Read please, an interesting mail that about N5 and Hitachi

[Paul]

How can you say that when there is no hardware back from the fabs yet. Its still in patent form . Where as

How can I say that IBM is shooting for 1TFLOPS performance and they have been working on Cell for years now? Wha....?

Is already showing some proof of concept. the equivalent of 4 3ghz p4s . The chip was running at 100mhz on .18u . This was done a year ago before 3ghz p4s were around and on less advance microns . .18-.13 . Not only that but much less than the .65 cell is going to use .

Oh and what was the floating point performance on this thing? Exactly. And is it practical to put inside a closed box such as a game console? Or are they cooling this thing with some type of insane cooling solution.

This chip has the best chance of becoming the first TFLOPS consumer chip? Nah, it doesn't.

Your still trying to dodge my issue, if NEC and Hitachi started yesterday could they make a 1tflops microprocessor to beat what Cell is supposidly going to do? This was my point the whole time, you not even adressing it.

 

[jvd]

How can you say that when there is no hardware back from the fabs yet. Its still in patent form . Where as

How can I say that IBM is shooting for 1TFLOPS performance and they have been working on Cell for years now? Wha....?

Is already showing some proof of concept. the equivalent of 4 3ghz p4s . The chip was running at 100mhz on .18u . This was done a year ago before 3ghz p4s were around and on less advance microns . .18-.13 . Not only that but much less than the .65 cell is going to use .

Oh and what was the floating point performance on this thing? Exactly. And is it practical to put inside a closed box such as a game console? Or are they cooling this thing with some type of insane cooling solution.

This chip has the best chance of becoming the first TFLOPS consumer chip? Nah, it doesn't.

Your still trying to dodge my issue, if NEC and Hitachi started yesterday could they make a 1tflops microprocessor to beat what Cell is supposidly going to do? This was my point the whole time, you not even adressing it.

Paul this is going to be my last post to you. Its really not worth the time.

If nec and hitachi stared yesterday then I don't know. But i do know that nec and hitachi did not start yesterday. I also know the cell chip was not started a few years ago. It was all of the work ibm has ever done plus years more of on going r&d to make this chip. The billions the cell crew dumped into it was for fabs that didn't exist. To liscense the tech from ibm including the fab tech. Then all the money to make sure there are no bugs in the chips , the yields are good , the masks , the first samples , each die respin. So till you actually know what your talking about . Can show me any proof of concept of this chip. And more than just a patent on it . I will refuse to believe that the cell chip will be a) ready on time ,B) hit 1 tflop in the ps3 system C) Be the first single chip set up to hit 1 tflop.

Till then its all just sony hype . Like when the first ps2 numbers came out and it was said to do 80million polygons. Yes sony never said 80 million. But the fan boys got a leaked number and ran with it. Same thing you are doing now.

Btw why didn't you answer my other post ?

 

[Paul]

If nec and hitachi stared yesterday then I don't know. But i do know that nec and hitachi did not start yesterday.

Umm this was my point since the begining, IF THEY started yesterday. With nothing. Why you made it into more than this I don't know.

So till you actually know what your talking about . Can show me any proof of concept of this chip. And more than just a patent on it .

What in the HELL are you talking about? I ALREADY SAID that 1TFLOPS Cell will be a MIRACLE, and I NEVER SAID that Cell would be 1TFLOPS in this post. EVER. End of story, your WRONG. I said they were AIMING for a tflops chip.

I will refuse to believe that the cell chip will be a) ready on time ,B) hit 1 tflop in the ps3 system C) Be the first single chip set up to hit 1 tflop.

I already said that this is what STI wants Cell to do, whether they REACH IT or not is another matter. But THIS IS WHAT they are aiming for. I never said any of this, what's more you can't get a direct quote from me in this topic.

Like when the first ps2 numbers came out and it was said to do 80million polygons. Yes sony never said 80 million.

They clearly said 75 million polygons RAW.

 

[Grall]

How can you say that when there is no hardware back from the fabs yet. Its still in patent form .

Cell's undoubtedly in the layout stage by now working its way towards tapeout. It's far further along than "still in patent form". Remember, they will start to fab this thing in 2H 04...

Is already showing some proof of concept. the equivalent of 4 3ghz p4s . The chip was running at 100mhz on .18u .

