normal Cell will compete with x86, both in the CE space. But who knows, ARM-based Cell is not impossible.

Ughhh... yes it probably is impossible. ARM merely produce processor designs which others can stick into whatever. This 'sticking into whatever' means they have a dedicated bus design, interfaces, instruction sets and other specific things that completely go against the grain of EIB and PowerPC. There is 0 chance of ARM being put onto a Cell. Perhaps there will be someone who develops a bridge for the buses to allow the ARM to use Cell (or maybe a single SPU?) as a co-processor (much in the way it currently does for FP; but ARM FP sucks performance wise) but that is still highly unlikely.

 

[mckmas8808]

Wow! Never thought I'd see such a big fight about if CELL can actually improve CE products. I mean aaronspink I respect that you think a mini-Cell is overkill, but do you honestly think that there will be NO improvements on using a one SPE CELL chip?

 

[Kryton]

Wow! Never thought I'd see such a big fight about if CELL can actually improve CE products. I mean aaronspink I respect that you think a mini-Cell is overkill, but do you honestly think that there will be NO improvements on using a one SPE CELL chip?

The whole point aaronspink is arguing is that this is ARMs space. Perhaps for Sony/Toshiba it will reduce the need to license their designs and technology but for everyone else it will be a case of ARM vs. mini-Cell and who is the best to license.

ARM have extremely efficent chips and extremely wide market support (think: any mobile phone, set-top box, PDA, DVD player, etc.). Because of this a lot of design work has already been done to interface with them and connect as co-processors, people won't reinvent the wheel for a mini-Cell they will stick to existing proven chips with cheap commodity extensions that are very prolific.

Cell SPUs seem to be well designed and I support the concept of a gaming/DSP processor based on multiples of them but no way are they going to undermine ARM. They follow almost identical design goals (except ARM is aimed at high MIPS/Watt rather than FLOPS/Watt) and are both RISC. I liken it to trying to x86-64, it won't be a radical transition because of software - Intel or ARM in this case has the hardware architecture everywhere, they (inc. AMD) can only extend in a compatiable way and not produce some readically new architecture as all the previous software will just break.

Edit: I forgot to mention this BUT... A lot of the hardware now is using generic FPGA chips which have become advanced enough to perform the computations for movie and CE purposes. This means you sidestep the programming issue as the FPGA can easily be reflashed with a new program that adds support. We have got to the stage now where producing these generic FPGAs and programming is cheaper than creating a chip as mask costs have gone through the roof (the downside to exponential speedup is that masks and initial fabrication investments have to be far greater) so it is no longer efficent to design a chip that can do specific H264 decoding - it is cheaper to buy say an Altera chip and their decoding libraries and flash them onto it. Need new codec support? No problem, distribute a new firmware et voila MPEG5 support!

Altera and others are gearing themselves specifically at doing this and are investing heavily in getting software engineers to develop the required Verilog etc. to perform this decoding so they can make such futureproofed appliances. They can and are doing this because they can outprice any custom designed chips because they have the customers and ability to produce millions of 'generic' chips rather than ASICs.

 

[Shifty Geezer]

Edit: I forgot to mention this BUT... A lot of the hardware now is using generic FPGA chips which have become advanced enough to perform the computations for movie and CE purposes. This means you sidestep the programming issue as the FPGA can easily be reflashed with a new program that adds support. We have got to the stage now where producing these generic FPGAs and programming is cheaper than creating a chip as mask costs have gone through the roof (the downside to exponential speedup is that masks and initial fabrication investments have to be far greater) so it is no longer efficent to design a chip that can do specific H264 decoding - it is cheaper to buy say an Altera chip and their decoding libraries and flash them onto it. Need new codec support? No problem, distribute a new firmware et voila MPEG5 support!

Altera and others are gearing themselves specifically at doing this and are investing heavily in getting software engineers to develop the required Verilog etc. to perform this decoding so they can make such futureproofed appliances. They can and are doing this because they can outprice any custom designed chips because they have the customers and ability to produce millions of 'generic' chips rather than ASICs.

