Announcement

Collapse
No announcement yet.

FStorm vs Redshift

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

  • FStorm vs Redshift

    Hello ! Somebody makes a comparison Fstorm RedShift entirely on one and the same scene? It would be interesting to know the results .. Particularly interested in the difference in speed, light shadows and other things.
    Thanks!

  • #2
    http://zchen.ca/gpu/
    FStorm 1.3.0d I Max 2018 I Win7 I i7 5930k I 64GB I 3 x 1080Ti

    Comment


    • #3
      He should use FStorm portals. They are very powerful.
      Attached Files

      Comment


      • #4
        Originally posted by Karba View Post
        He should use FStorm portals. They are very powerful.
        Better use portals instead of hidden lights in windows (if I remember right the scene has hidden fstorm lights)? Seems like...
        FStorm 1.3.0d I Max 2018 I Win7 I i7 5930k I 64GB I 3 x 1080Ti

        Comment


        • #5
          Originally posted by RobSteady View Post
          Better use portals instead of hidden lights in windows (if I remember right the scene has hidden fstorm lights)? Seems like...
          I didn't find any

          Comment


          • #6
            Originally posted by Karba View Post

            I didn't find any
            But there was a portal - at least when I downloaded the scene.

            FStorm 1.3.0d I Max 2018 I Win7 I i7 5930k I 64GB I 3 x 1080Ti

            Comment


            • #7
              Ok, first of all, there's a portal in the scene (max 2016 version for sure) but...this epic fight can't be closed with a simple "use the light portal!"
              104 min v 7.54 min (nearly noise free) can't be an usable result (maybe?) so...redshift users that use fstorm too could better compare pro and cons of these two renderer.
              My profane opinion: incredible time render speed for redshift (GI light seems not so good) and incredible light/tuning for fstorm (fast but not fastest).

              Comment


              • #8
                Originally posted by savat View Post
                My profane opinion: incredible time render speed for redshift (GI light seems not so good) and incredible light/tuning for fstorm (fast but not fastest).
                I haven't used RedShift myself yet, but... so far the images I have seen have that something that just don't feel right to me, that lumion look to them, is subjective

                I have read people saying that RedShift is kind of a VR running on GPU but at the same time I haven't seen a single RS image that can compare to lots of VR images produced even by amateurs and 3d beginners, and maybe this is subjective too, but I prefer a lot the FStorm lighting over the VR anytime and I'm using a FStorm veta, not even the last veta... so I guess it depends
                if you are ok with the RS results, and plus it's near 10 times faster than FStorm or anything, I would go blindly with RS and ignore all the comments about how bad or how fake it looks or anything else bad about it

                for exapmple, long time ago I had a work for a bowling chain, and RS would have been perfect

                Comment


                • #9
                  Ok, now I see it. Sorry
                  it is showed as a box, this is why I missed it

                  What I changed is GI clamp to 0.2, noise threshold to 0.02 and remove environment importance sampling (useless with constant environment)
                  GI clamp 0.2 is the reason why it has so less noise in my render then. it decreases render time from 104 minutes to about 30

                  The main difference is FStorm doesn't use any interpolation like Light Cache.
                  VRay and RedShift use Light Cache.

                  Comment


                  • #10
                    Originally posted by Iskánder ivp87 View Post

                    ...
                    if you are ok with the RS results, and plus it's near 10 times faster than FStorm or anything, I would go blindly with RS and ignore all the comments about how bad or how fake it looks or anything else bad about it

                    ...
                    Less unbiased and "only" 5 time faster could be a good recipe. "10 time faster then vray" is what people says about the brute force in Arnold, so...this could be a good goal for the unbiased roadmap that Karba wants to follow

                    Comment


                    • #11
                      bump map on the floor adds too much noise imho (slow-down?) + visually I even prefer it without (there may be a midde-ground found there...)
                      Click image for larger version

Name:	315s_nobump_7m54s_1440.jpg
Views:	1
Size:	220.7 KB
ID:	8370 Click image for larger version

Name:	clamp0.1_7m53s_1440.jpg
Views:	1
Size:	255.0 KB
ID:	8371
                      (both images are 7m53s on gtx1070 as in the write-up, only resized (h1440) for the forum)
                      Last edited by antonyebl; 12-30-2016, 09:11 PM.

                      Comment


                      • #12
                        Originally posted by antonyebl View Post
                        bump map on the floor adds too much noise imho (slow-down?) + visually I even prefer it without (there may be a midde-ground found there...)
                        [ATTACH=CONFIG]n8370[/ATTACH] [ATTACH=CONFIG]n8371[/ATTACH]
                        (both images are 7m53s on gtx1070 as in the write-up, only resized (h1440) for the forum)
                        maybe could you try to lower the bump amount?

                        Comment


                        • #13
                          Originally posted by Iskánder ivp87 View Post

                          maybe could you try to lower the bump amount?
                          1mm bump | 0.5mm bump (7m53 ~315samples)
                          Click image for larger version

Name:	315s_bump_1mm_7m55s_1440.jpg
Views:	1
Size:	238.3 KB
ID:	8381 Click image for larger version

Name:	314s_bump0.5mm_7m53s_1440.jpg
Views:	1
Size:	228.8 KB
ID:	8382

                          6000 samples finished in ~25m
                          Last edited by antonyebl; 01-01-2017, 06:57 PM. Reason: original setting was 5mm bump

                          Comment

                          Working...
                          X