Announcement

Collapse
No announcement yet.

Realtime Dynamic GI + Reflections + AO + Emissive plugin - AHR

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    #76
    How many bounces do you have? one or 2? Does it work with dynamic meshes or objects or characters? Also how faar can you apply it in open environment with mountain like settings?

    If you can, can you please tell me what is the difference between your technic and this: https://www.youtube.com/watch?v=_E1oVl2d01Q
    http://www.geforce.com/hardware/tech...xgi/technology

    https://developer.nvidia.com/gi-works (Dynamic specular is good in the dinosaur scene).


    And this : https://forums.unrealengine.com/show...y-supports-UE4

    And this (made using Cryengine though) : https://www.youtube.com/watch?v=8PCnkJBvIrY

    You can download the demo and try it , look in the description and see hsi channel for other videos. I noticed that the reflections caused by those emissive balls when applied on the spheres are blocky and look like vulgar pixels.
    [ATTACH=CONFIG]16047[/ATTACH]
    I hope we won't get those artifacts with your plugin. He has 2 bounces though. The Tomorrow Children uses a tech with 3 bounces and 16 light tracing directions per voxel. You can read about it here:

    http://www.dualshockers.com/2014/08/...hots-released/

    http://www.dualshockers.com/2014/08/...e-info-shared/

    http://www.dualshockers.com/2014/09/...p-screenshots/

    http://www.dualshockers.com/2014/09/...-ms-per-frame/

    http://www.dualshockers.com/2014/10/...by-new-videos/

    http://www.dualshockers.com/2014/10/...lighting-tech/

    http://www.dualshockers.com/2014/10/...-and-gameplay/

    http://www.dualshockers.com/2014/10/...nfo-next-week/

    http://www.neogaf.com/forum/showthread.php?t=918421

    http://blog.us.playstation.com/2014/...g-ps4-visuals/

    They are using Octane Render :http://www.neogaf.com/forum/showthread.php?t=887777

    You can download PDF here : http://fumufumu.q-games.com/archives/2014_09.php#000934

    To be honest, I am very interested in your tech since it has the best ratio quality/performance. Keep the good work man

    Edit: GDC 2015 PDF is out: http://fumufumu.q-games.com/archives...ldrenFinal.pdf
    Attached Files
    Last edited by Nudlegaru; 04-09-2015, 02:45 PM.

    Comment


      #77
      Sorry for taking this long, been having little spare time. Is one bounce, but could be extended to multiple bounces. Performance is probably a problem tough.
      At start it will work only for dynamic and static objects, but particles and others could be added later.
      Regarding NVIDIA's technique, they are quite similar, with the fundamental difference that I use raytracing instead of conetracing. I also have a way more simple voxelization pipeline and I use just one binary Voxel grid, the irradiance gets reconstructed using some dark magic that I can't talk about for now

      Comment


        #78
        Originally posted by RyanTorant View Post
        Sorry for taking this long, been having little spare time. Is one bounce, but could be extended to multiple bounces. Performance is probably a problem tough.
        At start it will work only for dynamic and static objects, but particles and others could be added later.
        Regarding NVIDIA's technique, they are quite similar, with the fundamental difference that I use raytracing instead of conetracing. I also have a way more simple voxelization pipeline and I use just one binary Voxel grid, the irradiance gets reconstructed using some dark magic that I can't talk about for now
        Haha cool! Looking forward to try out your version

        Comment


          #79
          Originally posted by RyanTorant View Post
          Sorry for taking this long, been having little spare time. Is one bounce, but could be extended to multiple bounces. Performance is probably a problem tough.
          At start it will work only for dynamic and static objects, but particles and others could be added later.
          Regarding NVIDIA's technique, they are quite similar, with the fundamental difference that I use raytracing instead of conetracing. I also have a way more simple voxelization pipeline and I use just one binary Voxel grid, the irradiance gets reconstructed using some dark magic that I can't talk about for now
          Question.
          Wouldn't Cone tracing (or rather Sphere Tracing), improve performance ? This way you could probably get away with smaller amount of traces.
          https://github.com/iniside/ActionRPGGame - Action RPG Starter kit. Work in Progress. You can use it in whatever way you wish.

          Comment


            #80
            Originally posted by iniside View Post
            Question.
            Wouldn't Cone tracing (or rather Sphere Tracing), improve performance ? This way you could probably get away with smaller amount of traces.
            I'm not sure really, as the few demos I tested about Voxel cone tracing where slower than AHR, sometimes AHR running even twice as fast. It might have to do with the way tracing works, using a binary grid to do the tracing. On the perfect case( a ray going on a straight Voxel line up) each sample loads 32 voxels

            Comment


              #81
              Quick update.
              I worked all the weekend, but made progress. Got most of the core classes working, and some of the shaders too. It's starting to come together

              Comment


                #82
                Originally posted by RyanTorant View Post
                Quick update.
                I worked all the weekend, but made progress. Got most of the core classes working, and some of the shaders too. It's starting to come together
                Awesome! So looking forward to playing around with this
                Twitter | dylan@incandescentimaging.net | View Caffeine On IndieDB

                Comment


                  #83
                  Originally posted by RyanTorant View Post
                  Quick update.
                  I worked all the weekend, but made progress. Got most of the core classes working, and some of the shaders too. It's starting to come together
                  Great to hear! Really looking forward to see more.

