What is the true spec of PSX3 CELL???

PiNkY said:
Quote Panajev
The funding in the next three years given to SCE for R&D on the Cell project is going to be around $4 Billions ( of course this is going to include fab costs for part of Nagasaki #2 and part of Oita #2 ).

AMD totaled its investments for bringing its 300mm fab in dresden online at approximatly 2.5 billion dollars. The quoted 400 mio figure is for isa / chip design developement as are all comparision figures.



[url]http://www.amd.com/us-en/Corporate/AboutAMD/0,,51_52_502_509,00.html
[/url]

Thanks for the link, still only one floor of Nagasaki #2 is for CELL production and Oita #2 is co-financed by Toshiba ( who has "the name" on the fab ).

So, just for fun ( these are not exact figures of course ;) ), we say that Sony is spending like $1.25 Billions on Oita #2 and let's say $2 Billions on the CELL part of Nagasaki #2( only the first floor of Nagasaki #2 is dedicated to 65 nm CELL production ).

This would account for $3.25 Billions... so that means we still have a $800 Millions head-room left and we can add the $400 Millions that Sony, IBM and Toshiba initially pledged for Cell R&D.

Sony also gave for other Semiconductor R&D spending other $4 Billions ( non CELL related R&D ).

I think the 90 nm part of Nagasaki #2 would fall in this other category.
 
Well investment figures may well be considered kind of a black art, but if a publicly held company openly qoutes a 400 mio $ dollar investment into a certain venture and unofficially injected another 800 mio $ into said effort from another publicy announced venture, this would implie a severe break of laws. Besides, i (no expert in this field) would think your figures are well on the low side for establishing to fabs with bleeding edge production lanes not even considering that the regarded company probably would have to license most of its know-how from other companies (mostly Toshiba and IBM) as Sony (by itself) can hardly be considered market leading in regards to process know-how when the likes of Intel, IBM, AMD, Toshiba and Infineon are considered. And partnership of (otherwise) competing economic entities certainly doesn't imply the free flow of key know-how (there is (imo) no way Toshiba delivers fabbing know how to Sony free of charge).
 
Pinky, you seem to make several assumptions that would skew the investment of a new 'Fab'. When Sony or Toshiba invest, say, half a billion dollars they're not breaking new ground and building a brand-new facility such as AMD Dresden or IBM's East Fishkill. Rather, AFAIK, these are production lines at established facilities. AMD Fab40 cost ~$2.5Billion for the entire facility.

Beyond this, process techology transfer would seem to fall under the aegis of the 2001 STI agreement, if not previous ones between Sony and Toshiba (eg. look at OTSS for precedence).
 
...

To pinky

1.) Comparison PS2 VU - PS3 FP APU: I don't expect these to differ as much in complexity as e.g. Deadmeat implies (for whatever reasons he has).
1. VU2 now does integer vectors, in addition to floating point vectors.
2. VU2 now is a self-running processor, instead of being a slave device attached to CPU.
3. VU2 has larger address and is capable of addressing external memory.
4. VU2 is to be clocked faster, so its pipe has to be stretched to support higher clock.

2.)Two-chip solution: I am a bit sceptical of the two chip version that many are expecting.
Two chip solution of CPU+GPU is a given. I am not disputing this configuration, but the actual number of CELL cores in the EE3.

To Vince

Lets forget that PSP is a portable device with power, thermal and size requirements.
Size requirement, yes. Power and Thermal requirement, no. eDRAM sucks up little power and heat generation is minimal, Sony didn't fit more eDRAM only because it couldn't, not because of power and thermal requirement. Sorry to disprove your belief that Sony is somekind of fabrication god and could do what others can only dream of(Like fitting 132 FPUs in single die)
 
Re: ...

