Do you think we will get Killzone CG level graphics next gen?

[London Geezer]

not. it's technically impossible. it's even impossible for PS3 (or Xbox 360) to have a good enough approximation that fools the eye into thinking they can do Killzone E3 2005 graphics.

Says who? Approximations and "tricks" can give some really, REALLY amazing results. SOTC and a few other games on PS2 showed that (HDR, Fur, DOF, motion blur, all with loads of geometry shown on screen... on PS2). It's all about spending enough time on the platform and have enough brains to come up with the best "trick" to approximate certain things the fastest way possible. It's all about swapping one "approximation" with another one that runs faster.

the PS3 CELL CPU is about 35 times faster/more powerful than the PS2 EE CPU. but in terms of graphics the leap is much smaller than 35x.

How is that quantified? Seriously. What framerate would a PS2 attain when trying to run a 1080p game with all the per-pixel effects, all the geometry and textures, animation, physics a good PS3 game has? That is, if there was enough RAM on PS2 to store such a game, and if PS2 even supported the per-pixel effects used on PS3, which is obviously not the case... My guess is that it would be MUCH less than 1fps, which is already 30x (or 60x, depending on the game) slower than PS3. By your "estimate", such a game would run at, say, 5fps on PS2 (a 6-12 times difference, which is "much smaller" than 35x), which is absolutely ludicrous.

 

[sunscar]

Approximations CAN be very convincing, yes, but there is a reason why we call them approximations - they attempt to approximate the look of something else, as closely as possible within the performance parameters given. It's important that it's understood that approximations aren't perfect, and while some are better than others, we're still talking about similarities between properties of only a small percent. If we wish to go beyond what could be describes as a passing similarity, then we have to move away from approximation, and begin moving toward using similar effects and assetts to what is used in our teasers. Unfortunately, right now the technology doesn't entirely allow for that - We can't just decide to display a few dozen 300,000 poly character models, and a few-dozen-million polygon environments, all lit up like Christmas with a ray-based lighting system (Not even counting the power needed to simulate all of that). The hardware right now just isn't designed for that kind of work.

Untill it is designed to deal with that kind of workload, approximations are what we have, and they will show. That doesn't mean everyone's going to notice, and it doesn't mean everyone's going to care, ofcourse, but the truth is the truth - I can't change that.

 

[sunscar]

How is that quantified? Seriously. What framerate would a PS2 attain when trying to run a 1080p game with all the per-pixel effects, all the geometry and textures, animation, physics a good PS3 game has? That is, if there was enough RAM on PS2 to store such a game, and if PS2 even supported the per-pixel effects used on PS3, which is obviously not the case... My guess is that it would be MUCH less than 1fps, which is already 30x (or 60x, depending on the game) slower than PS3. By your "estimate", such a game would run at, say, 5fps on PS2 (a 6-12 times difference, which is "much smaller" than 35x), which is absolutely ludicrous.

Didn't actually wish to create a new post for this, but it was a different thought.

The answer to that which you ask above is simple - It isn't seriously quantified at all. It's a layman assumption based on the PR numbers released for both CPUs - PS3CellBE @218Gflop/s in single percision math, and PS2EE@6.2Gflop/s in single percision math. It doesn't really speak volumes about anything. Similarly, the quote that Nvidia made a year or two ago about RSX being fourty times more powerful than the PS2GS - It doesn't state specifically "how" it's faster, or in what way.

 

[BobtheVila]

It could do 10,000+ silent hill models, render to texture stencil volumes like silent hill, grayscale bump mapping(only the per-pixel code is done on the vu1, and it's about 2 instructions one for each axis of the texture, though i'm aware instructions are different then operations), also vram only stores buffers or a 2d array image in bytes, and sony and nintendo put in vram for only the needed buffers(R2T then is done more by using back buffer rather adding a p-buffer), also GC's starwars rouge squad game looks good with high texture qualities and libraries and yet only has 2mb of frame buffer and 1mb of texture buffer, while ps2 could do 2mb of frame and 2 of texture by using the same or a little higher then GC.

