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

[Qroach]

Yessir mastah

You can read whatever you want into my post, I can make a similar post about things MS has done lately that would make YOU think they don't know what they are doing. So what? This has nothing to do with personal preference as none of the next gen consoles have been released yet. I formulate my personal preference when I know somethign concreate about the systems. I do however have the opinion that sony isn't invincible and I think that clashes with some people vision of sony as a company doing everything right.

This is my opinion on what happened and as always there are tons of factors invovled.

 

[London Geezer]

Yessir mastah

You can read whatever you want into my post, I can make a similar post about things MS has done lately that would make YOU think they don't know what they are doing. So what? This has nothing to do with personal preference as none of the next gen consoles have been released yet. I formulate my personal preference when I know somethign concreate about the systems. I do however have the opinion that sony isn't invincible and I think that clashes with some people vision of sony as a company doing everything right.

This is my opinion on what happened and as always there are tons of factors invovled.

Fine, and i said i don't agree on some things you said. That's it. You're the one who went all defensive.

 

[Qroach]

London boy,

You're the one who went all defensive.

First, you attacked my right to express my opinion by telling me that i'm not in a position to make judgments, even when I clearly was not making any sort of judgments.

Second, you try to belittle my opinion by saying I'm showing my "personal preferences" when I feel this has nothing to do with any preferences at all.

I haven't been defensive until now. I mean, what would you expect after the two points above?

 

[London Geezer]

London boy,

You're the one who went all defensive.

First, you attacked my right to express my opinion by telling me that i'm not in a position to make judgments, even when I clearly was not making any sort of judgments.

Second, you try to belittle my opinion by saying I'm showing my "personal preferences" when I feel this has nothing to do with any preferences at all.

I haven't been defensive until now. I mean, what would you expect after the two points above?

I was bored? And i saw your name and jumped to conclusions? Not a long jump to make anyway.

 

[Qroach]

Not a long jump to make anyway.

Longer for some than others... yes.

 

[pahcman]

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.

Well said post, this portion i must ask something.

I was so caught up in Vince (was it Vince?) forceful "whats a cpu whats a gpu" arguement, that i was surprise Nvidia got the PS3 GPU. Because i do remember Nvidia is against a unified system soon. So GPU deal may mean Cell *is* so powerful that Nvidia see unified system finally doable on Cell than on peecee. Or i guess Vince was being overly forceful and optimisic that Cell is *so* powerful, "gpu" is no more, since Nvidia Cell-PU and NV50 will be conflicting r&d.

Does this mean tflops PS3 is also out? Remember too Vince expect PS3 to be tflops ahead of gflops rivals, that difference will be discernable and such raw power can do anything internally as Nvidia/Ati designs can.

Please know, my memory is on Vince because he strikes me as a very knowledgeable Cell hobbyist researcher, i could be wrong to quote Vince, sorry Vince if so.

 

[London Geezer]

I have long legs.


Anyway, back to topic, kinda. I'm not sure i asked this already, but am i the only one who feels that now that NVIDIA is on board, this talk about "different rendering systems like REYES or whatever" is walking away steadily to the next next generation? Meaning, aint gonna happen this time around. I was skeptical before, and NVIDIA doesn't sound too different-rendering-y to me. It just sounds like more of the same, which is not a bad thing in itself, since we'll get much more of what we can see today on top PCs but hey...

 

[j^aws]

...
Anyway, back to topic, kinda. I'm not sure i asked this already, but am i the only one who feels that now that NVIDIA is on board, this talk about "different rendering systems like REYES or whatever" is walking away steadily to the next next generation? Meaning, aint gonna happen this time around. I was skeptical before, and NVIDIA doesn't sound too different-rendering-y to me. It just sounds like more of the same, which is not a bad thing in itself, since we'll get much more of what we can see today on top PCs but hey...

FWIW, I still think it's a possibility!

In this recent thread there are plenty of hints if you read it all,

http://www.beyond3d.com/forum/viewtopic.php?t=18849&start=0

and these series of posts,

http://www.beyond3d.com/forum/viewtopic.php?p=430107#430107

 

[nAo]

FWIW, I still think it's a possibility!

In this recent thread there are plenty of hints if you read it all,

http://www.beyond3d.com/forum/viewtopic.php?t=18849&start=0

Niet..that stuff is not about realtime rendering!