Which is also the approach Cell offers. A standard hardware that can be applied with stock software and customizability. As I hear it, the difference between flat-panel TV quality is as much in the software as the hardware. Different manufacturers wanting different algorithms are either going to need customs ASICs, FOGA's they can 'program' with their algorithms, or a processor that can run any software algorithm. As the software side becomes increasingly complex, surely the cost of ASIC's will increase dramatically, whereas a standard hardware like Cell will always be dropping in price and yet be both powerful enough for any tasks for the forseeable future, plus be scalable so the software and tools can be developed and ported to future products, plus existing products updated. That's why you don't see many standalone wordprocessors these days, whena PC can do the job of a word-processor and can be upgraded. If Sony and Toshiba can release TVs that can be improved over time to support new image processing techniques and codecs, for example, they'll be offering something other solutions can't.

 

[mckmas8808]

Kryton I totally understand your last post, but what I'm thinking and what I think Vince is trying to say is CELL can bring something different to the CE table that is worth something to the consumer. Years ago people started buying HDTVs even when the programming wasn't there to warrant it. Even today I don't think the programming is good enough for the masses, but you have to start somewhere.

If CELL in HDTV's can allow certain extra features that Vince was talking about then I'm all for it and so should all of you guys. This is a technical forum right?

 

[Kryton]

Which is also the approach Cell offers. A standard hardware that can be applied with stock software and customizability.

Which is what FPGAs are, stock hardware which you stick stock software on.

As I hear it, the difference between flat-panel TV quality is as much in the software as the hardware. Different manufacturers wanting different algorithms are either going to need customs ASICs, FOGA's they can 'program' with their algorithms, or a processor that can run any software algorithm.

A HDTV decoding algorithm is a HDTV algorithm... You have one algorithm you must implement - don't and you won't get proper results! Customisation as you would state are 'image enhancements' (making data up that isn't there but everyone suckers for it) and the FPGA's allow you to do this. See the Altera website and click on Consumers to see the things they provide as an example.

As the software side becomes increasingly complex, surely the cost of ASIC's will increase dramatically, whereas a standard hardware like Cell will always be dropping in price and yet be both powerful enough for any tasks for the forseeable future, plus be scalable so the software and tools can be developed and ported to future products, plus existing products updated.

Erm, I think you have things confused: an FPGA is a generic programmable chip which you can stick 'software' on (look up Verilog HDL or VHDL) - you program the hardware by altering gates etc. (this is hard to explain). An ASIC is an Application Specific IC and that is what I pointed out: there cost is going up rapidly as complexity increases so you end up with more complex mask-sets and high inital mask creation cost. FPGAs do not suffer this as the masks are made once and you make them do what you want after they are made. FPGA: Make one chip, program how you want. ASIC: Design, make and don't ever alter/program.

That's why you don't see many standalone wordprocessors these days, whena PC can do the job of a word-processor and can be upgraded.

Exactly.

If Sony and Toshiba can release TVs that can be improved over time to support new image processing techniques and codecs, for example, they'll be offering something other solutions can't.

FPGAs already allow this, that is my entire point.

If CELL in HDTV's can allow certain extra features that Vince was talking about then I'm all for it and so should all of you guys. This is a technical forum right?

FPGAs let you do this. As an example (you guys will start to think I work for Altera soon) take the Altera DVD chip solution - it has the necessary decoding programming provided and the ability to do other things such as OSD and 'image enhancements'. It is up to you as a H/W manufacturer to implement these additional features if you so wish, you still get a chip that with the basic package can play DVDs but you can write more into the basic package to do MP3 playback or WMV playback and put that in your player against your rivals who didn't even though you use the same chip. It's far cheaper to simply write those extras that make your product than have to do the entire thing (chip mask, DVD decoding on the chip, testing, fabrication for probably a 'small' run of say 1 million units for one model)

I don't see why the comment about this being a technical forum is really relevant. Technology is great that is why we are here but lets accept that economics also rules technologies applications and that is a topic of discussion that must be considered when we start to try and apply technology to things.

