Dave Baumann from B3D suggest Cell is a failure

Elios83

Member
According to highly respected insider Dave Baumann from B3D, the fact that Sony had to ally with nVidia means that Sony's "performance targets they had in mind for PS3 Cell processing are off" and that Cell won't be the "end all be all" CPU that STI promised it to be. So they need nVidia on board to save the day:

http://www.beyond3d.com/forum/viewtopic.php?t=18811&postdays=0&postorder=asc&start=60

http://www.beyond3d.com/forum/viewtopic.php?t=18811&postdays=0&postorder=asc&start=480

So, how about this? Do you think that Cell failed to deliver on Sony's part and that this new alliance proves a defeat for Sony initial intentions?
 
Yes that is exactly what it suggests please accompany this post with all the qualifiers and doomsday statements that are required for damage control.

Thank you and have a nice day.
 
Or it could suggest that the PS3 is some kind of crazy mix of brute power (CELL) and finesse (Nvidia). We just don't know yet.

People have been speculating that Nvidia might work with Sony in some capacity for a few years now though.

My feeling is this always seems to happen though right before Sony offically announces something like ...

- Before the PSP announcement, speculation was rife that the unit would cost $300 or maybe even more.

- Before the infamous Playstation at $299.99 announcement at E3 that floored Sega's Saturn, people thought the PSX would be $400 or so.

- When the PSP was initially announced people assumed PSOne level visuals.

- With the PS2 people said DVD playback was "impossible" because it would kill Sony's DVD division and that those early demos would never be indictive of in-game visuals, etc. etc. etc.


I just don't bet against these guys anymore. Any time there's been questions raised about their hardware, they always not only match what you expect, but go one further.

I think come March 2005, it's gonna be the same old story. Everyone is gonna be floored.
 
This is Sony you are talking about and sony is awesome!
Nintendo should be out of business!!!1

Uhm... I do think you have a point, but sony isn't debatable at GaF. Sony never makes mistakes!

I thought cell was CPU en Nvidia was for GPU. Different things.....
 
soundwave05 said:
Or it could suggest that the PS3 is some kind of crazy mix of brute power (CELL) and finesse (Nvidia). We just don't know yet.

People have been speculating that Nvidia might work with Sony in some capacity for a few years now though.

Hope you are right!
 
Maybe Cell is just too expensive to manufacture, maybe they planned ps3 to use 4 cells, 2 for AI/Physics tasks, and 2 for graphics, and have decided to just use 2 for AI/Physics and get a Nvidia GPU so PS3 is cheaper
 
I think it's unlikely that Cell not hitting performance targets caused Sony to turn to Nvidia. There's simply no way Sony could have known how Cell performs two years ago.
 
Cell dooom3d!


Sony goes to plan B (or A?) Sony makes Emotion Engine PS3 (tm) with tried and true MIPs core, several Super Vector Units, and Nvidia GeForce 7 GPU.
 
It also might be a bit of a development compromise. A completlely propietary set up might have alienated some developers, working with Nvidia though could potentially swing a lot of PC devs over to Sony's camp.

I think the CELL performance is still going to be really insane from a raw processing power standpoint, but perhaps they just felt they needed to be a little more conservative with the GPU aspect this time around.

My guess is the PS3 will be an architecture that holds an unbelievable amount of raw power for those devs that want to really push those parralell CELL units and get down to nitty gritty programming.

But for those devs that just want to easily get an engine up and running or a more familiar set up, having an Nvidia GPU in there certainly doesn't hurt.

Also perhaps Nvidia got a bit desperate themselves being dumped by Microsoft and made Sony an offer they couldn't refuse. It was looking pretty bad there for Nvidia with both Microsoft and Nintendo choosing ATi, Nvidia scores big here by getting the Sony deal, since the Playstation is far and away the biggest console brand.
 