 

[Qroach]

I think any fancy rendering ideas should be thrown out and we'll be bound by what ATI and nvidia are capable of doing for a long while. I don't expect either company to achive a renderman like rendeirng system anytime soon. Although marketing may make people think they are capable of that.

 

[London Geezer]

The thing is that the whole concept of Pixel Shaders (which will not be called Pixel Shaders on PS3 i guess as that is a DX term, but i could be wrong) was invented to fake effects offline renders render through raytracing. So, unless the same hardware can be used for a small-scale form of raytracing, we're still gonna see polygons and shaders for a long time.

 

[Qroach]

Pixel shading was designed to provide a procedural method of representing surface texture among other things. It wasn't really designed to fake the effects of a raytracer specifically. You can do a whole lot with a scanline renderer.

 

[j^aws]

FWIW, I still think it's a possibility!

In this recent thread there are plenty of hints if you read it all,

http://www.beyond3d.com/forum/viewtopic.php?t=18849&start=0

Niet..that stuff is not about realtime rendering!

I'm not talking about the patent...just remove the RT/GI stuff from the patents and your left with REYES. If you follow all the links in that thread, NV's goal is to make Offline rendering a Real-time prospect, which has also been Sony's goal with all the GScube R&D...

And scanline renderring isn't going to replaced by full RT, which alot people got the impression from that thread...their hearts still lie with Shaders!

 

[Qroach]

I don't think we'll see anythign like reyes next gen.

 

[j^aws]

I don't think we'll see anythign like reyes next gen.

If you can tell me that a GScube-64 (approx. 600-800 GFlops) was not capabable of real-time REYES because of lack of power, then I'll be more inclined to rule it out next-gen!

 

[nAo]

If you can tell me that a GScube-64 (approx. 600-800 GFlops) was not capabable of real-time REYES because of lack of power, then I'll be more inclined to rule it out next-gen!

it's not a matter of power.
You need a special hw support to make it (reyes) efficient and fast.
PURE sw reyes rendering is a suicide.

 

[London Geezer]

I don't think we'll see anythign like reyes next gen.

If you can tell me that a GScube-64 (approx. 600-800 GFlops) was not capabable of real-time REYES because of lack of power, then I'll be more inclined to rule it out next-gen!

GSCube had 64 GSs.
PS3 won't have a GS-like architecture.

 

[j^aws]

I don't think we'll see anythign like reyes next gen.

If you can tell me that a GScube-64 (approx. 600-800 GFlops) was not capabable of real-time REYES because of lack of power, then I'll be more inclined to rule it out next-gen!

GSCube had 63 GSs.
PS3 won't have a GS-like architecture.

GScube was a highly parallel architecture...What is PS3?

 

[j^aws]

If you can tell me that a GScube-64 (approx. 600-800 GFlops) was not capabable of real-time REYES because of lack of power, then I'll be more inclined to rule it out next-gen!

it's not a matter of power.
You need a special hw support to make it (reyes) efficient and fast.
PURE sw reyes rendering is a suicide.

If I was CTO and NV's goal was to make offline-rendering a real-time prospect and I was designing a new architecture, where would my GPU designs go without the bounds of Wintel? Why would I join allegiance with Sony who have a similar goal? And if you'd want to break away to a new paradigm, you don't do it in one big jump but allow the 'old' to exist with the 'new'...which seemed to be hinted by Hofstee when he said the architecture would allow you to 'try' new things!

 

[Panajev2001a]

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.

First the latencies with dedicated TMUs would be much much lower IMHO: an SPU/APU bi-linear filtering a texture would waste quite a bit of the unit's performance.

Second, I do not see how this changes my point: if they saw Vertex Texturing and VS 3.xx/4.xx as NEEDED then they would push the most optimal way to support it (performance and cost wise) and that would be to have a non PS only GPU (which might still be what nVIDIA is delivering, we do not know if it will be PS only).

The other alternative, if they really saw Vertex Texturing as important would be to bolt TMUs close by each SPU (I think the SPUs will be able to deal with the latencies that texture fetching+filtering would imply... the stall would be quite lower than if te SPU would be filtering the texture in software and the power consumption would be lower as we would have the SPU idle a bit or switiching to another thread while the TMU is working).