Suzuoki's CELL Patent application revisited....

[ChryZ]

Re: ....

but in opinion of some others the patent is just a nice reference/guideline about the architecture ideas and doesn't necesserily have a lot to do with the actual PS3

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

Quote somebody else! Those remarks were not made to support your claims nor those of anybody else. Let's talk about fixed specs, when Sony is ready to talk about fixed specs, everything else is going to be a guessing-game and nothing good will come from that ...

 

[Grall]

Re: ....

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

This will be very interesting, come next year and PS3 specs being revealed to the world... Deadmeat's psychic powers of future-seeing vs. reality.


*G*

 

[Paul]

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

STI intends to have the last laugh.

 

[JacksBleedingEyes]

Re: ....

but in opinion of some others the patent is just a nice reference/guideline about the architecture ideas and doesn't necesserily have a lot to do with the actual PS3

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

Now I realize I aint a regular poster, but I've been reading you feelas for a while now. I respect people like Vince cuz he knows more or less what he is talking about, even though he is quite right wing. PC-Engine gave me the shits to start with but he too is resonable.

Deadmeat on the other hand is a fucking pykie bastard
From what I gather and you all agree, he hates SONY, and tries to bring it down every chance he has got. Now this all might be old news to you, but this is a cry from the silent community of those who dare not say, tell the fucker, that either he starts complaining about Xbox architecture, or he shuts the fuck up about shit that has not concrete definition as of yet. Mr FUCKING princeton I can do my calculation and tell you that "no no you can not fit 4 CELL cores togehter"

You stupid FUCK

Post like this again and its the last post you ever make here .

 

[Evil_Cloud]

Re: ....

but in opinion of some others the patent is just a nice reference/guideline about the architecture ideas and doesn't necesserily have a lot to do with the actual PS3

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

Weren't there supposed to be low and high performance cell chips?
Like one Cell chip able to do 1 TFLPS?

 

[London Geezer]

ERRRRRRRRR.... choose one and only one of the following options:


1) Deadmeat is pissing "some" people off (can't think of who in particular )

2) "Some" people need to chill...

3) Other (specify) ................................................................................

 

[Josiah]

Remember when PS2 debuted, the chips were huge and had very poor yields. Then gradually they shrunk them down and eventually integrated them.

I could see something similar happening with PS3. They'd start out with a huge impractical chip with bad yields, and gradually get the cost of production down (maybe even integrate Cell and VS ~2008). I don't see why they would pour billions into the manufacturing facilities unless they were really trying to push the envelope.

 

[cthellis42]

Not going to quote Jacks, but um... DAMN, man! Didn't eat enough fiber this morning?

 

[Brimstone]

Re: ....

but in opinion of some others the patent is just a nice reference/guideline about the architecture ideas and doesn't necesserily have a lot to do with the actual PS3

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

I'm not so sure what the "consumer price range" will be for the next consoles. Sony kept their PS2 retail price high even when Nintendo and Microsoft launched and still had no problem selling out at a furious pace. With the PS3, Sony could get away with a $399 price tag (if not even higher) and still be able to sell out every unit they make for the first 18 months. Also the console buisness model will be a little different for Sony with a big focus on internet connectivity. It could turn out to be a good revenue stream with each PS3 being internet ready from day one.

 

[Brimstone]

Re: ....

but in opinion of some others the patent is just a nice reference/guideline about the architecture ideas and doesn't necesserily have a lot to do with the actual PS3

Numbers don't lie; you cannot fit 4 CELL cores into single die at consumer price range. This is why I know EE3 will have at most 2 CELL cores.

Now I realize I aint a regular poster, but I've been reading you feelas for a while now. I respect people like Vince cuz he knows more or less what he is talking about, even though he is quite right wing. PC-Engine gave me the shits to start with but he too is resonable.

Deadmeat on the other hand is a fucking pykie bastard
From what I gather and you all agree, he hates SONY, and tries to bring it down every chance he has got. Now this all might be old news to you, but this is a cry from the silent community of those who dare not say, tell the fucker, that either he starts complaining about Xbox architecture, or he shuts the fuck up about shit that has not concrete definition as of yet. Mr FUCKING princeton I can do my calculation and tell you that "no no you can not fit 4 CELL cores togehter"

You stupid FUCK

Deadmeat has been very consistent on the Cell subject. For people like you to attack Deadmeat instead of countering his conclusions is just plain sad. Grow up.