Technically speaking I have no idea how CELL works, or any other processor for that matter. But what seems obvious to me, and I maybe wrong but all the next gen machines are going to be using PC graphic solutions. All will be the new generation of Graphic GPU's and all are using powerful CPU. The time difference between the machines is short so ultimately looking at everything thats gone on before in the console world and cutting through the PR bluster released by MS, Sony,ATI and Nvidia I've come to the conclustion that the differences in machine capabilities will not be an argueing point. However the software thats made for them should be, because that where Sony could lose, hold on to or even increase its market share and off course the same holds true for MS and Nintendo. I'm really not worried about what the specs are, I want to know what the games are and so should most gamers.
 
xexex said:
Cell dooom3d!


Sony goes to plan B (or A?) Sony makes Emotion Engine PS3 (tm) with tried and true MIPs core, several Super Vector Units, and Nvidia GeForce 7 GPU.

It will be... The greatest invention since the printing press.
 
soundwave05:

> working with Nvidia though could potentially swing a lot of PC devs over to Sony's camp.

It won't make a difference if Cell is a bitch to code for. Besides, the fact that it's a PlayStation is a lot more important than the name on the graphics chip.

Noone but M$ cares about PC devs anyway.
 
FriScho said:
CELL is IBM and GPU is NVIDIA... what did Sony do with PS3? Overseeing the projects and giving directions?

CELL is Sony/IBM/Toshiba
GPU is Sony/Nvidia

I'd say "overseeing the projects & giving directions" is an understatement, accurate to a small degree, but still seriously understating their involvement.
 
PS3 is still Kutaragi's baby, and its going to be his design philosophy -- which is a sh-tload of brute, raw power.

The Nvidia move is Sony probably just covering their bases. You like a PC-centric dev environment?

Fine, Sony gives that to you to too.

I think the Sony philosophy will be for those that want to just get a game up and running with minimal fuss ... fine let the Nvidia GPU be your main go through. But if you're more adventurous and willing to work beyond those limits, then you can work through the CELL processors and use the Nvidia GPU as an auxiliarly component.

I think the GPUs in both Xenon and PS3 will be comparable in power, but PS3 will pull away with the CELL processing units, which will be more than just "a regular CPU core".
 
soundwave05 said:
The Nvidia move is Sony probably just covering their bases. You like a PC-centric dev environment?
You don't get a PC centric development environment just by having a PC graphics chip (and I suspect this one will be heavily modified).

PC developers will still go to M$ who they know will offer money and development assistance.
 
We'll see.

I think Sony is giving the developers the options to use the PS3 how they want.

If you want a more traditional architecture, then just go through the Nvidia GPU and the CELL units can just act as a regular CPU. This is more for the dev team thats under a tight deadline or doesn't have the staff/patient to deal with a more challenging CELL set up.

If you want to really unlock the power of the unit though, the CELL processors will likely be highly programmable with a lot of raw power to use however the developer sees fit.

I don't think the PS3 will be a tougher architecture than the PS2. IBM has already said its a priority that it is not.

Of course developers will face daunting challenges due to the technical prowess of these machines, but that's going to be the same deal all around the board.
 
a) this didn't need its own thread. we have a massive and active nvidia/ps3 thread already

b) you're misrepresenting his quotes. He used words like "seems" etc. that you have dropped. He's not stating these things as fact. He also said "Cell isn't the be all and end all of all types of processing.", not ""cell isn't the be all and end all" that STI promised". And that's true - if you look at what IBM has been saying, Cell is biased toward apps that require massive floating point performance (like graphics, digital media etc.). I doubt he would be happy with how you are characterising his quotes - a couple of other people jumped in afterwards with words like failure, and Dave was quick to stamp that out.

and

c) with all due respect to Dave Baumann, I'm sure he'd be the first to acknowledge that he isn't as aware of what's going on with cell as he is with the PC graphics landscape.

IMO (and in the opinion of many others who replied to those comments on B3D), choosing NVidia to co-design the PS3 GPU isn't necessarily a reflection on Cell's performance. For a start, we don't even know that the GPU won't be cell based (although that is unlikely, imo), but even if it's not, there are many reasons to go with an experienced graphics partner asides from hardware (software, tools etc., developer pressure, "features", cost).

We'll see in February, and after, how cell is shaping up. But from what we know from the monday before last, things seem to be on track.