Isn't it that massive-array-of-8-bit-CPUs-thingy you're talking about here? If so, Cell would crush it to a pulp using one of its virtual little fingers. That array processor is AFAIK an integer only device so it would be no good for 3D processing, physics, AI etc, and with only 8 bits of precision it's more of a gimmick than anything else. An 8-bit CPU = crap today, and a massive array of 8-bit CPUs = a massive amount of crap.

Maybe you're talking about something else than me, but if I'm right that thing is no threat to anything. Least of all an 128-bit Cell processor.


*G*

 

[PC-Engine]

It's a proof of concept as opposed to a patent

Upgrading the 128 8-bit SIMD units to 32 or 64 bits isn't difficult. Sure more transistors are required to do that but that's what sub 0.10u fabs are for. The prototype was manufactured on 0.18u and ONLY ran at 100 MHz

How powerful would it be if it was clocked at 1-2GHz with 128 64-bit SIMD units fabbed on a 0.09u process?? Pretty damn powerful.

Why people assume STI has some sort of magical technology that allows them to pack in more logic into CELL than other companies is beyond me. These same people think that FPUs was invented by STI

See the difference here is that STI wants to fab a HUGE chip that INCLUDES eDRAM. NEC can also fab a HUGE chip too and they don't even need eDRAM, but why would they??? The GPU will be able to do loads of GFLOPS anyway

The question isn't whether they can or cannot, rather it's whether or not they NEED to.

BTW when the EE was available, was it the most powerful CPU on the market? Um...NO...

So who had the most powerful CPU? Hitachi and NEC at 8 GFLOPS each.

 

[Grall]

It's a proof of concept as opposed to a patent

...Meaning what, exactly? Are you doubting Cell will ever appear on the market, despite Sony stating it's on schedule and things are going well with the development, or something?

I don't see your point. Volkswagen has been working on the Veyron 16-4 supercar for years, some said it was vaoprware, yet it made a public appearance just a few weeks ago. Sony will intro Cell one day too. Patience, my young padawan. Patience.

Upgrading the 128 8-bit SIMD units to 32 or 64 bits isn't difficult.

...Except they'd still be integer units and unsuitable for what's needed in a modern console. Guess you could run one mean version of duke nukem 3D on that array, but nextgen gaming will not be its forte. Re-jigging it to do flops will require re-working the thing from the ground up as ALUs are totally different from FPUs from an architectural standpoint.

How powerful would it be if it was clocked at 1-2GHz with 128 64-bit SIMD units fabbed on a 0.09u process?? Pretty damn powerful.

Except it would have to be re-jigged AGAIN to run at speeds 10-20x faster than the first implementation, so your "proof of concept" wouldn't be ANYTHING like a product that could actually be used in a game console. It would do flops instead of ops, it would be 64 or 128-bit instead of 8, be produced on sub-.10u instead of .18 and run at possibly 2GHz instead of 100MHz.

Nothing much in common between the two is there?

Why people assume STI has some sort of magical technology that allows them to pack in more logic into CELL than other companies is beyond me.

Which people assume that? I haven't seen anyone except Deadmeat championing this theory - and now you. Deadmeat does his famous 'pick apart one thing to predict how another entirely different thing will look on the inside' analysises (hm), and comes up with the idea Cell is impossible, yet still we have three multinational conglomerate corporations working on such a thing! Aren't THEY stupid or what!

One doesn't HAVE to believe STI to be able to perform miracles. All that's required is taking a wait-and-see approach. They ARE developing some kind of chip, exactly what it is we don't know yet. No need to get one's panties in a bind, rip EE apart and use it as a basis for an illogical argument that Cell is impossible and then accuse people of thinking STI can do things others cannot.

Just wait and see dammit!

These same people think that FPUs was invented by STI

Ok, now you're trolling. Please stop making stuff up, thank you.

See the difference here is that STI wants to fab a HUGE chip that INCLUDES eDRAM. NEC can also fab a HUGE chip too and they don't even need eDRAM

The eDRAM is there for a reason. Remember when Faf said GS renders an entire scene in the time it took a GF2 GTS to finish the shadow volumes? The GF2 beats GS on raw fillrate, yet eDRAM gives a massive real-world performance advantage to GS.