 

[nonamer]

I think Sony, assuming the original patent is true, can fit 4 cores on a single chip, if just barely. I based this on the guess that a single core with all it associated eDRAM and APUs should be comparable in size to an Itanium on .13u today, so with .065u 4 will fit. However I don't see this monster going anything like 4 Ghz. More like 1-1.5 Ghz like I've said before.

 

[Grall]

3) Other (specify) ................................................................................

3) We haul that stinky ol' carcass outta here. Peace and quiet resumes.

Problem solved.

However I don't see this monster going anything like 4 Ghz. More like 1-1.5 Ghz like I've said before.

Uh, dude... A CPU, 2005, at .65u, at ONLY 1 GHz, what are you smoking man? Hell, even VIA cracked the 1GHz barrier long ago with their toy processors...


*G*

 

[jvd]

Remember when PS2 debuted, the chips were huge and had very poor yields. Then gradually they shrunk them down and eventually integrated them.

I could see something similar happening with PS3. They'd start out with a huge impractical chip with bad yields, and gradually get the cost of production down (maybe even integrate Cell and VS ~2008). I don't see why they would pour billions into the manufacturing facilities unless they were really trying to push the envelope.

Only problem is if the yields are really bad trhey wont ship enough systems. Not only this but when the ps2 launched for the first year it went up against a dieing dreamcast. Not a newly launched xbox2 and gamecube 2 .

 

[Grall]

Only problem is if the yields are really bad trhey wont ship enough systems.

I don't think you need to worry about their yields, Sony + gang are starting Cell production around a year in advance of launch.

Leave the worrying to the pros, man. They know their stuffs!


*G*

 

[jvd]

Only problem is if the yields are really bad trhey wont ship enough systems.

I don't think you need to worry about their yields, Sony + gang are starting Cell production around a year in advance of launch.

Leave the worrying to the pros, man. They know their stuffs!


*G*

A year in advance ? That means they would have already has samples back from the lab . Since a 2005 launch is a year and 4 months away. You can bet they wouldn't want to send out a buggy chip. Yet the fabs are not done yet so how can that be possible ? fabs aren't going to be finished before second fy 2004. So they will have a half a year at most to make the chip. But then you have to factor in making masks, tweaking the lines and of course making sure there are no flaws in the design.

 

[DeadmeatGA]

...

To nonamer.

I think Sony, assuming the original patent is true, can fit 4 cores on a single chip, if just barely.

Such a 4-core "Server" chip will measure something like 450 mm2 and will be quite unsuitable for console use.

Recall what Suzuoki patent application said, two cores for workstation chip, one core for visual station chip, and one GS3 core for PDA chip.

However I don't see this monster going anything like 4 Ghz. More like 1-1.5 Ghz like I've said before.

Agreed. Going into 65 nm doesn't guarantee 4 Ghz.

To Grall

A CPU, 2005, at .65u, at ONLY 1 GHz, what are you smoking man?

Going into smaller geometry does not automatically guarantee higher clock. Pipe architecture has more to do with it.

Hell, even VIA cracked the 1GHz barrier long ago with their toy processors...

All X86 processors long long stretched pipes for some time now, including VIA. There is no guarantee that CELL VUs will have long pipes; to the contrary they will try to keep the pipe short to save on transistor counts.

Sony + gang are starting Cell production around a year in advance of launch.

The first batch of chips are for testing and debugging only.

 

[Grall]

A year in advance ? That means they would have already has samples back from the lab . Since a 2005 launch is a year and 4 months away.

2005 launch is more likely 2+ years off... They wouldn't launch first of january 2005 of course. Cell goes into production H2 04, this has been confirmed already by Sony.

Going into smaller geometry does not automatically guarantee higher clock.

Assuming only 1GHz or slightly over it for a .65nm 2005 design seems being overly pessimistic to me. While it would be a great surprise to me if Cell hits 4GHz (though not exactly an unpleasant one), I do expect slightly more than one and a quarter. Say 2.5 or so doesn't seem unreasonable, maybe even 3. EE ran at 300, Sony aiming for a 10x increase would be a nice marketing number for them if nothing else.