IMO, if it's true that Sony have been pitting different options against each other as an alternative to a preferred Cell-based GPU, the reason they chose a "plan B" probably came down to cost - Cell may well be expensive to manufacture at the moment, but instead of compromising CPU power to include a second cell-based chip, they've decided to go with their next best option. I.e. I doubt it has anything to do with missing performance targets, and possibly more to do with missing cost targets. IF that particular scenario is true ;)
 
I think that Sony is getting help from Nvidia because Sony knows it does not have the graphics expertise. the ability to have all the needed hardwired functions. or the ease of development (a secondary concern for Sony, a top concern for developers). a Sony GPU, while no doubt massively powerful, could have lacked some of the functions that even NV2A in Xbox has. Sony does not want to be embarrased with PS3 in any way. they want PS3 to be 'perfect'. Sony knows that Xenon and Revolution could have nicer looking graphics than PS3, even if PS3 was more powerful. Sony needs Nvidia and Nvidia needs Sony. that's what it comes down to, IMO.
 
Another thing that I was thinking of with this announcement is how it kind of goes against the Cell paradigm. From what I understand each Cell is pretty much exactly the same and the power comes from putting a bunch of them together and having them run in parallel. Want more power? Just add some more Cells. The point is that it's all uniform, it's this array/grid of Cell chips, but now in PS3 you're going to have this grid, and then this anamolous GPU stuck in there somewhere. It just doesn't feel natural and conducive to the whole Cell paradigm.
 
rastex said:
Another thing that I was thinking of with this announcement is how it kind of goes against the Cell paradigm. From what I understand each Cell is pretty much exactly the same and the power comes from putting a bunch of them together and having them run in parallel. Want more power? Just add some more Cells. The point is that it's all uniform, it's this array/grid of Cell chips, but now in PS3 you're going to have this grid, and then this anamolous GPU stuck in there somewhere. It just doesn't feel natural and conducive to the whole Cell paradigm.

The GPU will be seperate from the CPU. If you want to make the CPU faster, you can just keep adding PEs (until you reach the bounds of feasibility, of course) - sure, you may not be able to do that with the GPU, but the CPU will be perfectly in line with the Cell paradigm.

Plus, we don't know yet what architecture the GPU will use. I'm not saying it'll be cell-based - I doubt that myself - but I wouldn't rule it out completely yet.
 
If you think sony has been working with nvidia on a new GPU for two years, keep dreaming. There's no way that they could have received money from Sony or spent r&d money on this without notifying stockholders. The conrcrete deal is much likely very recent and it's not going to be any more custom than the Xbox gpu was.
 
gofreak said:
The GPU will be seperate from the CPU. If you want to make the CPU faster, you can just keep adding PEs (until you reach the bounds of feasibility, of course) - sure, you may not be able to do that with the GPU, but the CPU will be perfectly in line with the Cell paradigm.

Plus, we don't know yet what architecture the GPU will use. I'm not saying it'll be cell-based - I doubt that myself - but I wouldn't rule it out completely yet.

My point is in communication. The way all the Cells communicate and share data with each other is very unified. For the GPU to exist in this situation it's either going to be outside, or inside the Cell network. If it's outside then you introduce another line of communication problems (bottleneck). If it's inside then nVidia/Sony will essentially have to wrap the GPU with it's Cell tech, in which case what are they really gaining from using the nVidia GPU if their Cell chips are really as powerful as specified?

Just remember, when the Cell specs were first released everybody was doing calculations on the polygonal pushing power of these chips and rejoicing in that. But now that the GPU is going to be there, what happened to all that power? Looking at the Xbox, it's CPU is a friggin Pentium (Celeron?) 733MHz, which was pretty far behind the curve when it was released, but it was the graphics chip that made it really stand out. I mean, with that puny 733 chip you have Halo 2 doing some pretty complicated Havok physics calculations that are on par with Half Life 2. I mean if you really look at what a more powerful CPU offers with people specifying AI and Physics, well those are two areas that Halo 2 compares, if not surpasses (AI) Half Life 2. If you can follow me it's that in the current game programming paradigm, the CPU doesn't make that much of a difference when you have an uber-powerful GPU. If PS3 is going to have an uber-powerful GPU what's all the hype about it's CPU processing for? It's all very suspect.