It's one thing having 128 SIMD thingies in parallel on a chip, and another thing entirely to keep all these processors fed with data. How do you propose to keep 128 64 or 128-bit processors running at 2GHz using only off-die memory? You'd need literally hundreds of GB/s worth of very low-latency bandwidth, and such a memory sub-system would be very VERY expensive. That's what eDRAM (and the scratchpad RAM in each APU AND the big register files) is for in Cell, spread the bandwidth load around to prevent bottlenecking.

The GPU will be able to do loads of GFLOPS anyway

You seem to be slightly short-sighted here. What if you want to do complex AI, wind simulation, cloth physics, inverse kinematics etc etc? A phat GPU will be NO HELP WHATSOEVER there. All a GPU is good for is eye candy, but the environment will be dead just like in current games if there isn't a strong CPU behind it to drive the universe the game's located in.

BTW when the EE was available, was it the most powerful CPU on the market? Um...NO...

Was it the most powerful CPU in its price class? Um...YES...

So who had the most powerful CPU? Hitachi and NEC at 8 GFLOPS each.

How many thousand $ apiece did they cost?


*G*

 

[jvd]

...Meaning what, exactly? Are you doubting Cell will ever appear on the market, despite Sony stating it's on schedule and things are going well with the development, or something?

I don't see your point. Volkswagen has been working on the Veyron 16-4 supercar for years, some said it was vaoprware, yet it made a public appearance just a few weeks ago. Sony will intro Cell one day too. Patience, my young padawan. Patience.

Two points , a chip in hardware and a chip in patent form are two diffrent things. Even if things are going grat on the other there is no hard proof of how it performs or if its up to snuff.


Second nivida used to say the nv30 was on schedule and things were going well. 5800 ultra ? Horrible horrible card.

And for years it was vaporware . There are cars that drive themselves in the test labs. Do you consider them vaporware ? I do. Till they are announced and shown they are vaporware. The sledgehammers were vaporware till they were demo'd , P4 5ghz are vaporware. I don't doubt though that one day it will come out. Right now ps3 , xbox 2 and gamecube 2 are all vaporware . One day they will come out. But till they are demo'd they are still vaporware .

 

[Deepak]

Two points , a chip in hardware and a chip in patent form are two diffrent things. Even if things are going grat on the other there is no hard proof of how it performs or if its up to snuff.

Second nivida used to say the nv30 was on schedule and things were going well. 5800 ultra ? Horrible horrible card.

And for years it was vaporware . There are cars that drive themselves in the test labs. Do you consider them vaporware ? I do. Till they are announced and shown they are vaporware. The sledgehammers were vaporware till they were demo'd, P4 5ghz are vaporware. I don't doubt though that one day it will come out. Right now ps3 , xbox 2 and gamecube 2 are all vaporware . One day they will come out. But till they are demo'd they are still vaporware .

Why are you comparing 5800 with Cell, or to put it correctly why are you comparing NV with Sony, 5800 was delayed becse of TMSC whereas Sony intends to manufature Cell in its own fabs.

 

[Grall]

Two points , a chip in hardware and a chip in patent form are two diffrent things.

Duh. You're stating the obvious! However, unless there's any reason to doubt Sony's ability to deliver I don't see why we should. And we don't actually HAVE any reason to doubt, because no hints of trouble in Cell paradise have leaked out to the public.

All we got are the crowings of a few trolls like DM, etc.

Even if things are going grat on the other there is no hard proof of how it performs or if its up to snuff.

Yes? So?
Like I said, unless there's a concrete reason to doubt Cell, why doubt it...?

We don't know what Cell will be, and we don't know what it WON'T be either, so why not just chill instead. Wait and see, like I said. Arguing that it might not be the best thing since sliced bread is pointless. We have no details of what it will be plain and simple!

Second nivida used to say the nv30 was on schedule and things were going well. 5800 ultra ? Horrible horrible card.

Yes, so what? Apples and oranges. ATi said R300 was on schedule and things were going well and YOU KNOW WHAT? THEY WERE RIGHT!

Don't bring up NV30, thank you, it is completely irrelevant in this discussion. To re-iterate: we have no reason to doubt Sony at this point so why do it?