 

[seismologist]

I didn't check Altera's website but are they really making 3.2ghz FPGA's that are cheap enough to go into consumer products? I wouldn't think that FPGA's are relevant here...

 

[Kryton]

I didn't check Altera's website but are they really making 3.2ghz FPGA's that are cheap enough to go into consumer products? I wouldn't think that FPGA's are relevant here...

No they don't make 3.2Ghz FPGAs. The whole point of an FPGA is you don't require this power - name ANY consumer appliance where you need it as well? No one makes 3.2Ghz chips for consumer appliances. The reason you don't is because you don't need the power because the chips can be turned from general-purpose (read Cell SPUs) to specific purpose very easily. We have already seen how Cell is amazing at floating points because it was targetted at performance in that area, similar to how FPGAs can have amazing performance in any target area because of domain-specific knowledge application.

They are relevant because every argument that has been mentioned for Cell being 'revolutionary' in this area has already been answered by FPGAs.

 

[Shifty Geezer]

Which is what FPGAs are, stock hardware which you stick stock software on.

I don't know quite you're arguing because I've agreed with you. Cell works like FPGAs. The big difference that I see is that Cell will be, in theory, far more wide-reaching. If someone comes up with a new fancy effect, it can be used in software and programmed using a wealth of experience. That is, if someone writes a Photoshop .8bf plugin filter service for Cell as part of a graphics package, this can be added to the software of the Cell TV. Of course one could design an FGPA to do the same thing, if they allow such complex pathways, but the likelihood of someone creating such an FPGA is low. Whereas the development of software algorithms will be high. When it comes to versatility, a standard CPU (like there's anything standard about Cell!) is going to win every time, surely. After all, FPGA's haven't found their way into computers. We don't have an FPGA that switches between audio decoding and graphics rasterization. The only FPGA I know used in a computing device was in the Amiga CD32 console.

Obviously I'm no expect on FPGAs, but it seems to me that these chips offer good value and flexibilty up to a certain complexity of program, but beyond that you hit limits, and I guess they lack the developer tools of traditional language based programming. How easy would it be for me to write a 3x3 convolution matrix given knowledge of how to do this in C++? Are my existing skills portable? What's the performance limit of FPGA's at the moment and how is that likely to progress?

 

[expletive]

Uh, it's clear in their adoption of products of increasing speed, convinence and convergence. EV-DO/3G, DVR's, the Ipod & iTunes, WiMAX, the PSP. People want pervasive computing and this extended functionality which you get from devices which aren't just single function.

And none of these things were made possible becuase of a revolutionary processor, they exist becuase someone had the vision to make them interoperable. None of these examples (granted i dont know what EV-DO is) imo, are examples of how something like the cell processor is necessary.

You totally miss the point, not surprisingly,

Your tone here is condescending and unnecessary. How about you treat me with the same respect ive afforded you in my comments and not resort to this simply becuase i'm challening your arguments.

and invoke the friends -- who are ultimately the standard and the only necessary bar of what everyone wants -- argument. The point is that by utilizing a processor like Cell you get all this additional ability, utility, as added value, it's basically free and transparent to the end-user. They don't need to care about what type of algorithms are running, but if they can buy a TV that has the flexibility enabled by a Cell processor and get the types of functionality mentioned earlier or later for close to the cost of the competition; that's a strong selling-point.

If you think its any different for the mass market consumer than what we see from our friends and family then by all means provide a response to the contrary. The bottom line, imo, is that companies are not going to increase the BOM for features that are "transparent" to the end user. You talk about this flexibility afforded by the cell yet i havent read anything that proves it is something that cant be produced by smaller, cheaper, dedicated designs. If a manufacturer decides certain features are necessary theyll have plenty of more cost-effective options besides a cell processor to meet those requirements.

