will PS3's GPU be more modern than PS2's GS for its time?

[Panajev2001a]

My hunch is NV in dire need to replace the loss of income from failing to get Xbox2, most likely outbid Toshiba, thus making PS3 or other cell device cheaper for Sony to produce.

that would have been my second quess ; )

My question is this: why cannot we maybe accept that nVIDIA's GPU is also the more featured and faster of the two ?

no.. because italian wine and pasta is just so darned delish!

Mamma-miaaaaa..... Singh-ah... you know in the fAmily, you see the Pope-ah and eating past-ah while playing du mandulin-ah.

.

No, that is not how I would talk to "Teflon" Ken .

 

[V3]

No... that is still not it... you are assuming that the internal solution was the one they set aside when they finally gave the contract to nVIDIA.

Panajev2001a, I only assume that Sony went with cheaper solution, and they end up going with with NV. So NV is the cheaper solution.

In the past Sony always goes with cheaper solution too. Sony is cost concious. They'll do it internally or able to partner someone other than NV, if it was cheaper. NV just outbid everyone and won the contract. Sony would not go with a more expensive solutions just for the sake of having superior pixel engine compare to competitors, that's for sure. They would just get the cheapest pixel engine available, internally or externally.

 

[Inane_Dork]

Panajev2001a, I only assume that Sony went with cheaper solution, and they end up going with with NV. So NV is the cheaper solution.

In the past Sony always goes with cheaper solution too. Sony is cost concious. They'll do it internally or able to partner someone other than NV, if it was cheaper. NV just outbid everyone and won the contract. Sony would not go with a more expensive solutions just for the sake of having superior pixel engine compare to competitors, that's for sure. They would just get the cheapest pixel engine available, internally or externally.

But that's only true to an extent. They could have used multiple GS's for less money, I'll bet. But that solution was unreasonable for feature set.

 

[Panajev2001a]

I guess it will all come out in the wash. however I don't think sony is really saving any money on graphics tech, based on all the "cell taking a revoloutionary approach to rendering graphics" nonesense people around here have been muttering for months.

Qroach, if Sony ever going to bite a bullet, they would go with cheaper technology not an expensive one. In that way they may had more inferior pixel engine. That's Sony biting the bullets. Not MS style, like you thought.

We don't know anything about Toshiba solution (or NV), yet you thought of their solutions to be inferior because Sony went with NV instead. Sony will go with a cheaper solution. If it is better than its a plus isn't ?

It is wrong on many levels this topic of "Sony wanted the cheapest solution".

It is really misleading and I am glad Vince caught the the argument by the head and understood the need of clarifying this.

If you normalize the relative preformance to be equivalent, it would take a larger amount of capital to reach that static level with the internal Sony or Toshiba designs. Or, said in other words, the external design is cheaper.

I think we need to read these lines again as they clear the confusion we are having here.

Time and money could bring a design from Toshiba or an internal one to be in parity or be superior to the design nVIDIA is proposing.

The problem is that time is not something they have an unlimited resource of and it is getting even shorter than money now: the console has a certain release date they have always planned to hit and miss it by 6 months to go with an internal or semi-internal solution is not an option as that would give MS too many months being the first next-generation console on the market.

If Toshiba would have had a fully PS 3.xx/4.xx by December 2006 then it is kinda worthless if the console must ship before that date.

Also, if money was no problem (meaning that they have an unlimited supply of money) they could simply put more Toshiba GPU cores there tille the nVIDIA solution is clearly outperformed and by far, but Sony/SCE does not have an infinite budget.

I think performance is one of the key that pushed the contract to nVIDIA, but I think that a lot of pressure is coming on the software side of things (PlayStation 3 SDK, GPU related tools, etc...) which I think nVIDIA was more prepared to provide than Toshiba GPU-wise (nVIDIA also has a very nice and estabilished Development Relations group and a lot of experience with PC developers [which I think SCE wants to bring in as well and not just leave all of them to the Xbox 2-side] and standard APIs like OpenGL) as well as the money side.

