CONFIRMED: PS3 to use "Nvidia-based Graphics processor&

[Qroach]

That's what I've been thinking to some extent. I do beleive ATI was also in the running, but didn't go down that route or wasn't picked for some reason.

Anyway, I just don't beleive that nvidia had any real significant input in what sony was doing without a mention of his to investors. Even if it was only consulting. I'm sure Nvidia has been talking to sony the past two years, or longer, but I don't think they started work on anything for them without a hint of money transfered. I was a damn shareholder for over a year and that is teh type of information I would want to know about.

 

[McFly]

What if Ati was also "running" for the position?

I'm actualy sure they did. And it could very well be that the Nvidia/Sony deal was made before the MS/ATI deal.

Fredi

 

[London Geezer]

Well maybe Ati has already a bit too much to worry about, with their next gen PC chips, the Xbox2 GPU and apparently the next Nintendo GPU, which are all supposed to be somewhat different... And another new architecture to couple with the Cell CPU might have stretched them a bit too much...
But i do have the feeling they at least had a few meetings down the line.
I don't see why Sony would pick Nvidia without checking what Ati had to offer too.

 

[PC-Engine]

SONY likes to be special so they wouldn't want graphics technology used in other consoles in their own. They would rather use thier own tech however they realized using beefed up Voodoo tech wasn't gonna cut it so they smartly chose Nvidia as last ditch effort.

 

[London Geezer]

*Imagines Sony in their Simon Cowell outfit, lining up representatives from S3, Matrox, PVR, NVIDIA, ATI against the wall, going "Mmmm, No... No... Maybe... Yes... Yes..."*




GPU IDOL!!!!

Oh my god i had too much coffee.......

 

[j^aws]

Sony has been working on CELL for more than double that amount of time ... if they had intended to work with NVIDIA from the start, they most likely would have began working with them sooner.....

From that patent you can see, that they had two type of cell chip in mind, one for data plane processing, the other for graphics processing.

On this board, people always question who are going to design that part, specifically the one with pixel engine. At first it was assumed it was just going to be another Sony graphic chips, some more patents searches gave Toshiba as possibility too. Also rumours from various and questionable sources points to NVIDIA. These rumours been around for quite sometimes too.

On this board, this was once the achilles heels for PS3 because Sony was assume to be doing it alone. Now its official that NVIDIA got the deal. So I wonder where they'll poke next

Now the next interesting question is, how does the Power core and Synergistic core from STI, help NVIDIA in designing the the graphic processing cell, if at all. That's the interesting bit.

Say that the GPU that was racing nVIDIA's solution as the GPU for PlayStation 3 was aa Pixel Shading only part, that the plan was still doing all VS work on the CELL based CPU and all the PS work on the GPU.

I am saying... not that I know anything at all .

Wouldn't you at least wonder why, if the GPU were CELL based like the Visualizer why then that GPU was doing only PS work and it could not share VS work at all with the CPU ?

I mean... Apulets can migrate and should be processable by any CELL based chip with enough APUs as the Apulet specifies (the PEs in the Visualizer were supposed to have 4 APUs each).

I know you lurking PS2 devs are under NDA's!

CPU= 32bit vertex engines
GPU= 32bit pixel engines

And IIRC, S|APUs dont have to be 4-way SIMD based for CELL to work with Apulets (Software Cells). The patents only say preferrable...

Am I hot or Am I cold!

APUs were supposed to have the same ISA all across and if you want to call the GPU CELL based, if you want your GPU to be CELL based is because you do want to share the workload dynbamically between CPU and GPU.

Why would a CELL based GPU only do Pixel Shading work and could not share Vertex Shading workload ?

I mentioned this in another thread of mine with salc/salps...and pixel engines...

Both CPU and GPU can do vertex and pixel shading. However, the CPU is optimised for vertex shading and the GPU is optimised for pixel shading. They can also exchange work if necessary. You have two types of execution units with two degrees of granularity. Sony are supplying the vertex side and nVidia are supplying the pixel side. However the CPU+GPU combination is CELL based as they can operate on Apulets (Software Cells). Hot or Cold!

 

[one]

Well maybe Ati has already a bit too much to worry about, with their next gen PC chips, the Xbox2 GPU and apparently the next Nintendo GPU, which are all supposed to be somewhat different... And another new architecture to couple with the Cell CPU might have stretched them a bit too much...
But i do have the feeling they at least had a few meetings down the line.
I don't see why Sony would pick Nvidia without checking what Ati had to offer too.

nVIDIA can offer the software environment (plus closely-tied engines from 3rd parties, such as UE3). They might offer lower, competitive price compared to ATi, too.