Personally, I can't wait for companies to come out with PPU and IPU (Physics Processing Units, and Intellegence Processing Units) Of course maybe that's the whole point with this multiprocessor approach.
 
"Personally, I can't wait for companies to come out with PPU and IPU (Physics Processing Units, and Intellegence Processing Units) Of course maybe that's the whole point with this multiprocessor approach."

Arn't there rumours over on B3D of a 3rd chip in Xbox 2 that nothing to do with GPU or CPU?
 
rastex said:
My point is in communication. The way all the Cells communicate and share data with each other is very unified. For the GPU to exist in this situation it's either going to be outside, or inside the Cell network. If it's outside then you introduce another line of communication problems (bottleneck).

The interface will certainly be interesting. As I understand it, though, the PUs split the work up into Apulets, for the APUs to work on, and the receives back the results. Those results are combined if necessary - presumably they could be passed on to the GPU as simply data. In other words, the PU would act as the crossover point between data/instructions formatted for the "outside" and data/instructions formatted for Cell. Though I'm not sure..perhaps Pana could elaborate.

rastex said:
If it's inside then nVidia/Sony will essentially have to wrap the GPU with it's Cell tech, in which case what are they really gaining from using the nVidia GPU if their Cell chips are really as powerful as specified?

More specific graphics performance? Something we don't know about? ;) Software tools? A more familiar environment for developers?

rastex said:
Just remember, when the Cell specs were first released everybody was doing calculations on the polygonal pushing power of these chips and rejoicing in that. But now that the GPU is going to be there, what happened to all that power?

It hasn't necessarily gone anywhere. The CPU could still be used for vertex processing, but if not, it just leaves more power for everything else. And trust me, you can never have enough power for physics/AI etc..HL2 etc. certainly are not the limit! The graphics next gen will be so sophisticated..it'll have to be backed up by ever more accurate physical models etc, or the "reality" of the game will be lost.
 
Sometimes I think Sony overestimates the importance of a CPU in a dedicated gaming machine. Unless of course it ISN'T a dedicated gaming machine, but in fact more of a set top box.
 
More specific graphics performance? Something we don't know about? ;) Software tools? A more familiar environment for developers?

More pie-in-the-sky speculation and hype? Anyway, you've actually touched upon the concern I'm going with this, and that's how introducing this new different chip will actually make the dev environment even MORE complicated. It's obvious that Cell is a centerpiece of new tech that's going to exist in many many devices, so IBM will obviously invest a lot of time in making its compilers very optimized and straightforward to work with its Cell architecture. Now you introduce this new GPU which may or may not follow the Cell philosophy and it's going to seriously complicate/muddy your compiler creation since you'll have to take into account this new type of device. OR what's even scarier is that you don't incorporate this tech into the compiler and instead pass controlling the device directly onto the devs which could be a major pain in the ass and harkens back to PS2's DMA chains.

In other words, the PU would act as the crossover point between data/instructions formatted for the "outside" and data/instructions formatted for Cell.

If that's the case, then that's just extra processing overhead required for every single frame that is displayed. Obviously you're just guessing at a solution here, but this is my point, is that introducing this GPU causes problems, so the question is WHY would Sony do this if Cell was all it's cracked up to be?
 
rastex said:
More pie-in-the-sky speculation and hype?

The "something we don't know about" is, of course, and that's why I added a ;) The rest is obvious.



rastex said:
Anyway, you've actually touched upon the concern I'm going with this, and that's how introducing this new different chip will actually make the dev environment even MORE complicated. It's obvious that Cell is a centerpiece of new tech that's going to exist in many many devices, so IBM will obviously invest a lot of time in making its compilers very optimized and straightforward to work with its Cell architecture. Now you introduce this new GPU which may or may not follow the Cell philosophy and it's going to seriously complicate/muddy your compiler creation since you'll have to take into account this new type of device. OR what's even scarier is that you don't incorporate this tech into the compiler and instead pass controlling the device directly onto the devs which could be a major pain in the ass and harkens back to PS2's DMA chains.

