My view on what PS3 games could/will look like.

Possible on PS3 in game?

gino_dammers1.jpg


meny_hilsenrad1.jpg


john_fox1.jpg


willi3.jpg


petrovic2.jpg
 
Pictures:

1. This picture should almost definitely be recreatable on the PS3 in real time. There's nothing that great going on in the background and the character doesn't look exactly nice either. It looks good, but of course better has been seen.

2. Yes. What's with all the aliasing?

3. Racing games on PS3 should be leaps and bounds better than this.

4. Yes and with ease.

5. Yes.
 
I think PS3 will be able to do all the models from that movie, but not all the fx, fsaa, lighting, etc. if you are expecting FF movie quality in PS3 games, you're going to be disappointed.

If the ps3 provides what has been announced, it should be capable of rendering the FFmovie. Keep in mind that near FFMovie, ok not technically(but to the average gamer it should be very difficult to distinguish.), demos have already been shown... look at the fairy. Now the ps3 should be capable of having similar models ingame which to most will be FFmovie quality, with artistic cleverness that can be pushed to near photo real.
 
zidane1strife said:
I think PS3 will be able to do all the models from that movie, but not all the fx, fsaa, lighting, etc. if you are expecting FF movie quality in PS3 games, you're going to be disappointed.

If the ps3 provides what has been announced, it should be capable of rendering the FFmovie. Keep in mind that near FFMovie, ok not technically(but to the average gamer it should be very difficult to distinguish.), demos have already been shown... look at the fairy. Now the ps3 should be capable of having similar models ingame which to most will be FFmovie quality, with artistic cleverness that can be pushed to near photo real.



zidane, not *the whole thing*... that is just impossible.... unless u take into consideration *heavy* cuts in the special effects. i remember reading that in some scenes, frames had to be rendered like 20 times one over the other. that is
20layers x 60 frames per second....

im not talking about texture layers, im talking about proper whole-frame-layers... one frame for some special effects one for other stuff etc etc.....

a bit too much even for PS3 considering all the stuff that goes in one frame.... but of course u can always get similar results in different and more time efficiant ways.... still, it would not be *the same* thing
 
See the problem with talking about this is sony fans will claim it lived up to its hype even if the games look like ps2 games.

People don't seem to understand that the further along we go the more power each step is going to take. The jump from 16-32 bit was huge. From 2d sprites to 3d graphics were big. From 200kpolys a sec to 4 million is more night and day than 3 million - 20 million. Its hard to believe but thats the way its happening. Look at fsaa . Look at the hit it takes on current hardware. Same with fsaa. Its not going to change much. The hit will just be covered up with more power going to it. Then there are lights. Current hardware (r300) takes a big hit with anything more than 8 lights. This is 2003 hardware. Do you think that by 2006 hardware will be able to do the hundreds of lights being done in the ff movie ? Not only that but the sizeable amount of texture layers done.

Set your goals lower. Lets see toy story quality first. Show me that. After that we will move to Ants and bugs life. Then toy story 2 , then monster inc
 
et your goals lower. Lets see toy story quality first. Show me that. After that we will move to Ants and bugs life. Then toy story 2 , then monster inc

The thing is, realtime rendering hardware (Video cards) has become very different from offline rendering hardware (CPU farms). Video cards are highly specialized. If they want to do one thing well, they can. Bumpmapping is a great example; it is done with extreme speed and efficiency even by current-generation video hardware, but you don't see a lot of bumpmapping in old computer-rendered movies. However, due to the memory requirements of realtime renders, they will never approach the polygon counts, texture quality, or antialiasing level of an offline render.

With DirectX9 level hardware, realtime rendered games can have a lot of very impressive lighting and bumpmapping effects. We are already seeing from games like Doom 3, Deus Ex 2, and Halo 2 that real-time shadows will become very prevalent in the future. Also, the high dynamic range of DX9 class videocards will allow much more realistic lighting effects. With DX9 shaders, we should also see much better surface effects for difficult-to-render stuff like skin and glass.

On the other hand, we simply aren't going to see characters with 2 million polygons anytime soon, nor are we going to see 50 shadow-casting lightsources dancing around a scene, or 20-layered textures. (FFTSW) CG movies can use countless terabytes of data to store their vertices, lights, and textures. Videocards are currently limited at 128 MB, and their memory capacity does not improve at the same rate as their processing power; this is a big limitation for large textures as well as anti-aliasing. In addition, stencil shaders and overbright lighting inflict a large performance overhead for each lightsource; an offline rendered scene can afford to use ungodly numbers of lights, but an application cannot.