Look at the PSP, at the additional functionality they keep adding to it... possible because of their inclusion of such flexible processors, like the VME, over standard DSPs.

You dont know that the upgrades we're seeing now were ALWAYS part of the desgin spec of the PSP. If it was then they included a chip that could meet those requirements. If a manufacturer decides that they want a mobile computing device that requires an upgrade path theyll include the ASICs that provide that fliexibility. That doesnt mean you need a Cell proessor.

Ok, so how many DSPs are necessary to replicate the functionality of a Cell Processor? The POINT is that we're coming to the age in which processors like Cell are going to start replacing all of your DSPs and do it by having more functionality, easier use, and doing it in a single chip solution.

Does it really matter? A manufacturer will figure out what a product needs to do and theyll put together a BOM that can meet those requirements. You assume that the there is functionality out there that can ONLY be accomplished by a cell processor. This leads to your question how many DSPs does it cost to replace a cell, the real question we need to ask is, do we NEED to even come close to approximating what a cell processor can do?

Vacuum Tubes worked, will work and are still working. But that doesn't mean you would use them now that you can integrate (Notice that general trend across the entire computing industry for it's entire history? Yeah, it's not going to stop) half a billion transistors on a single die.

If vacuum tubes were smaller, cheaper, used less power, and could perform the same function as a solid state counterpart you can be assured we would still be using them.

And on your comments of "dedicated chips," it's related to my comment to Aaron on implications of NFLT in computing; DSPs are only useful for their specific and static program landscape; they are useless outside of it. NFLT dictates that you look to the niche and back-propogate for the strategy which is optimal if you want to be sucessful. Cell is dynamic and yeilds a balance between the preformance of a DSP and the general computation of a CPU, targeting the digital media computational landscape. They looked at the near future needs and designed an architecture around it: HD stream manipulation; multiple HD stream decoding; HD recording, compression, archiving, searching; DVD archiving and HD upconversion; Video-over-IP to client devices; augumented reality, visual-based GUIs; Multiple Tasking, IM, SMS, video email; Seemless device connectivity; etc.

You will not find another system, especially a static DSP based on which can come close to this type of functionality, across such a broad spectrum, so effortlessly.

Ok so first off, the kind of HD video processing you talk about here is far from 'effortless', especially if you want to make an argument that the cell will provide image quality/functionality an order of magnitude better than what we see today.. Just becuase you have a processor with power to spare does not mean by any stretch that the work is close to done. Ask the best in the business of video processing: Realta, Gennum, DVDO, Lumagen, Terranex, what % of the work is in the hardware and what % of the work AND the performance is the SOFTWARE. I'll tell you right now, more than half the work is in software. In terms of video processing features and power, the cell is a smaller part of the equation than you think.

DSPs being only useful in their specific function is fine. A product is designed to do X. You put the chips in it to do X. If youre saying hteres a model where TVs and camcorders are going to be upgraded in the future to add functionality that wasnt part of its original design spec, then make that argument as to why Sony isnt going to try and sell you an upgrade instead.

 

[mckmas8808]

No they don't make 3.2Ghz FPGAs. The whole point of an FPGA is you don't require this power - name ANY consumer appliance where you need it as well? No one makes 3.2Ghz chips for consumer appliances. The reason you don't is because you don't need the power because the chips can be turned from general-purpose (read Cell SPUs) to specific purpose very easily. We have already seen how Cell is amazing at floating points because it was targetted at performance in that area, similar to how FPGAs can have amazing performance in any target area because of domain-specific knowledge application.

They are relevant because every argument that has been mentioned for Cell being 'revolutionary' in this area has already been answered by FPGAs.

I see your point and I pretty much agree. But where I differ from you is that I believe it could be possible that a mini-Cell could be benefit companies that use it in their CE products. I'm in no way discrediting FPGAs.