It is strange that Toshiba seems to have been caught off-guard about this announcement, a decision made recently (this does not mean that nVIDIA had not been following PlayStation 3's development and hadn't asked questions, etc...) must mean that nVIDIA made some very BIG and persuasive argument to SCE's management that changed their mind to the point of changing the contractor for the GPU.

I do not think that the GPU was supposed to be CELL even when they decided to go with the Toshiba GPU as well as now that they are going with the nVIDIA design.

This is not a loss for CELL, nobody said that CELL systems were not going to use any other chip: Toshiba is going to use CELL with thei MeP architecture.

I think that architecturally wise there are not many solutions to the perfomance and data flow problems for parallel-architectures in the context of multi-media driven solutions so it is not hard to see IMHO how forward looking architectures such as ATI R5xx, nVIDIA NV5X and S/T/I CELL might share some if not many ideas. I saw an nVIDIA patent not too long ago, it almost looked like a CELL system in one of the diagrams (that is before you started noticing differencies here and there).

With this said, I still say that IMHO the PlayStation 3 GPU is not CELL based, it does not have the SPUs/APUs.

Let's look at the stance ATI and nVIDIA are taking for the short-to-medium term (which includes the solutions that, time-frame wise, could ship in PlayStation 3, Xbox 2/Xenon and Revolution/NES 5).

ATI believes that developing a single unfiied shading unit is the key, that is what the GPU market needs to agree on as the best path to take for optimal performance and flexibility in 3D graphics.

This path leads to a situation in which Shading ALUs and Texture units are decoupled and we have a pool of Shading Units computing data, writing and reading from memory and interacting with a pool of Texture Units that process the requests from the Shading Units pool (separate scheduling and separate resources for each pool).

This is the path that CELL 2.0 might take, maybe in PlayStation 4.

This is the vision you have Vince: unified Shading Hardware (the SPU/APU) connected with a sea of fixed-function/hardwired units providing AA, Texture Sampling/Filtering, etc...

This is not what nVIDIA believs in.

nVIDIA believes that the time has not come yet for having the same Shading ALU hardware do both Vertex Shading work as well as Pixel Shading work, they believe that each unit should be heavvily optimized in a certain area (Vertex Shading or Pixel Shading) as they believe that the two areas, even though the ISA is converging, have still quite different needs that claim different hardware configurations.

I do not see the APUs in the Broadband Engine having instructions optimized to do texture sampling/filtering work or having separate TMU silicon. If the PlayStation 3 GPU ends up being a Pixel Shading only design as the other challenger design was then it means that Sony/SCE has a different view on Vertex Shading that includes versatility, but not features like Vertex Texturing (or at least not an use of Vertex Texturing so heavvy that it would require dedicatd hardware).

Vertex Shading might have been removed (as part of the customization work nVIDIA is doing with SCE on the GPU) and it would then be performed on the CELL based CPU. This path might be interesting as nVIDIA could re-utilize the silicon-space freed for e-DRAM and more Shading ALUs.

From what we hear, it is possible to guess that FP-wise the CELL based CPU in PlayStation 3 is aiming at an over-all higher-peak rating than what the Xbox 2/Xenon CPU is.

It would seem to me a not bad idea to assign all the Vertex Shading work to the CELL based CPU: we do not have tons of idle units if the application does not put much pressure on Vertex Shading (which would happen if we had a fixed number of Shaidng ALUs dedicated to Vertex Shading only) and we have lots of performance on the Pixel Shading side as we have space for more and very optimized Pixel Shading Units.

 

[Panajev2001a]

No... that is still not it... you are assuming that the internal solution was the one they set aside when they finally gave the contract to nVIDIA.

Panajev2001a, I only assume that Sony went with cheaper solution, and they end up going with with NV. So NV is the cheaper solution.

In the past Sony always goes with cheaper solution too. Sony is cost concious. They'll do it internally or able to partner someone other than NV, if it was cheaper. NV just outbid everyone and won the contract. Sony would not go with a more expensive solutions just for the sake of having superior pixel engine compare to competitors, that's for sure. They would just get the cheapest pixel engine available, internally or externally.

I disagree, they are cost conscious, but they also have a clear idea of what their performance and feature targets are which is what engineering is about.

 

[V3]

It is wrong on many levels this topic of "Sony wanted the cheapest solution".