The next generation of hardware will have the basic architecture necessary to perform extremely impressive lighting and surface effects; perhaps not FF:TSW level, but at least as good as "Shrek". However, these will be limited by the raw power of the videocard; game developers will still have to keep careful count of the number of lights in a scene, and texture size will continue to be limited by memory requirements. Polygon count is set to improve substantially, but I don't imagine that UT2k3-like polycount will be possible with universal stencil shadowing anytime soon, and given a choice between better lighting and more polys, lighting wins out any day. Due to the low resolution of the TV screen, low poly bumpmapped models should look almost as good as high poly models.

The cost of Antialiasing techniques continues to drop; currently, owners of Radeon9700 videocards can keep antialiasing on 4x all the time. The PS3 and Xbox2 will definitely have antialiasing on all their games; even current-day PCs can afford FSAA, and as games become more shader-dependent (like DOOM3) the incremental cost of antialiasing will actually decrease. However, games will still look considerably more aliasy than offline cinematics; a cinematic might use 16x or more stochastic SuperSampling, while games use 4x grid-based multisampling; it's a big difference.

I suspect that within a few generations, perhaps even with PS3/Xbox2, our games will start looking very close to cinematics. The major problem with the current generation is aliasing and unrealistic lighting effects, and both should be solved by the next generation. The big difference between games and cinematics will remain the same as it has always been; scale. Movies will always have much more lights and many more objects in a scene compared to a game.
 
I believe the point was never to convince anybody that FF:TSW will be doable on PS3 as it is - I think what Zidane was trying to point out is that using different methods and tricks, a game that will look pretty damn equal to what the movie is. Perhaps it won't have the millions of polygons per character - but lets not forget how good Doom 3 already looks - and those characters don't even have an amazing amount of geometry. The movie might contain scenes where each frames were rendered 20 times - that however does not mean that using other "cheaper" effects, it would not look much worse in-game. I think that was the whole point.

Also, it has been pointed out before - but for the sake of it, again, at what resolution was the movie rendered in? And at what resolution will PS3 need to render them? Big difference IMO. With the correct art direction, I think a machine as in PS3 will be able to do wonders.

Hell look at MGS2 or GT3 and look how it compares to current efforts on 'more powerful' hardware...
 
Set your goals lower. Lets see toy story quality first. Show me that. After that we will move to Ants and bugs life. Then toy story 2 , then monster inc

No I don't think so, If I'm not mistaken, both stories have per frame polycounts lower than 20M polys, or slightly higher than that in some frames(again I'm not sure, I read it quite a LONG TIME ago.), I too recall reading that in the first one some frames had as little as 1M polys.

Today's h/w is capable of 500,000+ polys per frame, albeit with FAR less effects. a 10-40X jump... and voila you have Toy story level geometry, and as for the effects, they usually jump quite high from gen to gen(look at psx, and n64...), not to mention we ALL KNOW THE next gen will have at the very least a 100X jump in perf. over the current one, it's the least possible jump expected.

Now I don't know about the FFMovie total poly counts, but I've heard most of the char. models have about 100,000, or maybe more, but it is around that figure, and those are likely the most detailed thing there.(again, it was in ars I read that and I don't remember, but at most it was a few 100K polys.)

If I recall correctly the psone games had around 10K for The ENTIRE FRAME in most games. Today what once was for a WHOLE FRAME LAST GEN. goes to a single CHARACTER MODEL, and even more than that models of upwards of 20-30K have been mentioned in the current gen... All with FAR MORE effects, FAR better textures, rez, etc... And in far more detailed enviroments that are suitable for such models.

TODAY the entire frames go to around 250K-500K, if we assume A SIMILAR JUMP to that of the previous gen(although it is likely to be a bigger one.), than that my friends will likely be the geometrical detail of the character models nextgen... and again those will be accompanied by equally detailed enviroments.

Shaders for such models and enviroments should at least offer a significant jump over those currently in use in modern pc gpus. Thus we have models and env.s with similar geometry to those in the FFMovie, with Shaders SIGNIFICANTLY ahead of the current ones in the pc market, and more detailed textures, effects, rez, etc...

Will all that be equal to FFMovie, maybe not technically, but something tells me, that with artistic and programmer cleverness, it will be nigh impossible to tell the difference for the average consumer... and that my friends is what matters.
 
zidane1strife said:
Now I don't know about the FFMovie total poly counts, but I've heard most of the char. models have about 100,000, or maybe more, but it is around that figure, and those are likely the most detailed thing there.(again, it was in ars I read that and I don't remember, but at most it was a few 100K polys.)