                  Comment


                    #85
                    Nvidia released their useless Apollo tech demo: http://blogs.nvidia.com/blog/2014/11...l-apollo-demo/
                    http://www.nvidia.com/coolstuff/demos#!/apollo-11


                    It needs at leats a GTX 970 to run while I have a GTX 660M -_-.While the tomorrow children uses much better tech with ray traced voxel cones with 3 light bounces and in 16 directions per voxel while this nvidia tech only has one bounce. ALso tomorrow children has 3 types of AA and the list of the awesome features is long (I gave you links before) and this a game not just a demo. Morever PS4 is not a GTX 980. A good way for Nvidia to sell their overpriced cards that only 2 customers will buy and a good way to deprive us from using any type of dynamic GI . I don't blame epic though, they are good devs. RyanTorant, the ball is in your camp now. You are the only and last savior.

                    Comment


                      #86
                      Originally posted by PLASTICA-MAN View Post
                      Nvidia released their useless Apollo tech demo: http://blogs.nvidia.com/blog/2014/11...l-apollo-demo/
                      http://www.nvidia.com/coolstuff/demos#!/apollo-11


                      It needs at leats a GTX 970 to run while I have a GTX 660M -_-.While the tomorrow children uses much better tech with ray traced voxel cones with 3 light bounces and in 16 directions per voxel while this nvidia tech only has one bounce. ALso tomorrow children has 3 types of AA and the list of the awesome features is long (I gave you links before) and this a game not just a demo. Morever PS4 is not a GTX 980. A good way for Nvidia to sell their overpriced cards that only 2 customers will buy and a good way to deprive us from using any type of dynamic GI . I don't blame epic though, they are good devs. RyanTorant, the ball is in your camp now. You are the only and last savior.
                      I'm not sure what they're trying to achieve with their VXGI work - if they are seriously going to stick with their 'Maxwell only' line then no one will ever use it for games. You would always need to have a fallback to Lightmass for AMD / non-Maxwell users which are going to be about 98% of your playerbase, and for that 2% the extra work doesn't seem worth it. And it won't be supported on consoles because PS4 and Xbone are AMD APU's.
                      Gangster Weapon Pack | Free Weapon Pack | Various Free Models

                      Comment


                        #87
                        I'm gonna sounds like a total old fart, but I remember when we had to set up our bounces by hand even in non-realtime renders and then wait overnight to see if the single frame turned out how we hoped. MAN, we've come a long way baby
                        Trevor Lee

                        Comment


                          #88
                          Don't be so fast at judging Nvidia for the Apollo tech demo. People have said the same thing every new GPU generation, that stuff like this is useless because only one manufacturer or model supports it. Then 1-2 generations later, everyone uses it. This is R&D for new hardware to show the potential for the future, not necessarily what we're going to use at this exact minute.
                          Programmer, Rigger, Animator - Demo Reel 2015 - My RnD / WIP thread - www.dennylindberg.com
                          Worked on: Kitten'd - Guardians of Valor (Android, iOS)

                          Comment


                            #89
                            Originally posted by RyanTorant View Post
                            I'm not sure really, as the few demos I tested about Voxel cone tracing where slower than AHR, sometimes AHR running even twice as fast. It might have to do with the way tracing works, using a binary grid to do the tracing. On the perfect case( a ray going on a straight Voxel line up) each sample loads 32 voxels
                            Hm.Excuse my noobness. I was thinking along the lines that you can shoot spheres in cone, and just using single cone trace gather all voxels it touches. So for example single sphere can gather 64 voxels in sphere of influence while line will be always limited to gathering things only in line.

                            That being said you would probabal want more cone traces in more or less the same direction to gather more accurate information to average diffuse bounce (though diffuse doesn't need to be terribly accurate).

                            Or it doesn't work this way ?
                            https://github.com/iniside/ActionRPGGame - Action RPG Starter kit. Work in Progress. You can use it in whatever way you wish.

                            Comment


                              #90
                              Originally posted by iniside View Post
                              Hm.Excuse my noobness. I was thinking along the lines that you can shoot spheres in cone, and just using single cone trace gather all voxels it touches. So for example single sphere can gather 64 voxels in sphere of influence while line will be always limited to gathering things only in line.

                              That being said you would probabal want more cone traces in more or less the same direction to gather more accurate information to average diffuse bounce (though diffuse doesn't need to be terribly accurate).

                              Or it doesn't work this way ?
                              The "gather" part is the problem. Loading the data from the memory (aka sampling the voxels) is what takes most of the time, and even if you change the shape of the sampling, you still need to load the data at some point. Sadly it doesn't work like you said, or maybe I'm misinterpreting you.

                              BTW, the forums seems to have unsubscribed my from the thread for some reason. **** computers
                              I just wanted to say I'm still working on this, and I'm getting there, but I'm about to start the exams period at college and that goes until 20 of December, give or take a few days, so progress will be slow for a while.

                              Oh BTW, I wanted to talk a bit about memory usage.
                              For a single 512x512x512 voxel grid, wich is what I'm using now, the memory usage is 16 MB. To that you need to add the memory of a RSM per light, but that's not usually taken into account when talking about voxel-based GI algorithms, so I think it's ok to talk about just 16 MB of memory usage.
                              That's not taking into account emissive materials.
                              A quick breakdown will be like this:

                              - Just AO : Only the Binary Voxel Grid, so 512*512*512*4/32 bytes = 16 MB
                              - AO + GI: Binary Voxel Grid : 16 MB + RSM ( Depth + 32 bits per pixel buffer). About 32 MB on my demo, so looking at 48 MB total
                              - AO + GI + 16 Emissive materials : 16 MB + RSM + 512*512*512*4/8 = 64 MB. So a total of 112 MB
                              - AO + GI + 256 Emissive materials : 16 MB + RSM + 512*512*512*4/4 = 128 MB. A total of 176 MB

                              When I implement cascaded voxel grids, the memory will probably be kept at the same 16 MB, maybe 32 MB in the worst case
                              Last edited by RyanTorant; 11-16-2014, 05:56 PM.

                              Comment

                              Working...
                              X