I don't feel qualified to answer this, but I will point out that we don't know yet whether or not it's cell based. Assuming it's not, I'm presuming/hoping Sony knows what it's doing in integrating a different chip with cell from a dev tool/compiler point of view.


rastex said:
If that's the case, then that's just extra processing overhead required for every single frame that is displayed. Obviously you're just guessing at a solution here, but this is my point, is that introducing this GPU causes problems, so the question is WHY would Sony do this if Cell was all it's cracked up to be?

The PUs won't be the stressed part of the system (the APUs are the workhorses), I think, so they could possibly/probably accomodate that kind of overhead. Again, though, I think Pana could give you a more assured answer on this.

I should point out again that everything being said here, optimistic and pessimistic, is obviously speculation.
 
gofreak said:
I should point out again that everything being said here, optimistic and pessimistic, is obviously speculation.

Exactly, so I hope you won't be offended if I choose to duck out of the discussion here. The reason I'm being so skeptical is that all this rampant speculation when it leads to "SUCK IT MS!" and "What will Nintendo/MS do now??" "PS3 will destroy EVERYTHING" is actually rather destructive to gaming as a whole and I want to counteract that.
 
It was planned from the beginning to have some kind of GPU in PS3. The Visualizer from the patents was cell based, but it was completly different then the main Cell CPU. Many GPU related task simple can't be done fast enough in a software only situation running on a Cell CPU (or any other CPU). Now many people had the idea that Sony would do the GPU (Visualizer) part and that it would be a fill rate monster, but without many hardwired features, but that's not the case. Sony is developing a GPU together with Nvidia, a completly new architecture, cell based or not, but it will be a GPU that combines the ideas Sony had from the beginning with the knowledge of Nvidia to create a GPU that is much better balanced then just a fill rate monster. I really don't know how people could think this is a last minute decision. Two and a half years ago was the time when Sony knew exactly what Cell would be and how a GPU has to work to be a perfect companion, that was the time they've searched and found a perfect partner. Why Nvidia? If they intended to not only have fill rate, then they needed a partner that could help with the features and they needed a partner that was willing to build something completly new based on their and Sonys knowledge. I'm sure they not only asked Nvidia, you can practicaly bet they asked ATI as well, but for whatever reason, Nvidia was the better choice.

Fredi
 
McFly said:
I'm sure they not only asked Nvidia, you can practicaly bet they asked ATI as well, but for whatever reason, Nvidia was the better choice.

Probably because nVidia low-balled to keep from getting shut out of the console market altogether by ATI.
 
MightyHedgehog said:
Probably because nVidia low-balled to keep from getting shut out of the console market altogether by ATI.

It's actualy possible that this deal was done before the MS/ATI deal. Not that MS would have chosen Nvidia again (or that Nvidia was interested to work again with MS), but if the two years of Sony/Nvidia codevelopment is true (why should they lie?), then this deal would have be done months before the MS/ATI deal was official.

Fredi
 
McFly said:
It's actualy possible that this deal was done before the MS/ATI deal. Not that MS would have chosen Nvidia again (or that Nvidia was interested to work again with MS), but if the two years of Sony/Nvidia codevelopment is true (why should they lie?), then this deal would have be done months before the MS/ATI deal was official.

Fredi

I doubt the deal was done 2 years ago. I'd say they were talking two years ago.

So I'm sure in configuring whatever Cell chip is being used for PS3's cpu, they left it open to accomodate the various options they were evaluating (including a NVidia chip). It'll be interesting to see how it all fits together though, that's for sure.
 
Isn't it possible that, even with a powerful CPU like the Cell, Sony realized that it is still the best use of resources to pair it with a cutting edge dedicated GPU that supports lots of forward thinking features in hardware? Doing everything in software just because you can doesn't sound like the best way to leverage Cell.

Combine that with the fact that NVidia's chief service is to design GPUs for others to manufacture, it makes sense. We should all be happy about this regardless of the Cell's capabilities.
 
mashoutposse said:
Isn't it possible that, even with a powerful CPU like the Cell, Sony realized that it is still the best use of resources to pair it with a cutting edge dedicated GPU that supports lots of forward thinking features in hardware? Doing everything in software just because you can doesn't sound like the best way to leverage Cell.

