Nv40
01-09-04, 04:16 AM
Originally posted by Arrghman
Simple raytracing, sure. Production CGI level? Not for a lonnng time...
indeed..
but realtime raytracing is not the goal ,for Nvidia (it doesnt look like it is ,not in the near future),doing similar quality like raytracing and photonmaping ..it is.. thats why we have Pixel shaders.. in fact im skeptic about if Nv3O can do -real- photomapping.
it looks to me that their photonmapping demos are more hacks to push the hardware in a very controled enviroment. that only works in special situations.
Here's an example. This (http://www.newtek.com/products/lightwave/tutorials/fullprecision/images/cornell_montecarlo.jpg) image is a radiosity image rendered in LightWave. There are no "traditional" light sources in the shot; all the light comes from a bright spot on the ceiling polygon itself which casts a shadow on the floor (a very soft shadow mind you), illuminates the top of the sphere and the blue and red walls which diffuse light onto the sides of the sphere.This kind of effect is used in many production level CGI shots and has been for the past several years as computers became powerful enough to render such frames in a reasonable amount of time. I keep saying this, but the point doesn't seem to get through, this is an enormously complicated light simulation, so much more so then any of the shaders that games are just barely starting to use now and hardware is going to have to mature a LOT to be able to even consider being able to do something like this in real time. Simple raytracing scenes can be rendered in less than a second on modern CPUs, by contrast radiosity shots can take several hours depending on complexity. There's no magic graphics card that's going to come along any time soon that can do this in 1/30th of a second.
yes ..the effect you are describing in Lightwave is something called FINAL GATHERING
,but dont know how its called in Lightwave since its renderer is a RADIOSITY Engine.
a more advanced technique than raytracing..that it is comparable with Photomapping.
its very complex ..indeed.. however it will be good to know that Photonmapping and Radiosity the most advanced techniques in CG used in today movies are very carefully used in few scenes (IF ANY) in 100% CG movies.. not because of the quality ,which is the best a computer can do ,but because of the Long TIme it takes.
it will be good to classify movies in two cathegories..
1)CG movies.. (like toy story/ANts/SHrek/ a bugs life/FF/Monsters/Nemo
2)and Digital FIlms.. that use real camera shots ,with some scenes of CG graphics..
Starwars,lord of the rings,Hulk ,Matrix..Terminator
the first group almost doesnt use photonmapping.
PIXAR movies..Toys STory1/2 use ->[nothing ] of radiosity or raytracing,same with 95% of Pixar movies ,same with Square FInal fantasy. ;) Shrek/Ants/ a bugs life? . however MOnters use it at some shots..finding NEMO, maybe more.. Star wars:atack of the clones is a diferent story :) but most of that movie is real life film with real people and scenarios ..not like FF ..yes nothing can match the real thing ;)
Lord of the RIngs use it some times ,only in some shots -and this is what people dont know .. that most CG movies use rendering techniques cheapier and faster than phtonmapping and radiosity. ;) Pixar REnderman support for Photonmapping only happened last year. and still they will tell you that Photonmapping is something that should be touched with a 10 foot Rod,in big films ,avoided whenever its possible. thats why Nvidia or ATI can claim Cinematic CG quality in real time.. since most of the movies that are CG use SCanline rendering like games.. ;) its just that the shaders in movies are very high quality and complex.. now you know all the fuss and all the hype for very long instructions shaders in directx9 Gpus.. because they are necessary for cinematic quality ;)
photonmapping renderers like this..
http://www.finalrender.com/products/products.asp?UD=10-7888-35-788&PID=36
http://www.splutterfish.com/sf/gallery_index.php
http://www.mentalimages.com/4_1_motion_pictures/index.html
have more advanced lighting techniques that most movies that are 100% CG. like Pixar movies.. the day to see a 100% CG holywood movie ,like FF but using photomapping -at all times- can be even longer that the dates of gaming cards showing games of FF quality in real time.
So yes.. since the topic here is FF graphics.. not Star Wars /or Nemo.. :)
1)we are very close to see most of FF in real time.. from 5-10 years..
http://www.arstechnica.com/wankerdesk/01q3/ff-interview/ff-interview-1.html
2)we can be as close as 10-15 years to match the basic of the most advanced that we see in today CG movies in graphics cards.
3)and probably Video cards -single cards-will never catch what photonmapping renderers can do in software (used mostly in digital films) ..since thats not their purpose.
we have pixel shaders for that. and as you say Renderers will have better techniques in the future.but still video cards will be able to fake something of that quality in real time ,in special situations. it is unclear for me which way will be the future of graphics.. pixel shaders programming or Photonmapping? because not only we see DIrectx9 demos ,but now also PM demos? in a few more years ~5 we will have a clear vision of what way it will be ...i will like more the last one aproach ,but still it seems that we are long time away to see that in real time -games- graphics at -every possible scene- at cinematic quality. if you ask Nvidia about this.. you will find that they even have diferent opinions about wich way will be.. maybe in the end ,it will be a mix of both techniques. :)
Simple raytracing, sure. Production CGI level? Not for a lonnng time...
indeed..
but realtime raytracing is not the goal ,for Nvidia (it doesnt look like it is ,not in the near future),doing similar quality like raytracing and photonmaping ..it is.. thats why we have Pixel shaders.. in fact im skeptic about if Nv3O can do -real- photomapping.
it looks to me that their photonmapping demos are more hacks to push the hardware in a very controled enviroment. that only works in special situations.