Are you sure about this? I remember hearing well over a million polys. I might have hear wrong though.


zidane1strife said:
If I recall correctly the psone games had around 10K for The ENTIRE FRAME in most games. Today what once was for a WHOLE FRAME LAST GEN. goes to a single CHARACTER MODEL, and even more than that models of upwards of 20-30K have been mentioned in the current gen... All with FAR MORE effects, FAR better textures, rez, etc... And in far more detailed enviroments that are suitable for such models.

TODAY the entire frames go to around 250K-500K, if we assume A SIMILAR JUMP to that of the previous gen(although it is likely to be a bigger one.), than that my friends will likely be the geometrical detail of the character models nextgen... and again those will be accompanied by equally detailed enviroments.

If you are talking about UT2k3, yes. However, few games have quite that many polys, and frankly UT2k3's lighting sucks. If you look at games with good lighting, like Splinter Cell, DOOM3, and Deus Ex 2, the polycounts are obviously lower. Due to performance limitations, High Poly and good lighting are often in conflict.


zidane1strife said:
not to mention we ALL KNOW THE next gen will have at the very least a 100X jump in perf. over the current one.

Sorry, but that is just marketing bullsh!t. PS3 will be 1000x more powerful than current hardware only in certain synthetic benchmarks, if at all. Remember, the GeForce3 was said to be "7 times more powerful than the GeForce2 Ultra". Now find me an application where a GeForce3 gets 7 times the framerate of a GeForce2Ultra.

You can't trust what companies say about unreleased products. I mean, hey, the GeForceFX delivers "free antialiasing at all resolutions", right?

And when was the last Xbox game you saw that looked as good as the pre-launch Xbox demo screenshots?
 
Are you sure about this? I remember hearing well over a million polys. I might have hear wrong though.

I read that too in a discover mag. Maybe it's for one or two of the highest detail models, but I dunno in ars. i think it said the poly counts were a few 100ks.

As for the jump well psx is 320x240(or something like that) with 10K at 30fps, and barely any textures or effects. PS2 is at the very least about 150K+ at 60fps, at higher rez, and with more effects than psone, according to mister Jason rubin. That I consider 100x+ increase in the ingame perf.( Xbox, cube and even ps2 are said to reach upwards of 300+Ks which exceed the ingame polycounts of either n64 or psx for an entire second)

A similar ingame perf jump might be possible for next gen, placing 15-20M+ polys as the per frame geometry, if the jump is the same as the previous one. All with effects that would put the previous gen to shame.
 
BoddoZerg said:
zidane1strife said:
not to mention we ALL KNOW THE next gen will have at the very least a 100X jump in perf. over the current one.

Sorry, but that is just marketing bullsh!t. PS3 will be 1000x more powerful than current hardware only in certain synthetic benchmarks, if at all. Remember, the GeForce3 was said to be "7 times more powerful than the GeForce2 Ultra". Now find me an application where a GeForce3 gets 7 times the framerate of a GeForce2Ultra.

You can't trust what companies say about unreleased products. I mean, hey, the GeForceFX delivers "free antialiasing at all resolutions", right?

And when was the last Xbox game you saw that looked as good as the pre-launch Xbox demo screenshots?



Yeah he said a 100x jump, not 1000x.... 100x is pretty realistic in my opinion and rather conservative....

now let's keep it in the console world without taking into the game various Geforce's.....

the PS2 IS in certain aspects more than 100X more powerful than PS1... technology is accellerating at a faster pace today than it was in the 90's therefore it's safe to assume the jump between PS2 and PS3 will be even greater....

and comparing the FMV in FFX with FF:TSW is a bit odd since the styles are very different. FF:TSW has the whole realistic thing going on for it, FFX is more *final fantasy*.... still, i will be very happy to play FF13 on PS3 looking like FFX FMV's......
 
Due to performance limitations, High Poly and good lighting are often in conflict.

Yes, IN THIS GEN. As u can see in my posts, my point is that WITH EFFECTS AND LIGHTING THAT SIMPLY CRE@M THE PREVIOUS GEN ONES, the games feature FAR FAR more geometry than the previous one.

I belief the same shall hold true for next gen, superiour lighting with superiour effects, with superior geometry.
 
Are you sure about this? I remember hearing well over a million polys. I might have hear wrong though.

http://www.digitalsculptor.net/FinalFantasyMovie/akiMaxim/akiMaxim_wireframe.jpg

this is pipelined trough renderman Subdivision surfaces.