It is really misleading and I am glad Vince caught the the argument by the head and understood the need of clarifying this.

I know they have their performance target they need, (that's why they didn't reuse the old GS), but given options that meet their performance target, they will go with the cheapest option. That's just business.

If they have to bite the bullet, they will lower their performance target instead of increasing cost as suggested by Qroach. You don't increase your cost on something that will sell for $300 and less. Its really stupid, Sega and MS done it, it just doesn't work.

I disagree, they are cost conscious, but they also have a clear idea of what their performance and feature targets are which is what engineering is about.

Engineering is about meeting performance and features target at a given cost. That's engineering.

 

[V3]

But that's only true to an extent. They could have used multiple GS's for less money, I'll bet. But that solution was unreasonable for feature set.

To re engineer GSs and produce Multiple GSs to meet what they wanted aren't exactly cheap for them. And they didn't went with it, they went with NV instead, its just a cheaper solution to Sony.

 

[Panajev2001a]

It is wrong on many levels this topic of "Sony wanted the cheapest solution".

It is really misleading and I am glad Vince caught the the argument by the head and understood the need of clarifying this.

I know they have their performance target they need, (that's why they didn't reuse the old GS), but given options that meet their performance target, they will go with the cheapest option. That's just business.

The other options did not meet their performance+features target IMHO (given the time constraints and the R&D available to fund the project... hey if they could afford 500 mm^2 chips for the GPU I do not doubt their internal solution would have been more than competitive enough ).

 

[V3]

The other options did not meet their performance target (given the time constraints and the R&D available to fund the project... hey if they could afford 500 mm^2 chips for the GPU I do not doubt their internal solution would have been more than competitive enough

That's a speculation Pana Sony has their budget. They won't increase their budget, not on $300 product. They won't take an increase in budget for PS3. They will cut corners to meet that budget. Its just the real world, Pana.

Its nice though if we live in deamland of some sort, where 500 mm^2 chips grow on tree

 

[Megadrive1988]

I guess it will all come out in the wash. however I don't think sony is really saving any money on graphics tech, based on all the "cell taking a revoloutionary approach to rendering graphics" nonesense people around here have been muttering for months.

Qroach, if Sony ever going to bite a bullet, they would go with cheaper technology not an expensive one. In that way they may had more inferior pixel engine. That's Sony biting the bullets. Not MS style, like you thought.

We don't know anything about Toshiba solution (or NV), yet you thought of their solutions to be inferior because Sony went with NV instead. Sony will go with a cheaper solution. If it is better than its a plus isn't ?

It is wrong on many levels this topic of "Sony wanted the cheapest solution".

It is really misleading and I am glad Vince caught the the argument by the head and understood the need of clarifying this.

If you normalize the relative preformance to be equivalent, it would take a larger amount of capital to reach that static level with the internal Sony or Toshiba designs. Or, said in other words, the external design is cheaper.

I think we need to read these lines again as they clear the confusion we are having here.

Time and money could bring a design from Toshiba or an internal one to be in parity or be superior to the design nVIDIA is proposing.

The problem is that time is not something they have an unlimited resource of and it is getting even shorter than money now: the console has a certain release date they have always planned to hit and miss it by 6 months to go with an internal or semi-internal solution is not an option as that would give MS too many months being the first next-generation console on the market.

If Toshiba would have had a fully PS 3.xx/4.xx by December 2006 then it is kinda worthless if the console must ship before that date.

Also, if money was no problem (meaning that they have an unlimited supply of money) they could simply put more Toshiba GPU cores there tille the nVIDIA solution is clearly outperformed and by far, but Sony/SCE does not have an infinite budget.

I think performance is one of the key that pushed the contract to nVIDIA, but I think that a lot of pressure is coming on the software side of things (PlayStation 3 SDK, GPU related tools, etc...) which I think nVIDIA was more prepared to provide than Toshiba GPU-wise (nVIDIA also has a very nice and estabilished Development Relations group and a lot of experience with PC developers [which I think SCE wants to bring in as well and not just leave all of them to the Xbox 2-side] and standard APIs like OpenGL) as well as the money side.