*G*

 

[jvd]

Two points , a chip in hardware and a chip in patent form are two diffrent things. Even if things are going grat on the other there is no hard proof of how it performs or if its up to snuff.

Second nivida used to say the nv30 was on schedule and things were going well. 5800 ultra ? Horrible horrible card.

And for years it was vaporware . There are cars that drive themselves in the test labs. Do you consider them vaporware ? I do. Till they are announced and shown they are vaporware. The sledgehammers were vaporware till they were demo'd, P4 5ghz are vaporware. I don't doubt though that one day it will come out. Right now ps3 , xbox 2 and gamecube 2 are all vaporware . One day they will come out. But till they are demo'd they are still vaporware .

Why are you comparing 5800 with Cell, or to put it correctly why are you comparing NV with Sony, 5800 was delayed becse of TMSC whereas Sony intends to manufature Cell in its own fabs.

actually not all was tmsc's fault. Then nivida lieing about if the chip taped out or didn't tape out let people think the chip was on time and would be a good performer. I was using it as a point that not everything a ceo or a company says is the full truth. Also cell will be made in fabs not yet built and have no prior history of performance.

 

[jvd]

Yes? So?
Like I said, unless there's a concrete reason to doubt Cell, why doubt it...?

We don't know what Cell will be, and we don't know what it WON'T be either, so why not just chill instead. Wait and see, like I said. Arguing that it might not be the best thing since sliced bread is pointless. We have no details of what it will be plain and simple!

Sorry but for a realtive unknown cpu producer , with un built fabs that have no prior history with making chips. There is no reason to believe that everything will go just peachy or that all of thier goals will be met. Just because money was pumped into this and its sony does not make it an automatic succes. which is where it compares perfectly to the nv30.

 

[Grall]

J:
It's *not* as if these people are without experience. Sony's behind the two most successful (stationary) gaming platforms EVER. Toshiba and IBM have tons of experience building fabs and designing powerful microprocessors. These guys design SUPERCOMPUTERS for a living for cryin out loud.

I don't see where you get all this scepticism from. Sure it's a good thing not to fall for the words of snake-oil salesmen, but this is something entirely different. You're starting to border on the irrational here. Instead of going overboard one way and preaching teh c€11 g0$pEl, you go overboard the other instead.


*G*

 

[London Geezer]

be cool. be like me. i don't care if Cell ends up with 300 million transistors instead of 500...

all i want is play my next generation SH and ZOE and MGS and FF games..... see, keyword: PLAY.

 

[jvd]

J:
It's *not* as if these people are without experience. Sony's behind the two most successful (stationary) gaming platforms EVER. Toshiba and IBM have tons of experience building fabs and designing powerful microprocessors. These guys design SUPERCOMPUTERS for a living for cryin out loud.

I don't see where you get all this scepticism from. Sure it's a good thing not to fall for the words of snake-oil salesmen, but this is something entirely different. You're starting to border on the irrational here. Instead of going overboard one way and preaching teh c€11 g0$pEl, you go overboard the other instead.


*G*

I don't get why you think i have all this scepticism. I just don't believe the cell chip will be a 1tflop performer . I don't think it will be cheap. I don't think it will be on time and i don't think it will be trouble free.

I never said it would fail. I think it will succed. I just don't think it will be the be all end all others think.

I'm vocal because there is no proof that the cell chip will be a 1tflop monster yet people are talking about it as if thats the offical released spec. That is what i don't like. And because of that I point out reasons why it wont be.

Btw toshiba is the one with the experiance in the ps2 chips. The ibm tech hasn't allways work well. Look at all the troubles amd had with help from ibm.

 

[PC-Engine]

Meaning what, exactly? Are you doubting Cell will ever appear on the market, despite Sony stating it's on schedule and things are going well with the development, or something?

I don't doubt there will be a powerful CELL processor in PS3, however if CELL is capable of 1 TFLOPS then a company like NEC or Hitachi can also have a cpu just as capable based on the fact that STI doesn't have some magical technology. However I don't believe CELL will be capable of 1 TFLOPS anyway.

As for the proof of concept, I brought that up to make the point that NEC already has the technology to have an array of processing elements running in parallel giving very high performance in the form of a working chip as opposed to a patent.