Francisco Cortina: Our model pipeline consisted of modeling in polygons and rendering in Renderman subdivision surfaces. Unlike the multi-leveled hierarchical implementation of subdivision surfaces that Maya created, Catmull-Clark Sub-D's are much faster and more efficient. They are a variation of cubic B-spline surfaces, having all the advantages of NURBS and none of the awful disadvantages

Catmull-Clark generates many thousands (tesselation somewhat controlable) of polys from a such wire.Not all the generated polys are needed for realtime game,it' would be overkill.

What i can't see soon in realtime is what can a micropolygon architrecture such renderman can produce.
 
_phil_ said:
Are you sure about this? I remember hearing well over a million polys. I might have hear wrong though.

http://www.digitalsculptor.net/FinalFantasyMovie/akiMaxim/akiMaxim_wireframe.jpg

this is pipelined trough renderman Subdivision surfaces.

Francisco Cortina: Our model pipeline consisted of modeling in polygons and rendering in Renderman subdivision surfaces. Unlike the multi-leveled hierarchical implementation of subdivision surfaces that Maya created, Catmull-Clark Sub-D's are much faster and more efficient. They are a variation of cubic B-spline surfaces, having all the advantages of NURBS and none of the awful disadvantages

Catmull-Clark generates many thousands (tesselation somewhat controlable) of polys from a such wire.Not all the generated polys are needed for realtime game,it' would be overkill.

What i can't see soon in realtime is what can a micropolygon architrecture such renderman can produce.


wow i know this might sound a bit odd, but i thought that Akin's model would have so many polygons i wouldnt be able to distinguish them in a wireframe view... but those are some pretty big polygons.... i mean... they're not THAT many if u can clearly see them..... u know what i mean? they DEFINATELY are not 1 million polygons.....
 
Look, if we get visuals in PS3 that are as good as 5-10 year old,
lower budget CG television shows like Voltron3D, i will be very very satisfied.

forget FF TSW graphics all together. you might not get those
exact graphics in Playstation4 - with all the effects and
layers that are in that film.

I agree with setting your standards lower. lets see Toy Story 1
done first. then Antz, etc. but i suggest even lower than that.
try television show or game cut scene CG, before we even get
into flim quality CG.
 
megadrive0088 said:
Look, if we get visuals in PS3 that are as good as 5-10 year old,
lower budget CG television shows like Voltron3D, i will be very very satisfied.

forget FF TSW graphics all together. you might not get those
exact graphics in Playstation4 - with all the effects and
layers that are in that film.

I agree with setting your standards lower. lets see Toy Story 1
done first. then Antz, etc. but i suggest even lower than that.
try television show or game cut scene CG, before we even get
into flim quality CG.


yeah i'd love to play... whats the name.... u know the CGI series of the Transformers.... i mean i watched it quite a long time ago so it may look a bit outdated now even in my eyes, but i remember i was glued to the TV for the sheer beauty of it....
 
that's funny.. when I think of Toy Story 1.. what was so impressive about it? Add some anti aliasing to current good looking games and I think you'll get something as impressive... of course, I'm not thinking of the effects etc, but just at the end result and what the casual out there might see when comparing the two.

Or should I go back and watch Toy Story? Hehe, my sister has it on DVD somewhere I think.... 8)
 
Phil said:
that's funny.. when I think of Toy Story 1.. what was so impressive about it? Add some anti aliasing to current good looking games and I think you'll get something as impressive... of course, I'm not thinking of the effects etc, but just at the end result and what the casual out there might see when comparing the two.

Or should I go back and watch Toy Story? Hehe, my sister has it on DVD somewhere I think.... 8)

what will really (REALLY) count in the next generation of hardware in my opinion is the Physics.
i'm sure u don't want a model as detailed as a FF:TSW model move like a FFX model... or let's take even MGS2 which still has at the moment some of the best animation in any game.
i would want more. even FF:TSW looks a bit odd at times and that's all because of the animation.
Fluids, Smoke, Fire, Cloth, Hair, Wind, even Light Physics will play a bigger role in the next generation. and they will all take a generous amount of vertices moving around decently to be considered realistic.
the animation is what gives a game character, not simple graphics. look at ICO and it's beautiful movements, and the water, the trees.... look at MGS2 and all the things it has going for it...
the human brain is much more receptive to movement of pretty things than to still pretty things. we are used to seeing the real world and its (obviously) perfect physics. for us to mistake a videogame (or any CG movie), the thing will have to feature an almost perfect recreation of our world's physics.
Cars will have to move like real cars, lift dust that moves like real dust moves....
and on top of that u would have all the rest....
but thats just my opinion.... :oops:
 
couldn't agree more. I still have that image stuck of GT3 plus some amazing storm like wind effects breezing across the road in front of you... this and better lighting and I think games will already look heaps better. :D
 
Back
Top