It is strange that Toshiba seems to have been caught off-guard about this announcement, a decision made recently (this does not mean that nVIDIA had not been following PlayStation 3's development and hadn't asked questions, etc...) must mean that nVIDIA made some very BIG and persuasive argument to SCE's management that changed their mind to the point of changing the contractor for the GPU.

I do not think that the GPU was supposed to be CELL even when they decided to go with the Toshiba GPU as well as now that they are going with the nVIDIA design.

This is not a loss for CELL, nobody said that CELL systems were not going to use any other chip: Toshiba is going to use CELL with thei MeP architecture.

I think that architecturally wise there are not many solutions to the perfomance and data flow problems for parallel-architectures in the context of multi-media driven solutions so it is not hard to see IMHO how forward looking architectures such as ATI R5xx, nVIDIA NV5X and S/T/I CELL might share some if not many ideas. I saw an nVIDIA patent not too long ago, it almost looked like a CELL system in one of the diagrams (that is before you started noticing differencies here and there).

With this said, I still say that IMHO the PlayStation 3 GPU is not CELL based, it does not have the SPUs/APUs.

Let's look at the stance ATI and nVIDIA are taking for the short-to-medium term (which includes the solutions that, time-frame wise, could ship in PlayStation 3, Xbox 2/Xenon and Revolution/NES 5).

ATI believes that developing a single unfiied shading unit is the key, that is what the GPU market needs to agree on as the best path to take for optimal performance and flexibility in 3D graphics.

This path leads to a situation in which Shading ALUs and Texture units are decoupled and we have a pool of Shading Units computing data, writing and reading from memory and interacting with a pool of Texture Units that process the requests from the Shading Units pool (separate scheduling and separate resources for each pool).

This is the path that CELL 2.0 might take, maybe in PlayStation 4.

This is the vision you have Vince: unified Shading Hardware (the SPU/APU) connected with a sea of fixed-function/hardwired units providing AA, Texture Sampling/Filtering, etc...

This is not what nVIDIA believs in.

nVIDIA believes that the time has not come yet for having the same Shading ALU hardware do both Vertex Shading work as well as Pixel Shading work, they believe that each unit should be heavvily optimized in a certain area (Vertex Shading or Pixel Shading) as they believe that the two areas, even though the ISA is converging, have still quite different needs that claim different hardware configurations.

I do not see the APUs in the Broadband Engine having instructions optimized to do texture sampling/filtering work or having separate TMU silicon. If the PlayStation 3 GPU ends up being a Pixel Shading only design as the other challenger design was then it means that Sony/SCE has a different view on Vertex Shading that includes versatility, but not features like Vertex Texturing (or at least not an use of Vertex Texturing so heavvy that it would require dedicatd hardware).

Vertex Shading might have been removed (as part of the customization work nVIDIA is doing with SCE on the GPU) and it would then be performed on the CELL based CPU. This path might be interesting as nVIDIA could re-utilize the silicon-space freed for e-DRAM and more Shading ALUs.

From what we hear, it is possible to guess that FP-wise the CELL based CPU in PlayStation 3 is aiming at an over-all higher-peak rating than what the Xbox 2/Xenon CPU is.

It would seem to me a not bad idea to assign all the Vertex Shading work to the CELL based CPU: we do not have tons of idle units if the application does not put much pressure on Vertex Shading (which would happen if we had a fixed number of Shaidng ALUs dedicated to Vertex Shading only) and we have lots of performance on the Pixel Shading side as we have space for more and very optimized Pixel Shading Units.

Panajev, I enjoyed and savored your post like I would a nicely cooked steak. seriously, when i saw this was gonna be a substantial post, i ran downstairs to get a glass of coke to enjoy while I read 8) now don't get me wrong i amm not buttering you up or anything. i just enjoy your posts.

okay, can you re-explain your current thinking on how PS3 is going to render scenes and what PS3 is gonna be all about--are you no longer looking at a REYES style approach? Is PS3 going to be more conventional now, with texture mapping & pixel shading? was your eariler vision for PS3 now moved forward into PS4 or still pretty much the same?

take a look at this pre-rendered CG intro for Everquest 2.

http://www.gametrailers.com/player.php?id=21&type=mov (QuickTime)
http://www.gametrailers.com/player.php?id=21&type=wmv (Windows Media)

the first one-half or one-third of that CG sequence, especially the open landscape, do you feel that will be possible on the CELL-Nvidia based PS3 ?