 

[London Geezer]

Well maybe Ati has already a bit too much to worry about, with their next gen PC chips, the Xbox2 GPU and apparently the next Nintendo GPU, which are all supposed to be somewhat different... And another new architecture to couple with the Cell CPU might have stretched them a bit too much...
But i do have the feeling they at least had a few meetings down the line.
I don't see why Sony would pick Nvidia without checking what Ati had to offer too.

nVIDIA can offer the software environment (plus closely-tied engines from 3rd parties, such as UE3). They might offer lower, competitive price compared to ATi, too.

Oh i'm sure, and lots of pre-written graphics libraries and shaders the devs can cut and paste on their games! Kinda like Wreckless on the Xbox!! YAY!! j/k

 

[Dave Baumann]

I'm actualy sure they did. And it could very well be that the Nvidia/Sony deal was made before the MS/ATI deal.

I believe the actual deal was done very recently. I'd picked up on the talk some months back and tapped up a few developer comments; I mentioned these to our local NVIDIA rep at a recent awards presentation (knowing there would be suitable quantities of alcholo around! ) and he pretty much explicitly said "as of last week it was still internal".

Talks between Sony and other vendors had been ongoing for sometime, but the finalisation of the deal was likely very recent. Speaking to an analyst yesterday it seems that Sony's approach was "yes, we're listen to others, but at present thats not our preferred route and there's no money on the table yet"; that wasn't particulary attractive to ATI when they already had two console projects on the boil - it appears they did drop the ball on following the progress of the preffered solution and getting in there at the right time. Given that NVIDIA were out of the other two consoles they followed Sony's progress much closer.

 

[Dio]

...still, if next gen games are mostly beat'em-ups, Racing games, fps... is there really that much content that they need?

Remember that the current generation of consoles is far behind the PC space in terms of capability, and those console games still take years to write now.

PS2 and X-box games don't have pixel shaders (1.1 don't really count, there's hardly anything interesting you can do with them). Shaders cause every surface in Doom3 to need 3 textures instead of one - and Doom3 is using a single surface type.

There are algorithms being used now that happily consume 8 texturesworth of parameterisation per surface - and somebody has to draw all that. Someone has to develop the mathematical model for each surface type... the list goes on.

And this is just graphics. The really difficult and time-consuming stuff (from a coding and particularly testing point of view) like AI, physics, networked gaming, etc. are expected to jump ahead an order of magnitude in capability as well.

So take Doom3 or Half-Life 2's development time, and layer on compulsory netcode and platform approval. And then double it, if the specs of the next generation consoles are going to live up to the hype.

How long's that?

 

[London Geezer]

Those sneaky peeps. It seems NVIDIA got over the NV3X failure (now THAT was a failure) pretty quickly...

 

[Panajev2001a]

Sony has been working on CELL for more than double that amount of time ... if they had intended to work with NVIDIA from the start, they most likely would have began working with them sooner.....

From that patent you can see, that they had two type of cell chip in mind, one for data plane processing, the other for graphics processing.

On this board, people always question who are going to design that part, specifically the one with pixel engine. At first it was assumed it was just going to be another Sony graphic chips, some more patents searches gave Toshiba as possibility too. Also rumours from various and questionable sources points to NVIDIA. These rumours been around for quite sometimes too.

On this board, this was once the achilles heels for PS3 because Sony was assume to be doing it alone. Now its official that NVIDIA got the deal. So I wonder where they'll poke next

Now the next interesting question is, how does the Power core and Synergistic core from STI, help NVIDIA in designing the the graphic processing cell, if at all. That's the interesting bit.

Say that the GPU that was racing nVIDIA's solution as the GPU for PlayStation 3 was aa Pixel Shading only part, that the plan was still doing all VS work on the CELL based CPU and all the PS work on the GPU.

I am saying... not that I know anything at all .

Wouldn't you at least wonder why, if the GPU were CELL based like the Visualizer why then that GPU was doing only PS work and it could not share VS work at all with the CPU ?

I mean... Apulets can migrate and should be processable by any CELL based chip with enough APUs as the Apulet specifies (the PEs in the Visualizer were supposed to have 4 APUs each).

I know you lurking PS2 devs are under NDA's!

CPU= 32bit vertex engines
GPU= 32bit pixel engines

And IIRC, S|APUs dont have to be 4-way SIMD based for CELL to work with Apulets (Software Cells). The patents only say preferrable...

Am I hot or Am I cold!

APUs were supposed to have the same ISA all across and if you want to call the GPU CELL based, if you want your GPU to be CELL based is because you do want to share the workload dynbamically between CPU and GPU.

Why would a CELL based GPU only do Pixel Shading work and could not share Vertex Shading workload ?