It wouldn't look quite like the CG vid of Killzone on ps2(while it has been confirmed by Gurilla that they are recieving similar results in real-time now on ps3), but ps2 needs more fully bump mapped games like Path of Neo, also so does GC, they never use much of these systems realy. Even if they do in geometry, bump mapping the grayscale is quite small to normal in performance size graphically and even in code, even if it looks amazing, so it wouldn't take too much more power to make, it can be done in a simple pass(while normal mapping takes 2 to 4 even thouhg it realy looks pretty much the same in motion). A paper I read for ps2 optimization reads 300Mpixel eaten every 4 passes and 37million textured triangles(over xbox's 31million).

Also if you noticed from hitmanBM on ps2's and pc demo's low pixel specular texture shine that bump mapping does not change shade per-pixel induvidually from the light like a shadow volume or emboss would. So if you had a low quality light you'd see it like an animation sweep, interesting. It's like a grid graph comparison of the vectors point calculated every pixel vs the pixels in the light, so actually the number of operations may be for each axis rather then pixel.

And also a question about the multi-pass thing. How does ps2 do multi-pass, do the additional passes geometry read to texture or are 2d? It seems so with what that official paper shows. And also someone said it's the same for GC, that it's multi-pass rendering does not realy eat half your fill-rate for each pass, because it's rasterised. That and a dev says GC's multi-pass rendering doesn't eat much fill-rate. So i was wondering if ps2 was the same in doing MPR because of that paper that is the result of the performance anylyzer.

Also what is a fullscreen pass as said in the paper when trying bump mapping, is this meaning where the pass's geometry is done in 2d in performance?

 

[Laa-Yosh]

Okay, is it just me, or do you guys also have a problem comprehending the last post here?

 

[dantruon]

Okay, is it just me, or do you guys also have a problem comprehending the last post here?

it sounds interesting but i got no idea what is the main point of the post, so is not only you that cant comprehend it.

 

[BoardBonobo]

Yes. I don't see why we wouldn't be capable of that level this generation.

 

[assen]

Okay, is it just me, or do you guys also have a problem comprehending the last post here?

Somebody finally paired Markov chains with computer graphics

 

[Cheezdoodles]

This thread should just be locked to be honest. Its not even a real debate, this is gameFAQs quality.

On one side you got the nay sayers, who say "no" with technological explanations, and basing their statements on REAL FACTS (ie. PS3 specs). They explain why its impossible.

Then you got the "yay" sayers, who say yes, simply because they liked what they saw at E3 05, and want to believe hype, just because they either f@nboys or just wanna play such a game. But not using any technical explanation as to why.

They believe this can happend with arguments like:

1: PS2 could do amazing stuff later in the year
2: Consoles get better with time.

While both the above mentioned things are facts, it does in NO way, make the next statement any more likely:
3: Killzone CGI will be achieved on the PS3

What the yay sayers are using as arguments is beyond logic. Its just whishful thinking without using anything to back it up.

Its just as stupid as making this prediction:

Since the console marked has grown the last 2 generations, i predict that we will sell 10billion consoles this generation.

See? Im using the same argument as you guys?

Of course, the yay sayers might be right to some degree, (depends on how you measure the Killzone CG, are we measuring poly\counts effects, etc, or are we just after getting something that looks fairly close?), but right now, none of you have any real reason that can be used in these types of debates as to explain why it will be achieved on a PS3.

So please, let this thread die, unless you got a real technical explanation as to why it will be achieved, and how.


"Killzone CGI will be achieved on the PS3 because i like to believe Sony hype", is not a real argument.

 

[BobtheVila]

This thread should just be locked to be honest. Its not even a real debate, this is gameFAQs quality.