 

[London Geezer]

now don't get me wrong i amm not buttering you up or anything. i just enjoy your posts.

Hey hey hey! Let's leave the gay innuendos to the expert shall we...?


Anyway, to stay on topic, does anyone other than me think that now that NVIDIA is on board, the chances of PS3 having a "different" approach to rendering are getting slimmer?

 

[Fafalada]

If the PlayStation 3 GPU ends up being a Pixel Shading only design as the other challenger design was then it means that Sony/SCE has a different view on Vertex Shading that includes versatility, but not features like Vertex Texturing

Why? Obviously we don't know yet how easy we can deal with latencies in SPUs, but that problem would always remain, even if you bolted texture samplers directly onto each SPU.

 

[j^aws]

....
With this said, I still say that IMHO the PlayStation 3 GPU is not CELL based, it does not have the SPUs/APUs.
....

It would seem to me a not bad idea to assign all the Vertex Shading work to the CELL based CPU:
....

To me the above two statements seem to contradict each other.. :?

Just to re-cap we exchanged a few pages of posts here in this thread,

http://www.beyond3d.com/forum/viewtopic.php?p=427444#427444

and continued a few exchages here,

http://www.beyond3d.com/forum/viewtopic.php?p=432187#432187

....
What does it mean having a "Consistent ISA without having the SPU/APU construct" ? Are you talking about on-the-fly JIT translation ? Emulation ?
...

I'm just looking at them as a bunch of SIMD units...the patents say you can have more or less SIMD units depending on performace...so what I'm thinking is that they have a Cell template, like the Toshiba MEP template if you remember?

...
Do you think nVIDIA is extending the ISA their Shading ALUs use to support the same ISA as the SPU/APU ?
...

Depends if you think there will be any shading going on in the CELL CPU? i.e vertex shading?

And of course they'll be eDRAM, Sony-Tosh love eDRAM! ...unless NV implement this turbo-cache thingy!

And at the end you asked me two questions which I answered with two more questions above!

1. Basically, the way I see it is this...if you're gonna have vertex shading on the CELL CPU, then they're gonna be VS CELL threads, (aka software Cells), no?

If Hofstee was talking about two-way comms between CPU<=>GPU, then these VS CELL threads should run on the GPU also, no?

This should imply the GPU is CELL based.

2. OTOH, If you assume that these VS threads running on the CELL CPU aren't 'software Cells', i.e. VS non-CELL threads, then they're running in an evironment on the CPU with OTHER CELL threads. Why would you create a CELL architecture and run CELL and non-CELL threads on the CELL CPU?

Wouldn't this seriously fuck up the scheduling of these CELL threads (software Cells) from a load balancing POV in this environment?

3. However, if the VS is NOT done on the CELL CPU but on the GPU, then the entire GPU may OR may not be CELL based then, no?

This is how I see things and may be jadded due to the holiday season!

 

[version]

1. if on GPU dont run any vertex process, then 6 GigaPoly/s is the maximum without texture upload (1 poly about 16 byte,redwood 100GB/s)
about 3 Gigapoly/s with texture
32 pixelpipeline is enough on 800MHZ to exploit this


2. if on GPU use nurbs with displacementmapping then to require vertexprocess,
vertexshader or apulets or directhardware?
vertexshader not likely , this is nightmare for programming(cpu,APU, VS,PS, 4 different langugage and architect)

 

[London Geezer]

Huh?

 

[Qroach]

Here's my prediction on what has happened thus far with PS3 development


1. Many people working on PS3 hardware didn't know or expect this to happen.

2. The Nvidia's invovlement in PS3 did not start for one or two years before this annoucement. Nvidia's involvement was minimal, nothing more than NV execs talking on a high level to Sony execs, and sending out graphics cards for Sony to evaluate performance. Sony had been evaulating Nvidia tech for a while with zero cash exchanged between both parties. Nvidia didn't change what anything in thier roadmap to accomodate sony up until this annoucement.