I don't see your point. Volkswagen has been working on the Veyron 16-4 supercar for years, some said it was vaoprware, yet it made a public appearance just a few weeks ago. Sony will intro Cell one day too. Patience, my young padawan. Patience.

Read above

...Except they'd still be integer units and unsuitable for what's needed in a modern console. Guess you could run one mean version of duke nukem 3D on that array, but nextgen gaming will not be its forte. Re-jigging it to do flops will require re-working the thing from the ground up as ALUs are totally different from FPUs from an architectural standpoint.

I didn't say that the chip will be the final cpu with only higher clocking and scaling. I said it was a proof of concept therefore it can be used as a foundation for a more powerful cpu based on 128 parallel processors.

How does the P10 have 200 32-bit SIMD processors on a 76M transistor chip? You think NEC couldn't do the same based on their very simple proof of concept with modifications? You think it's impossible?

It's one thing having 128 SIMD thingies in parallel on a chip, and another thing entirely to keep all these processors fed with data. How do you propose to keep 128 64 or 128-bit processors running at 2GHz using only off-die memory? You'd need literally hundreds of GB/s worth of very low-latency bandwidth, and such a memory sub-system would be very VERY expensive. That's what eDRAM (and the scratchpad RAM in each APU AND the big register files) is for in Cell, spread the bandwidth load around to prevent bottlenecking.

You have onboard cache. Doesn't have to be huge. 1 MB of cache would be sufficient.

You seem to be slightly short-sighted here. What if you want to do complex AI, wind simulation, cloth physics, inverse kinematics etc etc? A phat GPU will be NO HELP WHATSOEVER there. All a GPU is good for is eye candy, but the environment will be dead just like in current games if there isn't a strong CPU behind it to drive the universe the game's located in.

You won't be seeing physics and AI in games that require 1 TFLOPS of computing power anytime soon

Was it the most powerful CPU in its price class? Um...YES...

Um and how do you know that??? How much did it cost to manufacture initially? What are you comparing it to???

How many thousand $ apiece did they cost?

So are you saying that SONY has some magical technology that allows them to fab the EE at a lower cost than Hitachi and NEC's 8 GFLOPS cpus that came out before it? What makes you think those 8 GFLOPS cpus cost thousands of dollars to make?? Why would it cost that much to fab??? Care to explain?

 

[Grall]

I just don't believe the cell chip will be a 1tflop performer . I don't think it will be cheap. I don't think it will be on time and i don't think it will be trouble free.

M-hm...

I don't get why you think i have all this scepticism.

I'm vocal because there is no proof that the cell chip will be a 1tflop monster yet people are talking about it as if thats the offical released spec.

Then just ignore them. We don't have any proof either way really. Either it will, or it won't, whichever alternative is the more likely is beyond any of us to judge at the moment.

And because of that I point out reasons why it wont be.

But you don't KNOW it won't be.

Btw toshiba is the one with the experiance in the ps2 chips.

There's no such thing as "toshiba is the one with the experiance (sic) in the ps2 chips", Sony's heavily involved in the process as well of course since it's THEIR FRIGGIN GAMES CONSOLE.

The ibm tech hasn't allways work well. Look at all the troubles amd had with help from ibm.

Ibaem is teh d00med!

Oh come on! All companies hit a snag every once in a while, intel did too a couple years ago, etc etc. AMDs problems are AMDs problems btw, and not IBMs.


*G*

 

[Grall]

I don't doubt there will be a powerful CELL processor in PS3, however if CELL is capable of 1 TFLOPS then a company like NEC or Hitachi can also have a cpu just as capable

COULD, yes of course it means they COULD. Doesn't mean they actually ARE making one though. 1Tflop from a single chip is such a massive increase it probably takes a lot of guts of whomever suggesting such a thing to be attempted (assuming 1Tflop is what Sony's aiming for), I doubt a company would try it just on a whim, without a specific buyer for such a product.

However I don't believe CELL will be capable of 1 TFLOPS anyway.

I don't believe either way. I will wait and see.

NEC already has the technology to have an array of processing elements running in parallel giving very high performance in the form of a working chip as opposed to a patent.

It's a working chip that isn't anything like what's needed in a future games console. It's basically irrelevant, it's like saying a firecracker's a proof of concept for a space shuttle.