DeadmeatGA said:
how much do you throw in there before it becomes appropriate? who decides?
Developers do. And they aren't happy with 10 MB of developer accessible RAM size...

that's not what I meant. Devs generally are insatiable (enough? plegh).

what what meant is who decides (not whom best decides) what the limits are in terms of the product they want to deliver? why not higher res screen? Firewire support if not why not?
 
Lets say PS3's CPU, the Emotion Engine 3 / Broadband Engine has
1 TFLOPs peak performance. as others have said, the actual performance will be well under 1 TFLOPs, and by the time we get to in-game realworld performance, it might only be 100~200 GFLOPs.
 
PiNkY said:
AMD totaled its investments for bringing its 300mm fab in dresden online at approximatly 2.5 billion dollars. The quoted 400 mio figure is for isa / chip design developement as are all comparision figures.
http://www.amd.com/us-en/Corporate/AboutAMD/0,,51_52_502_509,00.html
AMD's FAB 30 is a 200 mm facility that was build in the late 90s for initial production at the 180nm node and has been updated since for 130nm production.

I don't see how any valid comparison can be made between it and whatever Sony is building, the differences are just to staggering.

cu

incurable
 
1 TFLOPs peak performance. as others have said, the actual performance will be well under 1 TFLOPs, and by the time we get to in-game realworld performance, it might only be 100~200 GFLOPs.

... if they screwed the design and threw out the embdded ram it very well could...
 
jvd said:
Paul said:
where as an xbox 2 would most likely have more than 512 megs.

More than 512mb? Not a chance, too much money.

I would say expect 256mb main ram for PS3, 16-32mb eDRAM for Cell than 64mb for Rasterizer.

Xbox 2 with 512mb of unified.

Last month i got 512 megs of ddr 3700 for 30$ usd. So i really odn't see a problem for them putting in more ram.

This is an arguement similar to one used by individuals trying to claim that MS was breaking even on the xbox back when it was released. They'd price out equipment on pricewatch, add it up, and come up with some number for how much the thing cost. This is unrealistic. What you should be looking at is, how much did it cost to produce that 512MB of DDR3700? Take that number, add in some reasonable profit margin for the quantities being purchased, and you likely have a much more reasonable number.

Ram is an especially bad example because last generation memory is often sold at or even below cost to clear it out after it's become obsolete. MS or any other console maker isn't going to be bargain hunting for their console parts, they'll have contracts set up, and they'll be paying the cost + some profit margin.

Personally, I'm guessing we'll see 512MB+ of memory in the xbox only if they decide to go with mass production or other types of memory with low production costs. 256-512MB of memory is probably more likely if they are going to use anything reasonably exotic with a higher production cost. I'm guessing had the xbox used dual 2.7GB or 2.4GB per second channels, we'd probably see it with 128MB of memory rather than 64.

Nite_Hawk
 
Well if MS feels they can't trump the PS3 in certain areas (computational power, for example), they may opt for a whole wack of RAM instead of putting in a costly high end/custom CPU.

A PS3 at 1Tflop with 256 megs of ram vs an Xbox2 with a midrange CPU and ~1 gig of ram would even things out nicely.
 
Re: ...

DeadmeatGA said:
1. VU2 now does integer vectors, in addition to floating point vectors.
2. VU2 now is a self-running processor, instead of being a slave device attached to CPU.
3. VU2 has larger address and is capable of addressing external memory.
4. VU2 is to be clocked faster, so its pipe has to be stretched to support higher clock.

1: Why do you call Cell's PUs "VU2"? You have no idea if they'll be based on EE VUs or not.

2: VU2 never was a slave device. It's always been stand-alone.

3: Are you sure about this? Mind backing it up with a quote from some kind of technical document? Anyway, it would seem inefficient to me. Using the integrated DMAC to burst whole chunks of data into its private memory would lead to less bus contention and fewer page breaks.

4: Well, DUH. As for how much overhead this leads to, transistor-wise, nobody but the actual people involved knows.

Sony didn't fit more eDRAM only because it couldn't

LOL, you don't know this! You're just making shit up!

not because of power and thermal requirement.