3. Toshiba was left out in the cold with this change and didn't know it was going to happen. The work on the GS for PS3 will continue, but that graphics hardware won't be used in the PS3.

4. One of sony's offices led the charge to make this change, as internally issues were raised with the expected performance of the new GS and/or expected feature set/delivery time frame.

5. The nvidia licensing deal will cost Sony more money, but allow them to compete better on the graphics technology side and a later launch. The trade off was cost vrs performance & delivery time. Nvidia will incorporate support for the graphics API that Sony is jointly developing with other partners.

6. Nvidia will supply the entire GPU with minor modifications. It will be a custom version of a future Nvidia graphics chip (along the lines of Xgpu) however sony will be in charge of fabbing (as we already know). The Nvidia processor "can" handle all vertex and pixel shading, or the cell CPU can be used for vertex shading. They will not be incredibly integrated together.

7. Cell is a really good CPU, but NOT the second coming of christ.

8. There are few to zero actual Cell workstations available for developers to get cranking on. Sony is currently using emulators and PC's to perform their development, and the final devkits won't arrive until Nvidia has delivered their custom GPU late next year.

9. You will not see any realtime raytracying, radiosity lighiting, or any advanced rendering techniques beyond what is readily capable with the graphics hardware found on PC's of the same time frame. you might see real support for subdivision surfaces in hardware, but they will only be used on chracters.

Feel free to quote me on this or post this when we see real PS3 specs or actual performance out of developers. grab and save this post if you want I don't F'ing care. I think there's some people around here that are living in an imaginary dreamworld with regards to sony and PS3 and their plans for it. It's a game console, no company is invincible (MS could fuck up easily, just as sony or nintendo) and sometimes tradeoffs are made.

 

[London Geezer]

I don't agree with many of the things you just said, or at least i think you (and everyone else) are in no position to make judgements on what we know very little about.

 

[Qroach]

You don't have to agree with it, as I said I don't F'ing care. Btw, you don't know what information I may or may not know about what has happened thus far, so do NOT tell me I'm not in the postion to speculate or post my prediction. I can do what I please or are you my master

That's a common defense in forums when someone readons something they don't like, they say "nobody here is in a postion to speculate", but funny how they only say that right after they disagree with someone. Like I said, you can quote me on it, or throw it back in my face if I'm wrong for all I care.

 

[London Geezer]

You don't have to agree with it, as I said I don't F'ing care. Btw, you don't know what information I may or may not know about what has happened thus far, so do NOT tell me I'm not in the postion to speculate or post my prediction. I can do what I please or are you my master

That's a common defense in forums when someone readons something they don't like, they say "nobody here is in a postion to speculate", but funny how they only say that right after they disagree with someone. Like I said, you can quote me on it, or throw it back in my face if I'm wrong for all I care.

Hey gee don't go all aggressive on me, remember i'm the Master, BOY.
All i'm saying is that your post screams "Sony don't know what they're doing, this is a rush job, Toshiba got stuffed" from every pixel. And you are not in the position to say that.
Speculation is fine and dandy, but it's not a medium to show us your personal preferences of a company over the other, like you usually do anyway so why am i even questioning your ways? :?

 

[Cryect]

1. if on GPU dont run any vertex process, then 6 GigaPoly/s is the maximum without texture upload (1 poly about 16 byte,redwood 100GB/s)
about 3 Gigapoly/s with texture
32 pixelpipeline is enough on 800MHZ to exploit this

Something tells me we aren't going to see games pushing 6 billion polygons a second. I expect maybe at the maximum a magnitude under that somewhere.

2. if on GPU use nurbs with displacementmapping then to require vertexprocess,
vertexshader or apulets or directhardware?
vertexshader not likely , this is nightmare for programming(cpu,APU, VS,PS, 4 different langugage and architect)

I would assume the VS language would be similar to the PS language as has been in every shader language I've worked with. And really VS are pretty damn easy to write in comparison to trying to manage cell threads for doing vertex shading. The only time VS can be somewhat of a pain is when you are doing something completely different and kinda strange.