Well, it'll make developers' jobs easier because you're not forcing some paradigm shift on them. Sony's got more to lose than anyone at this point, so you definitely do not want to make yourself any less attractive than the competition.
 
gofreak said:
I doubt the deal was done 2 years ago. I'd say they were talking two years ago.

Why? I know for sure they where already talking for much, much longer. Maybe not about a colaboration, but they where talking about various gfx tech related things for many, many years. It's not like Sony ignored the rest of the industry completly before this deal.

Fredi
 
McFly said:
Why? I know for sure they where already talking for much, much longer. Maybe not about a colaboration, but they where talking about various gfx tech related things for many, many years. It's not like Sony ignored the rest of the industry completly before this deal.

Fredi

At most Sony may have held a carrot in front of them and asked them to make a proposal that they could evaluate verus whatever Toshiba might have been working on etc. i.e. asked them to submit a bid. I doubt they've been working on the chip together full tilt for the last 2 years.

Other NVidia people are also saying they started working together 18 months ago..

Who knows, maybe we'll have a better indication of things at a point in the future.
 
I know absolutely nothing about this "collaboration", but I have a feeling that it's mainly patent related, and that NVidia is probably doing very little of the work.

Nothing!
 
He's not known as a lier, but take it as a rumor for now:

Hey sorry guys, i have not posted in so long since last time. But anyways i am sorry i couldn't say anything about this sweet deal until it was announced since i was under very tight NDA's. Anyways, now it's official, and i would like to say that this is a co-joint collaboration between sony in-house graphics technology and Nvidia technology. What this means is that this "custom-gpu" is a totally different architecture that is not based on an existing architecture from Nvidia.

http://psinext.com/forums/viewtopic.php?p=84712#84712

Fredi
 
So, if true, Dave would be proven wrong and the new GPU actually IS something completely new by Sony AND Nvidia, not just nVidia. Can't wait for this beast!
 
Ok First:

In before Pana.



Ok, secondly: I think the only problem Sony had was from a graphical standpoint.
They needed people on the shaders. The Ps2 has no hardware shaders. Most of the time you see shaders on the ps2, it's done by software. The job listing Pana posted shows this. Sony has a hard time with the shaders.
 
cybamerc said:
You don't get a PC centric development environment just by having a PC graphics chip (and I suspect this one will be heavily modified).

PC developers will still go to M$ who they know will offer money and development assistance.

What would happen if they gave developers low level access as well as a high level access to graphics processing (a high level shading language and a shareable high level library maybe OpenGL based ;)) as well as tools and libs for CELL (IBM said that they will provide tools to help developers with the learning curve associated with parallel processing systems and the SPUs should be able to be programmed using high level code that gets compiled for them as Gschwind's IBM web page says in reference to his involvement in CELL as the chief architect [with guys like Peter Hofstee] of the SPU/APU architecture) ?

I would like having the ability to choose between two competent solutions to tackle PlayStation 3 games programming.
 
You can look at this situation from two extremes...

1) Sony has planned this from the very beginning and will be able to seemlessly integrate Nvidia's technology into the PS3 to create the powerhouse cell-based console that they're hoping to unleash on the world.

2) Sony overestimated their capabilities to deliver, especially on the cell and GPU fronts, and thus were forced to scrap some of their original projections and plans and opt to go with a more over-the-counter solution at the last minute and bring in Nvidia's expertise to salvage the cause.

I imagine the truth is somewhere in between those two extremes.
 
Now that Pana is finally here, I have a question.
Why did Sony go with nVidia and not ATI?
The past year or so, ATI have really started to own the PC GFX card market, and it doesn't look like being the fleeting lead position that Diamond Stealth and nVidia once had either, they could be at the top for some time. So why did Sony pick the #2 GFX card maker?

I don't mean to insinuate anything, by the way. It could be that nVidia has gone behind in the PC market because they are putting a lot of effort into making something that blows ATI away. Just wondering if anyone had a proper idea.
 
ATi's contract may have prohibited them from collaborating with Sony or they may have found it pretty much impossible to work with Sony without violating their NDAs with MS or maybe they even didn't have enough workforce to develop for Sony in addition to Nintendo, MS as well as graphics and chipsets for PCs and mobiles.. ATi do have their hands full these days
 
Top Bottom