How does the P10 have 200 32-bit SIMD processors on a 76M transistor chip?

Those processors are extremely limited, that's how they do it. They can't run game code, they're only made to function within the very firmly enclosed environment of a hardware 3D accelerator. Your comparison is somewhat flawed I have to say.

You think NEC couldn't do the same based on their very simple proof of concept with modifications? You think it's impossible?

Have I said it's impossible? No I have not, so shut up. Thank you.

That said though, you're discussing a theoretical scenario as if it was something that's ACTUALLY HAPPENING, that's rather trollish I have to say. So NEC has an 8-bit integer processor array which according to you can be made into a 128-bit FP 2GHz chip, except we have no information whatsoever that is the case! Do you see the futility of your argument? "Oh but they COULD!" doesn't get you very far unless THEY ACTUALLY ARE. So why don't you cool it until there's an article in EEtimes or something that NEC is developing such a chip, because I don't have the patience to counter your fanboyish rants for much longer.

You have onboard cache. Doesn't have to be huge. 1 MB of cache would be sufficient.

No it would not. SIMD data tends to be highly streaming in nature, look at EE for example, it has TINY on-chip memories, just having a cache and then hope that's enough will make you end up with a dud of a chip. Your 128 processors would rip through the cache and then die of starvation waiting for new data to trickle in. One megabyte is likely less data than what's required just for a next-gen 3D model...

You won't be seeing physics and AI in games that require 1 TFLOPS of computing power anytime soon

True, the devtools required for such complex behaviours in games aren't there at the moment. Difference is, if we had such a CPU it would at least be POSSIBLE. Devtools will appear once there's a need for them.

[/quote]

Was it the most powerful CPU in its price class? Um...YES...

Um and how do you know that???[/quote]

Do you know of any other computer system that gave you 6.4Gflops for a few hundred $ back in 2001?

So are you saying that SONY has some magical technology that allows them to fab the EE at a lower cost than Hitachi and NEC's 8 GFLOPS cpus that came out before it?

Not magical. Chip manufacturing is HIGHLY sensitive to the rules of mass-market production. It costs X dollars to develop, test and start production of a chip. Spread those costs out on a few (tens of) thousands of chips = very expensive chips. Spread them out on fifty million (or more) chips = very cheap chips.

What makes you think those 8 GFLOPS cpus cost thousands of dollars to make?? Why would it cost that much to fab??? Care to explain?

I think I just did. Anyway, there's a chance those 8Gflops CPUs were fabricated as gallium arsenide multichip designs (many supercomputer CPUs in the past were, while silicon still sucked when it came to high performance), and gallium arsenide's fast alright, but VERY expensive compared to silicon. Yields are much lower, and there's little to no mass-market appeal to offset the extra costs, it all ends up on the sticker price of the chip.

I think you're becoming so blinded by arguing against me you're losing touch with the actual goals of the discussion. Calm down dude, you're working yourself up for no reason. I don't see why you seem to think I'm claiming STI have some kind of "magic tricks" or whatever to realize Cell, I never ever claimed such a thing. That's entirely your own fabrication.


*G*

 

[PC-Engine]

Um...so what are YOU arguing???

I've already made my point clear. NEC or Hitachi can make a cpu just as capable as CELL, but they don't NEED to. Heck I didn't even need to bring up NEC's proof of concept because their previous single chip GFLOP cpus already beat the EE that came out AFTER it. If they beat the EE before why wouldn't they be able to beat EE3/CELL?

Nobody said they WILL create it, but there's no technological reason why they can't. Down playing the competition with only a patent to stand on is rather stupid don't you think?

It's a working chip that isn't anything like what's needed in a future games console. It's basically irrelevant, it's like saying a firecracker's a proof of concept for a space shuttle.

Do you even read??? Do you know what the definition of concept is??? Do things change from concept to final product???

Those processors are extremely limited, that's how they do it. They can't run game code, they're only made to function within the very firmly enclosed environment of a hardware 3D accelerator. Your comparison is somewhat flawed I have to say.

So can you explain what the difference is in transistor count between units that CAN run game code???

