Behold, CELL specs!!!! (I think it is..)

Status
Not open for further replies.
fast polygon rendering help to CPU in collision computing
first: 'render object 1' , after setup GPU and render 'object 2'
if the two object collision then gpu send Z FAIL error and send coordinates to cell



coll.JPG
 
Well, from what we know of Xenon´s CPU, it´s supposedly in the 90nm process and doesn´t seem to have those issues (or at least we haven´t heard rumors about it).

I understand that the investment included fabs as well, but it´s still a pretty hefty investment and less money could have been spent if the chip isn´t significantly more powerfull than Xenon´s.
who said the xenon's cpu will be as big , as highly clock and offer as much performance . I expect the cpu in xenon to be slower than the cell in the ps3 , but the gpu in the xenon to be faster than the gpu in the ps3 . WIth both launching with in 6 months of each other .

1 PU would offer 256GFLOPs, and offer sustained performance of what? 100GFLOPs or less? I understand that that´s still a very powerfull chip when put in context, but after such a seemingly big investment to come up with a chip that isn´t even half of what their stated target is, I feel I still would feel somewhat dissapointed.

Look i can put 500billion dollars into a space program and my goal would be to put a person on mars . That doesn't mean its going to happen .

money doesn't equal sucess .


Just because the money was spent doesn't mean the goals will be met .



Its nice to dream big but if sony was actually capable of making a cell chip on 65nm that would hit 1tflop performance and be cost effective enough to put in a 300$ console don't u think on 90nm they could put out a 500gflop chip and sell them for the pc for 50$ ?

I don't see any of this happening thus i see a realistic 256gflop peak with a high 80gflops sustained


From the xenon i see 128gflop peak or less with a sustained raiting of 50-60 gflop
 
Inane_Dork said:
It's true you can stop a lot of the memory problems by keeping the info streaming around the processors. But lots of programs are not amenable to that design. Excel, for instance. Oracle, for another.
Well..then it will not surprise you the fact CELL architecture was not designed to run spreadsheet softwares :rolleyes:
 
Only 27 TFlops?! I'm so disappointed now! The GPU must be ridiculously weak. How will I now keep my 2 GB DVD home database adequately refreshed??? :( :p Sony has hornswaggled me with da hype once again! Curses!
 
jvd said:
I expect the cpu in xenon to be slower than the cell in the ps3 , but the gpu in the xenon to be faster than the gpu in the ps3 . WIth both launching with in 6 months of each other .

Why do you expect the Xenon GPU to be faster?

It is possible but I am not sure we have any information to say one way or the other. It looks like both GPUs will be on a 90nm process which is really the important thing in many ways (you can only pack so much into a space). And if we consider relative die sizes (something we do not know of course) it would be safe to say they will possibly be comparable in performance.

But the devil's advocate says. These designs are so different that the PS3 GPU may end up substantually faster:

1. R500 taped out in January-ish, PS3-GPU wont tape out until ~fall of 2005 which is almost a year of extra development. That counts for something.

2. If the CELL is doing all the vertex shading, the GPU can dedicate all the Vertex Shader realestate to pixel shading. Lets say both have 24 Shader Units; what happens if in a game R500 uses 4 for VS and 20 for PS. The PS3 would use all 24 for PS and the CELL would do VS.

3. One has to consider the effeciency of the R500s unified shaders. Would 24 Unified Shaders do Pixel Shading better than 24 dedicted Pixel Shaders? I doubt it.

4. The PS3 may have twice the memory bandwidth of the Xenon.

5. Unified Shaders will take up more space than stand alone PS units. If R500 has 24 Unified Shader units, how many more stand alone PS units could nVidia fit in the same transistor space? 28? 32? More?

6. Sony will have the advantage of seeing the Xenon GPU. If they think it is hands down faster (or feel the need to push ahead in this area) Sony can add more pipelines and take the hit in yeild knowing that a 65nm shrink will be much closer for them than MS (the year headstart means MS needs to bite the bullet longer to so speak). They can afford to put out a low yeild GPU chip whereas MS would be stuck with a low yeild GPU longer and thus cost them significantly more.

Of course it seems that the Xenon CPU's may be able to do some VS. And obviously if CELL is doing VS one most consider that when they compare CPU performance since it is doing a task the Xenon GPU is doing. So if CELL is 50% faster one has to take into consideration that a big chunk of that speed is being spent on GPU tasks (not to mention it will be using CPU cache and such).