You're disputing that eDRAM draws power and generates heat? At 333MHz, you can bet your ass it's gonna draw power and generate heat bud.

Sorry to disprove your belief that Sony is somekind of fabrication god and could do what others can only dream of(Like fitting 132 FPUs in single die)

Rofl, you didn't 'disprove' anything. You simply made some shit up and expected everybody to accept that as the truth. As for your claim of 'fitting 132 FPUs', remember that a FMAC unit does not a FPU make. As usual, you mix stuff up.

Besides, of course there's nothing stopping anyone from putting 132 real FPUs on one die. There was this company once that were gonna do a graphics chip with 1500 processors on it on a single chip doing pure software rendering to accelerate 3D graphics, on a far coarser process than the 65nano Cell will be manufactured at. It never took off, unfortunately; was a really cool concept.

*G*
 
Re: ...

DeadmeatGA said:
To Vince:
Sorry to disprove your belief that Sony is somekind of fabrication god and could do what others can only dream of(Like fitting 132 FPUs in single die)

Last I heard an IEEE complient FPU isn't that large. I don't know the exact numbers, but the ARM10 (250K transistsor budget) was IEEE 754 complient.

Perhaps if someone could explain it to us?
 
zurich said:
Well if MS feels they can't trump the PS3 in certain areas (computational power, for example), they may opt for a whole wack of RAM instead of putting in a costly high end/custom CPU.

A PS3 at 1Tflop with 256 megs of ram vs an Xbox2 with a midrange CPU and ~1 gig of ram would even things out nicely.


well... not really to be honest.... the difference would be like ps2 and DC-with-more-memory now.


explanation: PS3 will be doing amazing things thanks to the monstrous number of polygons it will *apparently* push, while Xbox2 will be the same PC-level-graphics-with-hi-res-textures-and-that's-it.

example:

PS3 games would have trees with thousands upon thousands of realistically looking and moving single leaves on tens of branches, while Xbox2 would have trees with big hi-res panels of digitized branches and leaves.

or

ps3 would have water that not only looks real enough, but also moves like real water would, with realistic fluids physics, while Xbox would have bump mapped water that moves a bit better than the current generation.

DISCLAIMER: the above is *if things go down that road*. which we still dont know, but its all speculation isnt it...
 
Swaying trees and moving water can be done a GPU. Even the PS3 in the fanb.. dream config won't be able to simulate reality. Not to mention that development even today is already so time consuming that making even more advanced games will take an entire gen for some teams.
 
cybamerc said:
Swaying trees and moving water can be done a GPU. Even the PS3 in the fanb.. dream config won't be able to simulate reality. Not to mention that development even today is already so time consuming that making even more advanced games will take an entire gen for some teams.


the point i was trying to make, all in speculation mode OF COURSE, is that LOADS OF POLYGONS WITH TEXTURES HI-RES ENOUGH FOR 480P OR 720P compared to NOT SO MANY POLYGONS BUT HI-RES TEXTURES is gonna be the crucial difference....

wherever u do ur "swaying trees", that doesnt change anything... if the GPU hasn't got enough power to compare to a 1Tflop chip, be it CPU or GPU then the visuals won't be the same...

this is all speculation of course, we dont know anything yet so i'm just saying... for all we know the GPU in PS3 wont be able to display whatever comes out of the CPU... just like we dont know whether the CPU in Xbox2 will be powerful enough to sustain the GPU... :D
 
Not to mention that development even today is already so time consuming that making even more advanced games will take an entire gen for some teams.

There are ways around that, recycle content from others, or simply settle for less spectacular looking gphx than the top dev.'s games.
 
I hope to drop some real numbers on you soon.


I'm just waiting to collect a team of reviewers before sharing PS3 specs.
I see a lot of crap that just seems silly.
Decided "Why not post the real details?"

Coming soon.
 
Back
Top