Here's an example. This (http://www.newtek.com/products/lightwave/tutorials/fullprecision/images/cornell_montecarlo.jpg) image is a radiosity image rendered in LightWave. There are no "traditional" light sources in the shot; all the light comes from a bright spot on the ceiling polygon itself which casts a shadow on the floor (a very soft shadow mind you), illuminates the top of the sphere and the blue and red walls which diffuse light onto the sides of the sphere.This kind of effect is used in many production level CGI shots and has been for the past several years as computers became powerful enough to render such frames in a reasonable amount of time. I keep saying this, but the point doesn't seem to get through, this is an enormously complicated light simulation, so much more so then any of the shaders that games are just barely starting to use now and hardware is going to have to mature a LOT to be able to even consider being able to do something like this in real time. Simple raytracing scenes can be rendered in less than a second on modern CPUs, by contrast radiosity shots can take several hours depending on complexity. There's no magic graphics card that's going to come along any time soon that can do this in 1/30th of a second.
yes ..the effect you are describing in Lightwave is something called FINAL GATHERING
,but dont know how its called in Lightwave since its renderer is a RADIOSITY Engine.
a more advanced technique than raytracing..that it is comparable with Photomapping.
its very complex ..indeed.. however it will be good to know that Photonmapping and Radiosity the most advanced techniques in CG used in today movies are very carefully used in few scenes (IF ANY) in 100% CG movies.. not because of the quality ,which is the best a computer can do ,but because of the Long TIme it takes.
it will be good to classify movies in two cathegories..
1)CG movies.. (like toy story/ANts/SHrek/ a bugs life/FF/Monsters/Nemo
2)and Digital FIlms.. that use real camera shots ,with some scenes of CG graphics..
Starwars,lord of the rings,Hulk ,Matrix..Terminator
the first group almost doesnt use photonmapping.
PIXAR movies..Toys STory1/2 use ->[nothing ] of radiosity or raytracing,same with 95% of Pixar movies ,same with Square FInal fantasy. ;) Shrek/Ants/ a bugs life? . however MOnters use it at some shots..finding NEMO, maybe more.. Star wars:atack of the clones is a diferent story :) but most of that movie is real life film with real people and scenarios ..not like FF ..yes nothing can match the real thing ;)
Lord of the RIngs use it some times ,only in some shots -and this is what people dont know .. that most CG movies use rendering techniques cheapier and faster than phtonmapping and radiosity. ;) Pixar REnderman support for Photonmapping only happened last year. and still they will tell you that Photonmapping is something that should be touched with a 10 foot Rod,in big films ,avoided whenever its possible. thats why Nvidia or ATI can claim Cinematic CG quality in real time.. since most of the movies that are CG use SCanline rendering like games.. ;) its just that the shaders in movies are very high quality and complex.. now you know all the fuss and all the hype for very long instructions shaders in directx9 Gpus.. because they are necessary for cinematic quality ;)
photonmapping renderers like this..
http://www.finalrender.com/products/products.asp?UD=10-7888-35-788&PID=36
http://www.splutterfish.com/sf/gallery_index.php
http://www.mentalimages.com/4_1_motion_pictures/index.html
have more advanced lighting techniques that most movies that are 100% CG. like Pixar movies.. the day to see a 100% CG holywood movie ,like FF but using photomapping -at all times- can be even longer that the dates of gaming cards showing games of FF quality in real time.
So yes.. since the topic here is FF graphics.. not Star Wars /or Nemo.. :)
1)we are very close to see most of FF in real time.. from 5-10 years..
http://www.arstechnica.com/wankerdesk/01q3/ff-interview/ff-interview-1.html
2)we can be as close as 10-15 years to match the basic of the most advanced that we see in today CG movies in graphics cards.
3)and probably Video cards -single cards-will never catch what photonmapping renderers can do in software (used mostly in digital films) ..since thats not their purpose.
we have pixel shaders for that. and as you say Renderers will have better techniques in the future.but still video cards will be able to fake something of that quality in real time ,in special situations. it is unclear for me which way will be the future of graphics.. pixel shaders programming or Photonmapping? because not only we see DIrectx9 demos ,but now also PM demos? in a few more years ~5 we will have a clear vision of what way it will be ...i will like more the last one aproach ,but still it seems that we are long time away to see that in real time -games- graphics at -every possible scene- at cinematic quality. if you ask Nvidia about this.. you will find that they even have diferent opinions about wich way will be.. maybe in the end ,it will be a mix of both techniques. :)