New dev tools ? PlayStation 3 Edge

From what I understand of this post , joker is essentially basing his example, opinion and impression from a part of the whole graphics system whereas the other devs who seem to offer contradicting views have all along been talking about the whole. Of course, my impression could be wrong but I'm just looking at the numbers. This would also coincide with what I asked in this post regarding the balance between PS/VS in games and the question that was asked in another thread by joebloggs.
 
I think the missing ingredient however to understanding "the RSX vertex capabilities are pants" is compared to Xenos as joker had mentioned, perhaps not just compared to 7800.
Sure, and as I originally asked, it'd be nice to know why. Is it just a matter of Xenos having 3x the vertex capability? - which doesn't help too much because at some point it has to stop and do pixel shading! Or is the setup or some-such fundamentally different. If you recall the triangle setup of RSX 'broken' at 250 M tri's per second in an Inquirer article ages back, are we looking at that being the issue here? Or something else entirely? What is it about Xenos that's putting it so far ahead of RSX?
 
Is it just a matter of Xenos having 3x the vertex capability? - which doesn't help too much because at some point it has to stop and do pixel shading!

You practically don't do pixel shading for the shadow and Z-only passes, which can in some scenes be 2/3 of the vertex count of a scene. (But not the vertex load since their vertex shaders are also lighter.)
 
Probably some will think I sound boring or fanboish but to be honest I don't care I stand with my opinion on the subject :) ppl can write even 10k times that RSX sucks at VS (and it seems to me they don't even know why..) but they're wrong. Maybe they think the problem is in VS, but this doesn't automatically make this true.
I'd really wish I could say more but it's obviously not possible.
For my own experience HS is the vast majority of times not VS limited and I think we push quite a lot of geometry on screen, both with simple (zpre pass and shadow maps) and complex rendering passes.
Obviously things can change a lot from game to game and good practices on a platform can be rally bad on another one and vice versa.
 
Probably some will think I sound boring or fanboish but to be honest I don't care I stand with my opinion on the subject :) ppl can write even 10k times that RSX sucks at VS (and it seems to me they don't even know why..) but they're wrong. Maybe they think the problem is in VS, but this doesn't automatically make this true.
I'd really wish I could say more but it's obviously not possible.
The fundamental problem to the argument. All we can ever have, thanks to NDAs, is different sides expression a truth without an explanation of the truth. Those listening can only take sides with the truth that appeals to them, without understanding the reasons the opinions.

joker says RSX VS sucks
nAo says RSX VS doesn't suck

Who ya gonna believe and why? nAo because he's a hardcore pro? Or joker, because nAo's a Sony lacky on their payroll and joker's truly independent?

Without real technical insight, this discussion can't be discussed, and should be forgotten. Which is a real downer as the topic's an important one, especially when considering Cell's role in graphics work.
 
Or joker, because nAo's a Sony lacky on their payroll and joker's truly independent?

nobody has said that. :cry:

the only difference (as I have pointed out) is that Joker has BOTH consoles to compare too.

it does not make him any more unbiased than nAo but it does add more info to the equation of what happens when a game is run on RSX w/o Cell help (and we know HS uses Cell) compared to Xenos without Xenon help.
 
The fundamental problem to the argument. All we can ever have, thanks to NDAs, is different sides expression a truth without an explanation of the truth. Those listening can only take sides with the truth that appeals to them, without understanding the reasons the opinions.

joker says RSX VS sucks
nAo says RSX VS doesn't suck

Who ya gonna believe and why? nAo because he's a hardcore pro? Or joker, because nAo's a Sony lacky on their payroll and joker's truly independent?

Without real technical insight, this discussion can't be discussed, and should be forgotten. Which is a real downer as the topic's an important one, especially when considering Cell's role in graphics work.
True but there's always a but.

Now as a newbie, and a wannabe at best when it comes to graphics rendering, I got the impression that nAo works on Heavenly Sword, a title which doesn't seem to be haltered by lack of vertex shading.

And with the still pictures, videos, demos and released PS3 titles his claim seems even more valid.

Now I'm not saying that Lair, R&C Future, Uncharted (the latest EGM issue) can't be done on the 360, just that I have yet to see it (keeping fucus on vertex shading and nothing else, not interested in starting a peeing contest).

I may be reading Joker wrong but to me it seems like he claims that CELL is so dedicated to handling the vertex shading that there's nothing left for anything else/can't compete with Xenon afterwards which clearly isn't true.

Am I rambling? :oops:
 
It does cut into CPU resources that can be used for other things, but I guess it's a question of how much, and what CPU resources you have to spare. I wonder if there's a developer consensus on RSX vs. Xenos, one perhaps even nAo agrees with. For example, (not saying RSX is crippled) but would the PS3 have benefitted with Xenos instead?
 