But of course the flip side is the CELL has the power to burn. So saying, "I expect... the gpu in the xenon to be faster than the gpu in the ps3" is not really clear cut, nor does it really consider the importance of the system design. Neither of these systems are PCs with a clear cut CPU does 'X' and CPU does 'Y'. Obviously there is a lot more too this than just the CPU/GPU comparison--the total system design is more relevant because the PS3 may not be using a traditional "CPU does Engine/AI/Physics and GPU does rendering." If the CELL does the VS then that changes how we look at both systems CPUs and GPUs.

I think such comparisons miss the boat on the Xenon design. The Xenon is an extremely effecient design that is equal parts powerful and flexibility. You have the GPU and CPU doing cache reads, you have Unified Shaders but you also have the option for one of the CPUs doing Vertex Shading (or picking up the slack). The design is NOT focused on brute force, but more on clean effecient design. This is very similar to the GCN (no surpise there) and how the goal was not to just put power behind it, but to have an effecient design that had a lot of accessible power instead of brute force sitting idly by.

The PS3 sounds the exact opposite. CELL is massively parallel and has a huge thoroughput. If CELL does VS you are looking at a different design philosophy than Xenon. You have a GPU with a very fixed function--rasterizing. This could mean a LOT of performance in that specific area, and if the CELL has the performance we are considering you could see geometry intensive games. And with the fast memory interface there seems to be some flexibility there also.

Both will be very powerful, but I think based on what we know their design emphasi are totally different. Xenon seems to be about flexibility and general purpose computing power where CELL/PS3 is about more specific HW with a lot of power. Of course this is all general (there will be exceptions) and this is speculation on what we know, so I could be dead wrong.

Most of us are in the dark, but from what we know it seems both PS3 and Xenon are shaping up to be VERY good designs that meet their design goals. Both are powerful, but in different ways.

The question goes back to who will have the most hype among casual gamers, 3rd party support, can get killer apps out quickest, and who is easiest/most affordable to develop for. System cost and perceived value will also be big factors.

::Sigh:: It will be nice once we can begin adding Nintendo's plans into these talks.
 
Quaz51 said:
The real PS3 CPU:

PS3-CPU.JPG

CPU diagram is actual size of CELL die :oops: We are going to need to cool this puppy with a gold Heatsink on Titan in one of those liquid methane oceans!

:LOL:
 
Quaz51 said:
PS3 CPU wafer

Interesting, the same site that has that CELL CPU wafer also had a pic of the new PS3 case they will be using to ensure the CPU remains cool:

e22g1-01.jpg


I kinda like it. It would match my GCN well ;)
 
nAo said:
Well..then it will not surprise you the fact CELL architecture was not designed to run spreadsheet softwares :rolleyes:
I didn't say it was. I was just pointing out that not everything can utilize the advantages of the CELL architecture. Why? You didn't.

It's really that simple. I'm not trying to steal your precious.
 
That would seem to be true for any computing architecture to date. So this pretty much treads well into the stating the obvious territory.
 
Inane_Dork said:
I was just pointing out that not everything can utilize the advantages of the CELL architecture. Why? You didn't.
I did that Inane_Dork, a lot of times on this board. I'm sorry you're not aware of that.
It's really that simple. I'm not trying to steal your precious.
Actually I was one of the first people here to talk about/address many of CELL architecture real shortcomings (not this CELL can't effciently run microsoft office babbling) so I believe you should redirect your cheap jokes towards other poeple. Thank you.

ciao,
Marco
 
randycat said:
That would seem to be true for any computing architecture to date. So this pretty much treads well into the stating the obvious territory.
Fair enough. But if true, it certainly isn't worth this series of posts, is it? :p


nAo said:
I did that Inane_Dork, a lot of times on this board. I'm sorry you're not aware of that.
I'm not trying to offend or one up you. I was simply agreeing with what you said and stating qualifications to it to match what I think.

Actually I was one of the first people here to talk about/address many of CELL architecture real shortcomings (not this CELL can't effciently run microsoft office babbling) so I believe you should redirect your cheap jokes towards other poeple. Thank you.
Geez! I must've really hit a nerve!

I don't know what's with all the personal attacks, but I'm done here. For whatever reason, this conversation just didn't work out. Sorry.
 
Status
Not open for further replies.
Back
Top