On one side you got the nay sayers, who say "no" with technological explanations, and basing their statements on REAL FACTS (ie. PS3 specs). They explain why its impossible.

Then you got the "yay" sayers, who say yes, simply because they liked what they saw at E3 05, and want to believe hype, just because they either f@nboys or just wanna play such a game. But not using any technical explanation as to why.

They believe this can happend with arguments like:

1: PS2 could do amazing stuff later in the year
2: Consoles get better with time.

While both the above mentioned things are facts, it does in NO way, make the next statement any more likely:
3: Killzone CGI will be achieved on the PS3

What the yay sayers are using as arguments is beyond logic. Its just whishful thinking without using anything to back it up.

Its just as stupid as making this prediction:

Since the console marked has grown the last 2 generations, i predict that we will sell 10billion consoles this generation.

See? Im using the same argument as you guys?

Of course, the yay sayers might be right to some degree, (depends on how you measure the Killzone CG, are we measuring poly\counts effects, etc, or are we just after getting something that looks fairly close?), but right now, none of you have any real reason that can be used in these types of debates as to explain why it will be achieved on a PS3.

So please, let this thread die, unless you got a real technical explanation as to why it will be achieved, and how.


"Killzone CGI will be achieved on the PS3 because i like to believe Sony hype", is not a real argument.

The developers say they are already acheiving those graphic in a way on one site, and motor storm is the same, they acheived about the same thing as the CG e3 vids. Killzone will look about the same with some slightly lesser quality graphics, also i could even see the normal medium rez textures on some floors. Most people get confused with the MUCH BETTER LOOKING CG film 30fps with motion blur and moving cam to animations for impossible GFX.

Also the ps2 can be expected to see better games like up to SW:RS3, GC is a lesser powerfull console, 1.9gflops on cpu(10 overall, but those are hardwired fpu's like the pixel engine and hardware stuff, and also this does not mean T&L and it's TEV, they are a smaller portion of that number). 3mb of total vram, 20,250,000 realworld polys vs ps2's 36,750,000 textured polygons, yet just because the per-pixel codes are done for them people think it's better. So all i mean is more ps2 games should look like SW:RS3 with full bump mapping and projected textures, shadows, and other.



I'll ask else where for the multi-pass question, this is the wrong thread.

 

[swx]

Or you can try to invent a completely different approach to rendering; I've heard an idea that one should rely less on textures and more on HOS-based geometry on the PS3, tesselating and culling literally millions of polygons on the SPEs before sending the scene to RSX.

Can you please elaborate?

or some one that can explain this more?

I'm no 3d technician but i want to know how you can gain more detail without relying on more detailed textures....

 

[Diamond.G]

Can you please elaborate?

or some one that can explain this more?

I'm no 3d technician but i want to know how you can gain more detail without relying on more detailed textures....

I'll take a stab, and say by using really high poly models/meshes. Basically taking the Lightwave/Maya approach to rendering. Those programs cull (cut out) and tesselate (stretch) High Order Surface geometry into usable models, well more or less.

 

[Cheezdoodles]

The developers say they are already acheiving those graphic in a way on one site, and motor storm is the same, they acheived about the same thing as the CG e3 vids.

HAHAHAHAHAHA

Seriously, take a look at any motorstorm realtime video, and compare it to the CG. Its not even close, not by a longshot. The amount of particles, the polygon count, the lightning, is in a completely different league than what Motorstorm the game looks like.

Just because you dont remember the video, or does not understand graphical technicalities, does not mean they achieved the same thing as the CG E3 vids.

Motorstorm, is a very good looking game, but it not even close to the CGI video.

 

[Shifty Geezer]

...and motor storm is the same, they acheived about the same thing as the CG e3 vids.