No it would not. SIMD data tends to be highly streaming in nature, look at EE for example, it has TINY on-chip memories, just having a cache and then hope that's enough will make you end up with a dud of a chip. Your 128 processors would rip through the cache and then die of starvation waiting for new data to trickle in. One megabyte is likely less data than what's required just for a next-gen 3D model...

So how does EE function with it's TINY memory??? Doesn't it have 9 FPUs to feed??? Why didn't they add eDRAM???

Not magical. Chip manufacturing is HIGHLY sensitive to the rules of mass-market production. It costs X dollars to develop, test and start production of a chip. Spread those costs out on a few (tens of) thousands of chips = very expensive chips. Spread them out on fifty million (or more) chips = very cheap chips.

So the EE cost $10 at launch???

 

[Megadrive1988]

I very much want to see a custom Nintendo CPU with Teraflop-class performance.

companies that could be working on this CPU include:

NEC
Cray
Hitachi
Fujitsu
Mitsubishi

No reason why it cannot be done. its only a matter of willingness.

 

[Grall]

I've already made my point clear. NEC or Hitachi can make a cpu just as capable as CELL, but they don't NEED to.

If they could or not is totally outside the scope of this discussion. Hell, Motorola or Texas Instruments or any number of other manufacturers could make a 1Tflop single-chip processor, except that's irrelevant, for starters none of these guys are confirmed as designers of a next-gen console CPU. Second, I don't see your point at all really because nobody here's arguing that only Sony & co is capable of making 1Tflop chips, I don't understand where the hell that came from. It's probably deadmeat's fault.

Heck I didn't even need to bring up NEC's proof of concept because their previous single chip GFLOP cpus already beat the EE that came out AFTER it.

Except, it's integer. Except, it's 8-bit. Except, it's USELESS as a gaming device, and hence don't beat EE at all. EE will PULVERIZE that thing in a real-world gaming situation, emulating fp is dead slow with integer, and emulating higher bit ranges than what the chip's got in hardware makes it slower still.

If they beat the EE before why wouldn't they be able to beat EE3/CELL?

Like said before: lemme know when media says NEC is making a single-chip 1Tflop CPU, until then you better pipe down coz you sound like a g-d f*nboy. Wether they COULD and wether they actually ARE are still two entirely different things. We're not interested in what a company CAN do, but what they actually ARE doing.

Down playing the competition with only a patent to stand on is rather stupid don't you think?

1: Stop putting words in my mouth, you sound like a f*nboy when you do that. Nobody's saying anything like that! 2: You're not seriously suggesting Cell only exists as a patent? If so, why are they building silicon fabs to manufacture the thing, wouldn't printing presses be more suitable for the Oita #2 building, not to mention much cheaper?

Do you even read??? Do you know what the definition of concept is??? Do things change from concept to final product???

Yes I read, but do you even understand? You have a concept that isn't anything like what can be used in a games console! A massive SIMD array for 128-bit FP is massively more dense in transistors compared to your concept. They aren't *anything* alike! It's not just a matter of, oh we'll just make it 16x wider and 20x faster and make it do FP instead of INT, it's a question of TOTAL FROM THE GROUND-UP REDESIGN here!

Do you get it? They'd be NOTHING alike! Your "concept" isn't worth the silicon it's lithographed on when it comes as a concept for a games console CPU. You're comparing a two-stroke motorcycle motor to an afterburning turbofan engine. They got nothing in common other than propelling things forwards, and one is certainly NOT proof of concept for the other.

So can you explain what the difference is in transistor count between units that CAN run game code???

Well, PS2 vector units run game code. They occupy at minimum on the order of many hundreds of thousands of transistors.

So how does EE function with it's TINY memory??? Doesn't it have 9 FPUs to feed???

DUH! Because EE's designed for streaming data through it!

So the EE cost $10 at launch???

Your f*nboyish tendencies shine through again, where did I say it cost $10 at launch or even now? Entire console cost was a few hundred $ on the CONSUMER end, stores bought them for less still of course, no matter what the chip cost Sony to manufacture initially.

They knew of course they weren't just going to manufacture a few tens of thousands of chips, so what they took on the chin initially they started making back quite a while ago.

Can you find a single-chip CPU, or indeed a whole computer system that gave you more float performance for less money than EE/PS2?


*G*