Last edited by a moderator:
IMO, the fact that Joker is working on a sports title where Xbox360 appears to be the lead platform, with the PS3 version being essentially a port with minimal redesign, means that he's going to be most keenly aware of where PS3 is weaker/more difficult than Xbox360. That's the nature of ports: only the things that don't fit as they did on the previous platform figure prominently; any areas of untapped resources of the new platform are not going to be fully appreciated.
 
It does cut into CPU resources that can be used for other things, but I guess it's a question of how much, and what CPU resources you have to spare. I wonder if there's a developer consensus on RSX vs. Xenos, one perhaps even nAo agrees with. For example, (not saying RSX is crippled) but would the PS3 have benefitted with Xenos instead?

I'm also wondering if there couldn't be a potential gain from letting the CELL handling culling and such when it comes to more "dynamic" graphics (again, I'm a complete noob going on a limb here) procedurally generated water waves interacting with a ship.

Always wanted an object like a pillow act like an actual pillow and not just being a "rock" with a pillow painted on it presented in the context of a bed.

But maybe I'm just holding nVidea and Sony in too high regards, in reality they could just be fumbling around without a clue.
 
It does cut into CPU resources that can be used for other things, but I guess it's a question of how much, and what CPU resources you have to spare. I wonder if there's a developer consensus on RSX vs. Xenos, one perhaps even nAo agrees with.

I think the only consensus here is that there is none. Your mileage may vary has probably never been more true - different experiences will be borne out of different approaches to the system.

I think though, if one is considering relative technical capability, it only takes one game to demonstrate what is technically possible if the system is approached in a particular way. I think Heavenly Sword is, going by what we're told of it, a good example that there should not be a relative lack of capability with regard to geometric complexity, for example, of RSX on its own (or pretty much, from what I understand of nAo's comments). Now, there's many other points one could make around that and asides from that if more or less developers have one experience versus the other, but purely on the matter of technical capability, if it's even only shown once, then that capability has been proven. Hopefully we can look forward to that capability being proven more than once, of course, and I'm sure it will be (and is being), and hopefully tools like Edge will help developers for whom that capability isn't so relatively apparent or accessible, and for whom help from other parts of the system is welcome (even if in reality, more could be done without Cell help, for example, with a different or better approach).
 
IMO, the fact that Joker is working on a sports title where Xbox360 appears to be the lead platform, with the PS3 version being essentially a port with minimal redesign, means that he's going to be most keenly aware of where PS3 is weaker/more difficult than Xbox360. That's the nature of ports: only the things that don't fit as they did on the previous platform figure prominently; any areas of untapped resources of the new platform are not going to be fully appreciated.

It is the problem and why many port from 360 to PS3 sucks. I think something like Edge can help here. They can dedicate some CELL ressource to help the port. And the devs like nAo who works on PS3 exclusive have more time for optimizing on RSX.

And in the future Ninja Theory can probably work with the Sony internal version of Edge(ICE) and improving the 3d engine of Heavenly sword. I'm not sure that Ninja Theory has access to this technology since a long time, maybe it was only for 1st party at the beginning.

But it is only one equation of the problem because the multiplatform devs need to adapt physics, IA for the SPE. And they have less time than 1st/2nd party developer to do it.

I don't think multi plateform game are very interesting for judging the PS3/360 power. I prefer wait what the exclusive developer can do on consoles with 3D engine tailored for PS3 or 360. Some developers like Ninja Theory, Insomniacs, Naughty dog for PS3 or Bungie, Remedy on 360.
 
Last edited by a moderator:
Now as a newbie, and a wannabe at best when it comes to graphics rendering, I got the impression that nAo works on Heavenly Sword, a title which doesn't seem to be haltered by lack of vertex shading.
I think people are also a little confused because I don't really think vertex *shading* power is a major problem. Yeah, the Xenos could theoretically divert significantly more vertex shading power on demand, but it really doesn't happen that often because the pixel shading demands are pretty even if not far greater. On the light-load passes, both vertex and pixel shading are light, and on the major passes, vertex shading demand is next to nothing compared to pixel shading.

Even the original thread was all about vertex attribute throughput being a limit, and Xenos is obviously far less so. Being limited to one attribute per cycle means you get a lot of vertex shading cycles for free because you'd otherwise just be waiting for data to come in. But it also means you can't push too many vertices through the pipe even if you had infinite vertex shading power. It's up in the air just how big of a problem that is and that's really coming down to each project.
 
I think people are also a little confused because I don't really think vertex *shading* power is a major problem. Yeah, the Xenos could theoretically divert significantly more vertex shading power on demand, but it really doesn't happen that often because the pixel shading demands are pretty even if not far greater. On the light-load passes, both vertex and pixel shading are light, and on the major passes, vertex shading demand is next to nothing compared to pixel shading.

