Announcement

Collapse
No announcement yet.

nvidia gtx1080 :)

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

  • #16
    I doubt there is a magic page file or another trick that will suddenly allow to render large texture sizes. It seems nvdia is reluctant do up the ram on gtx cards dramatically. It makes sense though, probably quadro type cards will get 16/32 faster, and gtx will crawl behind. Sort of reminds me of how intel handles i7 and xeon. You can't seat two i7 cpus in one board, if you could it would put xeon out of competition.

    I guess its ok though, I don't feel comfortable switching to gpu rendering in next few years anyway - any time I try, there are just too many things that needs to be improved / implemented still.
    Dmitry Vinnik
    Silhouette Images Inc.
    ShowReel:
    https://www.youtube.com/watch?v=qxSJlvSwAhA
    https://www.linkedin.com/in/dmitry-v...-identity-name

    Comment


    • #17
      The new Pascal architecture is supposed to have true "virtual" GPU memory where the GPU can transfer on demand blocks of memory from the system RAM to the GPU, pretty much as the regular CPU can page out to disk, without having to do special changes to the GPU code. I don't know how this will work in practice, or whether it will be enabled on GTX, but it sounded exciting on the slides

      Additionally, using NVLink, the memory of multiple GPUs can be linked together and added, so if you have two GPUs with 8 GB, they can share each other's memory. However this seemed to a feature reserved for the Quadro and Tesla cards, not sure if it will make it to the GTX's.

      Best regards,
      Vlado
      I only act like I know everything, Rogers.

      Comment


      • #18
        Originally posted by vlado View Post
        The new Pascal architecture is supposed to have true "virtual" GPU memory where the GPU can transfer on demand blocks of memory from the system RAM to the GPU, pretty much as the regular CPU can page out to disk, without having to do special changes to the GPU code. I don't know how this will work in practice, or whether it will be enabled on GTX, but it sounded exciting on the slides

        Additionally, using NVLink, the memory of multiple GPUs can be linked together and added, so if you have two GPUs with 8 GB, they can share each other's memory. However this seemed to a feature reserved for the Quadro and Tesla cards, not sure if it will make it to the GTX's.

        Best regards,
        Vlado
        I did two 980 ti sli.But they didn't merge rams I missed this : (
        http://facebook.com/Avisgrafik

        Comment


        • #19
          Vlado and Blago, when do you think you guys will get your hands on one of these to do some tests?
          Kind Regards,
          Morne

          Comment


          • #20
            I have no idea when they will be available here to buy... will be some weeks I guess.

            Best regards,
            Vlado
            I only act like I know everything, Rogers.

            Comment

            Working...
            X