This shows a fundamentla flaw in the rationale of the 'ya' sayers (to coin a phrase). If you can say Motorstorm the game is approximately the same as Motorstorm the render, you are saying to offer considerable latitude to the idea of 'the same'. There are huge differences between the two incarnations of Motorstorm. It's fair to say they are the same conceptually and in flavour, but the realtime is not rendering as well as the CG. By that, you can argue that the KZ realtime will look similar to the render in terms of concept and flavour. It'll look roughly the same. But if it lacks all the high-quality aspects that made the CG trailer really stand out, it won't actually be the same. It'll just give the same impression. And asking whether Killzone CG graphics will be achieved, the answer is simply 'no', just because the CG had loads of AA! That's before any concerns of geometry, texture detail, smoke effects, and so forth.

 

[swx]

I'll take a stab, and say by using really high poly models/meshes. Basically taking the Lightwave/Maya approach to rendering. Those programs cull (cut out) and tesselate (stretch) High Order Surface geometry into usable models, well more or less.

Can you point me to an example of that? (link, pictures, video, whatever)

So basically, you don't draw what you won't see (keeping resources availables for more rendering/whatever) and you stretch/tile "high order surfaces" so you don't hace to use so much textures gaining bandwidth or whatever you gain...so at the end you have more power to create or draw more polys/effects/shadres/whatever?

 

[Diamond.G]

Can you point me to an example of that? (link, pictures, video, whatever)

So basically, you don't draw what you won't see (keeping resources availables for more rendering/whatever) and you stretch/tile "high order surfaces" so you don't hace to use so much textures gaining bandwidth or whatever you gain...so at the end you have more power to create or draw more polys/effects/shadres/whatever?

nVidia GDC 2001 HOS

Maya Personal Edition (Free)

Sadly that is about all I can say on that subject. Maybe Laa-Yosh can help a little bit more.

 

[Polygon Crusher]

The developers say they are already acheiving those graphic in a way on one site, and motor storm is the same, they acheived about the same thing as the CG e3 vids. Killzone will look about the same with some slightly lesser quality graphics, also i could even see the normal medium rez textures on some floors. Most people get confused with the MUCH BETTER LOOKING CG film 30fps with motion blur and moving cam to animations for impossible GFX.

Also the ps2 can be expected to see better games like up to SW:RS3, GC is a lesser powerfull console, 1.9gflops on cpu(10 overall, but those are hardwired fpu's like the pixel engine and hardware stuff, and also this does not mean T&L and it's TEV, they are a smaller portion of that number). 3mb of total vram, 20,250,000 realworld polys vs ps2's 36,750,000 textured polygons, yet just because the per-pixel codes are done for them people think it's better. So all i mean is more ps2 games should look like SW:RS3 with full bump mapping and projected textures, shadows, and other.



I'll ask else where for the multi-pass question, this is the wrong thread.

Lol, tell me a game on the PS2 that pushed more than 12M polygons/sec (divide with x frames to get fps) and the KZ2 devs has said nothing about graphics for KZ2 vs CG dream target. Lol youre whole post is incorrect, I give you a F as grade so far!

 

[Acert93]

This shows a fundamentla flaw in the rationale of the 'ya' sayers (to coin a phrase). If you can say Motorstorm the game is approximately the same as Motorstorm the render, you are saying to offer considerable latitude to the idea of 'the same'. There are huge differences between the two incarnations of Motorstorm. It's fair to say they are the same conceptually and in flavour, but the realtime is not rendering as well as the CG. By that, you can argue that the KZ realtime will look similar to the render in terms of concept and flavour. It'll look roughly the same. But if it lacks all the high-quality aspects that made the CG trailer really stand out, it won't actually be the same. It'll just give the same impression. And asking whether Killzone CG graphics will be achieved, the answer is simply 'no', just because the CG had loads of AA! That's before any concerns of geometry, texture detail, smoke effects, and so forth.

Thanks Shifty, you point out the core issues that cause so much strife on this issue. When people call bland sprite shurbs the same as CGI it really makes for a nonsensical discussion.

 

[StefanS]

There goes another Killzone thread...