I mentioned this in another thread of mine with salc/salps...and pixel engines...

Both CPU and GPU can do vertex and pixel shading. However, the CPU is optimised for vertex shading and the GPU is optimised for pixel shading. They can also exchange work if necessary. You have two types of execution units with two degrees of granularity. Sony are supplying the vertex side and nVidia are supplying the pixel side. However the CPU+GPU combination is CELL based as they can operate on Apulets (Software Cells). Hot or Cold!

I repeat the GPU that I am talking about was PS only (could not exhancge VS work with the GPU besides writing to Vertex Buffers from the Pixel Shaders, but I do not know if you could do that)... Birds say that it was handled sort of internally... well by one of their existing partners.

 

[Panajev2001a]

I'm actualy sure they did. And it could very well be that the Nvidia/Sony deal was made before the MS/ATI deal.

I believe the actual deal was done very recently. I'd picked up on the talk some months back and tapped up a few developer comments; I mentioned these to our local NVIDIA rep at a recent awards presentation (knowing there would be suitable quantities of alcholo around! ) and he pretty much explicitly said "as of last week it was still internal".

Talks between Sony and other vendors had been ongoing for sometime, but the finalisation of the deal was likely very recent. Speaking to an analyst yesterday it seems that Sony's approach was "yes, we're listen to others, but at present thats not our preferred route and there's no money on the table yet"; that wasn't particulary attractive to ATI when they already had two console projects on the boil - it appears they did drop the ball on following the progress of the preffered solution and getting in there at the right time. Given that NVIDIA were out of the other two consoles they followed Sony's progress much closer.

Your argument makes a lot of sense and it does work well with what the birds chirp out of my window.

The only problem is that they keep me awake with all their chirping... oh well... cute birds, I like their songs .

 

[London Geezer]

Your argument makes a lot of sense and it does work well with what the birds chirp out of my window.

The problem is that they keep me awake with all their chirping... oh well... cute birds .

Just feed them a lot, keep them happy so they keep singing!!

 

[j^aws]

Sony has been working on CELL for more than double that amount of time ... if they had intended to work with NVIDIA from the start, they most likely would have began working with them sooner.....

From that patent you can see, that they had two type of cell chip in mind, one for data plane processing, the other for graphics processing.

On this board, people always question who are going to design that part, specifically the one with pixel engine. At first it was assumed it was just going to be another Sony graphic chips, some more patents searches gave Toshiba as possibility too. Also rumours from various and questionable sources points to NVIDIA. These rumours been around for quite sometimes too.

On this board, this was once the achilles heels for PS3 because Sony was assume to be doing it alone. Now its official that NVIDIA got the deal. So I wonder where they'll poke next

Now the next interesting question is, how does the Power core and Synergistic core from STI, help NVIDIA in designing the the graphic processing cell, if at all. That's the interesting bit.

Say that the GPU that was racing nVIDIA's solution as the GPU for PlayStation 3 was aa Pixel Shading only part, that the plan was still doing all VS work on the CELL based CPU and all the PS work on the GPU.

I am saying... not that I know anything at all .

Wouldn't you at least wonder why, if the GPU were CELL based like the Visualizer why then that GPU was doing only PS work and it could not share VS work at all with the CPU ?

I mean... Apulets can migrate and should be processable by any CELL based chip with enough APUs as the Apulet specifies (the PEs in the Visualizer were supposed to have 4 APUs each).

I know you lurking PS2 devs are under NDA's!

CPU= 32bit vertex engines
GPU= 32bit pixel engines

And IIRC, S|APUs dont have to be 4-way SIMD based for CELL to work with Apulets (Software Cells). The patents only say preferrable...

Am I hot or Am I cold!

APUs were supposed to have the same ISA all across and if you want to call the GPU CELL based, if you want your GPU to be CELL based is because you do want to share the workload dynbamically between CPU and GPU.

Why would a CELL based GPU only do Pixel Shading work and could not share Vertex Shading workload ?

I mentioned this in another thread of mine with salc/salps...and pixel engines...

Both CPU and GPU can do vertex and pixel shading. However, the CPU is optimised for vertex shading and the GPU is optimised for pixel shading. They can also exchange work if necessary. You have two types of execution units with two degrees of granularity. Sony are supplying the vertex side and nVidia are supplying the pixel side. However the CPU+GPU combination is CELL based as they can operate on Apulets (Software Cells). Hot or Cold!

I repeat the GPU that I am talking about was PS only (could not exhancge VS work with the GPU besides writing to Vertex Buffers from the Pixel Shaders, but I do not know if you could do that)... Birds say that it was handled sort of internally... well by one of their existing partners.