It's funny seeing you you parrot stuff you've overheard in other conversations about pipeline lengths etc. Facts are you have no formal knowledge on the subject, you're a CMOS quack, and it shows. You toss out guesswork casually as if it was facts written in stone, I find it kinda annoying actually.

The first batch of chips are for testing and debugging only.

Ummm, duh man. They sure won't be test batching for a whole year though.


*G*

 

[DeadmeatGA]

...

Assuming only 1GHz or slightly over it for a .65nm 2005 design seems being overly pessimistic to me.

Well, most top end RISC processors are clocking at 2 Ghz or less currently(PPC G5, Power4, Itanium2, etc), so I am not being overly pessimistic.

I do expect slightly more than one and a quarter. Say 2.5 or so doesn't seem unreasonable, maybe even 3. EE ran at 300, Sony aiming for a 10x increase would be a nice marketing number for them if nothing else.

That is if the design allows it.

It's funny seeing you you parrot stuff you've overheard in other conversations about pipeline lengths etc.

I don't parrot things I overheard from others, I know more than most of them.

Facts are you have no formal knowledge on the subject

Standard CPU pipeline.

1. I-fetch
2. Decode
3. Address calculation(In case of memory instruction)
4. D-fetch
5. Instruction Counter Increment
6. Execute
7. Write-Back
8. Branch to No.1(Or quit)

You toss out guesswork casually as if it was facts written in stone, I find it kinda annoying actually.

I find you quite annoying myself too.

 

[Megadrive1988]

EE ran at 300, Sony aiming for a 10x increase would be a nice marketing number for them if nothing else.

PS1 CPU ran at 33MHz, PS2's EE runs at just under 300Mhz. PS3's CPU will no doubt run at 3GHz.

if PS3 CPU runs at 3Ghz, we would need more than 4 Processing Elements
to reach 1 TFLOPs, if that is the goal.

Mercury News reported that PS3's main chip would have 8 PPC cores and 64 APUs. 72 processors on one chip. it didn't mention the GPU/Visualizer
which would have its own set of processors.. thus, I'd say the Mercury News article ment that PS3 CPU would have 8 Processing Elements.

4 Ghz seems too high for PS3 CPU. but LESS than 4 Ghz with only 4 Processing Elements (4 PPCs, 32 APUs) would mean the CPU falls short of 1 TFLOPs. no?

Therefore, keeping the clockspeed down to reasonable levels means more processors (more than 4 PEs) are needed to reach Sony's minimum goal.

 

[zidane1strife]

Yet the fabs are not done yet so how can that be possible ? fabs aren't going to be finished before second fy 2004.

Didn't pana. say they fabs are finished in jan?

will measure something like 450 mm2 and will be quite unsuitable for console use.

...

Assuming only 1GHz or slightly over it for a .65nm 2005 design seems being overly pessimistic to me. While it would be a great surprise to me if Cell hits 4GHz (though not exactly an unpleasant one), I do expect slightly more than one and a quarter. Say 2.5 or so doesn't seem unreasonable, maybe even 3. EE ran at 300, Sony aiming for a 10x increase would be a nice marketing number for them if nothing else.

They've seen MS spec sheet strategy... indeed to pump the Ghz as high as possible they will...

Mercury News reported that PS3's main chip would have 8 PPC cores and 64 APUs. 72 processors on one chip. it didn't mention the GPU/Visualizer
which would have its own set of processors.. thus, I'd say the Mercury News article ment that PS3 CPU would have 8 Processing Elements.

...

PS


.25micron... .18micron... .15micron... .13micron... (.1micron lengedary barrier, at which a 500M cpu was said to be the goal, with teraflops perf... plans changed... technology improved, the fastest and smallest transistors were gained...)

90nm... 65nm(nigh 45nm)...

They thought, they'd give us peak teraflops perf at .1micron.... Now that perf... might very well be sustained...

TWO chips Heavily contributing to the gphx at nigh 45nm... likely 2B+combined transistors at multi-Ghz speeds...