Even the original thread was all about vertex attribute throughput being a limit, and Xenos is obviously far less so. Being limited to one attribute per cycle means you get a lot of vertex shading cycles for free because you'd otherwise just be waiting for data to come in. But it also means you can't push too many vertices through the pipe even if you had infinite vertex shading power. It's up in the air just how big of a problem that is and that's really coming down to each project.

From what I understand triangle setup is a limiting factor this gen more on PS3 but it could be the case on 360 and the next gen consoles can process much more vertex than the GPU can handle. Is it false or true?
 
Last edited by a moderator:
IMO, the fact that Joker is working on a sports title where Xbox360 appears to be the lead platform, with the PS3 version being essentially a port with minimal redesign, means that he's going to be most keenly aware of where PS3 is weaker/more difficult than Xbox360. That's the nature of ports: only the things that don't fit as they did on the previous platform figure prominently; any areas of untapped resources of the new platform are not going to be fully appreciated.
Could it be that it is mearly "different" rather than "weaker".

Maybe if nAo did get a 360 devkit he could start posting Xenos sucks (although I am sure he wouldn't as he seems a very reasonable guy) at this or that when his PS3 optimised graphics code fails to yeild the same results when he tries to run it on a 360.
 
I think though, if one is considering relative technical capability, it only takes one game to demonstrate what is technically possible if the system is approached in a particular way. I think Heavenly Sword is, going by what we're told of it, a good example that there should not be a relative lack of capability with regard to geometric complexity,


no one is saying that HS or PS3 is not pushing plenty of geometry... the only thing brought up was Xenos may be better at pushing that same geometry (in joker's real world experience) without work being done by the Cell or Xenon to assist.

that's it.

edit:
which really should come as no surprise. we've been talking for 18 mos now about Cell helping RSX to equal some of the things that Xenos can do.
 
Last edited by a moderator:
People without the technical knowledge are going to believe whoever backs up their bias, some go as far to put other opions into their signiture the same day it was posted as if it's gospel. I think we should let the technical people debate it and the rest should just observe and not look like console cheerleaders.
 
I think people are also a little confused because I don't really think vertex *shading* power is a major problem. Yeah, the Xenos could theoretically divert significantly more vertex shading power on demand, but it really doesn't happen that often because the pixel shading demands are pretty even if not far greater. On the light-load passes, both vertex and pixel shading are light, and on the major passes, vertex shading demand is next to nothing compared to pixel shading.

Even the original thread was all about vertex attribute throughput being a limit, and Xenos is obviously far less so. Being limited to one attribute per cycle means you get a lot of vertex shading cycles for free because you'd otherwise just be waiting for data to come in. But it also means you can't push too many vertices through the pipe even if you had infinite vertex shading power. It's up in the air just how big of a problem that is and that's really coming down to each project.

Thanks for clearing that up, it does seem to make sense (if I got it right) that the amount of VS/PS are balanced by nVidea according to the performance of the GPU (though not being able to handle the extremes).

I also see that the Xenos approach allows for more flexibility, being able to divide the VS and PS after need... but with the consoles being closed systems, is there really that big an advantage when the devs know exactly what they have to work with? Guessing here, but the idle shaders could potentially be used for other tasks like physics processing (remember reading something about Havock optimizing their engine to run on GPUs too) while the CELL potentially could be helping out in cases where all the shader pipelines are occupied.

(the closest I have been to redering was when emulatinng a very simple "GPU" on an FPGA, so please do not hit me :D ).
 
no one is saying that HS or PS3 is not pushing plenty of geometry... the only thing brought up was Xenos may be better at pushing that same geometry (in joker's real world experience) without work being done by the Cell or Xenon to assist.

that's it.

edit:
which really should come as no surprise. we've been talking for 18 mos now about Cell helping RSX to equal some of the things that Xenos can do.

I know this goes against human nature, but perhaps we shouldn't blindly embrace claims simply because they state what we want to hear.

Yeah, I'm no better but the RSX being skewed claims do remind me of some weird claims made about CELL not being able to do integers, it being severely hampered because the SPEs lack branching logic and the CELL main BUS being a bottleneck (still wonder why people kept posting the latter claim on Wiki when it's such an extreme not likely to happen scenario, not even considering the deactivated SPE).
 
I know this goes against human nature, but perhaps we shouldn't blindly embrace claims simply because they state what we want to hear.

Yeah, I'm no better but the RSX being skewed claims do remind me of...

nobody is saying it's a POS.

I tend to think that both systems are close to EQUAL (too close to call- see my sig) or at least will be, as Cell gets leveraged with tools like EDGE.

I'm not following anything blindly either, I've been reading about both of these architectures for 2 years with an open mind. It's common knowledge that Xenos is more efficient than RSX and that Cell needs to help RSX in some capacities. this is just confirmation from someone running both.

If it ruffles some feathers here to come to the conclusion that the subsequent strengths and weaknesses of each architecture could/would allow them to perform nearly equally in real world scenarios, then I'm sorry they feel that way.
 
Back
Top