This is just an analogy below,

CPU= 32 S|APUs (vertex shading)
GPU= 32 S|APUs (pixel shading)

CPU<=>GPU--->output

Replace 32 with whatever number you think but there's a 1:1 mapping. And call the S|APUs whatever depending on whether VS or PS units. And both CPU and GPU will be classed as CELL based as they should be able to execute Apulets (Software Cells). That's what I'm leaning towards...

 

[function]

Talks between Sony and other vendors had been ongoing for sometime, but the finalisation of the deal was likely very recent. Speaking to an analyst yesterday it seems that Sony's approach was "yes, we're listen to others, but at present thats not our preferred route and there's no money on the table yet"; that wasn't particulary attractive to ATI when they already had two console projects on the boil - it appears they did drop the ball on following the progress of the preffered solution and getting in there at the right time. Given that NVIDIA were out of the other two consoles they followed Sony's progress much closer.

Is there any speculation in these circles that if Sony had known their in house solution was not suitable sooner, they might have looked at licencing ATI technology before they were tied up with the R500 for Microsoft? Or is that speculation too far, even for the speculators?

 

[Jov]

Wow I can see those Sega fans turned Xbox worshippers preaching rumours like there's no tomorrow.I mean come on you guys really are far sighted.Nvidia has just announced(revealed actually) a collaboration with SCE and now you're already churning out your negative statements about their plans as though you know what what their plans are.

The best part is they have to use the Xbox as sacrificial ham.

 

[V3]

This is just an analogy below,

CPU= 32 S|APUs (vertex shading)
GPU= 32 S|APUs (pixel shading)

CPU<=>GPU--->output

Replace 32 with whatever number you think but there's a 1:1 mapping. And call the S|APUs whatever depending on whether VS or PS units. And both CPU and GPU will be classed as CELL based as they should be able to execute Apulets (Software Cells). That's what I'm leaning towards...

APUs for fragment shading is over kill, fragment shading unit is cheap to implement in term of silicon estate and could be useful for post processing. You don't want to waste APUs for that.

Maybe what you want is to unify the rendering pipeline, shade everything in micropolygon style, instead of treating vertex and pixel seperately.

Not like the way ATI solution for Xbox next, where it is just for load balancing. And the rendering pipeline still similar to current solution.

 

[j^aws]

This is just an analogy below,

CPU= 32 S|APUs (vertex shading)
GPU= 32 S|APUs (pixel shading)

CPU<=>GPU--->output

Replace 32 with whatever number you think but there's a 1:1 mapping. And call the S|APUs whatever depending on whether VS or PS units. And both CPU and GPU will be classed as CELL based as they should be able to execute Apulets (Software Cells). That's what I'm leaning towards...

APUs for fragment shading is over kill, fragment shading unit is cheap to implement in term of silicon estate and could be useful for post processing. You don't want to waste APUs for that.

Maybe what you want is to unify the rendering pipeline, shade everything in micropolygon style, instead of treating vertex and pixel seperately.

Not like the way ATI solution for Xbox next, where it is just for load balancing. And the rendering pipeline still similar to current solution.

I said it was an analogy!

These PS S|APUs are specilized for the GPU. Not like your VS S|APUs on the CPU. What are the differences, well this is what I think nVidia is doing!

They could even be Salc/Salps and do not even have to be 4way SIMD units or they could be other pixel engine type units from nVidia or Sony.

And I agree with your micro-polygons shading style which is why I think this is likely,

http://www.beyond3d.com/forum/viewtopic.php?t=18849

 

[Nite_Hawk]

I'm actualy sure they did. And it could very well be that the Nvidia/Sony deal was made before the MS/ATI deal.

I believe the actual deal was done very recently. I'd picked up on the talk some months back and tapped up a few developer comments; I mentioned these to our local NVIDIA rep at a recent awards presentation (knowing there would be suitable quantities of alcholo around! ) and he pretty much explicitly said "as of last week it was still internal".

Talks between Sony and other vendors had been ongoing for sometime, but the finalisation of the deal was likely very recent. Speaking to an analyst yesterday it seems that Sony's approach was "yes, we're listen to others, but at present thats not our preferred route and there's no money on the table yet"; that wasn't particulary attractive to ATI when they already had two console projects on the boil - it appears they did drop the ball on following the progress of the preffered solution and getting in there at the right time. Given that NVIDIA were out of the other two consoles they followed Sony's progress much closer.

What do you think about the 18 months comment? Do you think perhaps that's just how much time they had been working on their next generation gpu anyway, and they can use it as a convenient "been working on this with sony for 18mo" excuse?

I'm just curious why you seem so convinced they missed their performance target. Just from the general information available, that seems like one possible explanation for going with nVidia, but certainly not the only one.

Nite_Hawk