Jump to content


Camera change makes difference to render time


  • Please log in to reply
7 replies to this topic

#1 jim c

jim c

    Illusionist

  • Members
  • PipPipPip
  • 385 posts
  • Joined: 01-November 08
  • Name:Jim Crafton

Posted 22 May 2012 - 07:29 PM

OK, I know, at first this sounds obvious - duh, change the camera, change what you're looking at, and yeah, I'd expect the render time to change. However I have one camera that includes most of my objects (maybe 350K polys) plus several that use displacement maps. This takes about 1 minute or less - good stuff.

Then I switch to another camera that is closer in, and of the 8 or 9 different objects visible in the first camera, only 2 are visible here, only one of which is using a displacement map. This takes over 3 minutes to render, with most of the time spent spinning mantra's wheels doing something, then all of a sudden it completes.



The lights using Deep Shadows are all loading the shadow map from disk, not generating them.

I tried using the Performance monitor (which by the way is really quite cool!), but it doesn't seem to show anything specific that's chewing up time. I'm sampling everything but Node Draws (GPU) and Thread Stats.

Any ideas what I might look for? I have the nasty suspicion that I've done something incredibly stupid, but I'm not seeing it.

#2 eetu

eetu

    Illusionist

  • Members
  • PipPipPip
  • 479 posts
  • Joined: 30-May 07
  • Location:Helsinki, Finland
  • Name:eetu m

Posted 22 May 2012 - 10:17 PM

My first guess would be that the displaced object's displacement bounds are a lot bigger, when measured in pixels, and that slows it down.
See if you could tighten the displacement bounds without getting render errors?
A shitty theory is better than no theory at all

#3 ChristianW

ChristianW

    Illusionist

  • Members
  • PipPipPip
  • 340 posts
  • Joined: 10-November 10
  • Location:London
  • Name:Christian Waite

Posted 22 May 2012 - 11:58 PM

Which render method are you using? Raytrace or Micropolygon?

#4 jim c

jim c

    Illusionist

  • Members
  • PipPipPip
  • 385 posts
  • Joined: 01-November 08
  • Name:Jim Crafton

Posted 23 May 2012 - 06:01 AM

I'm using the Micropoly renderer.

#5 ChristianW

ChristianW

    Illusionist

  • Members
  • PipPipPip
  • 340 posts
  • Joined: 10-November 10
  • Location:London
  • Name:Christian Waite

Posted 23 May 2012 - 06:30 AM

Swap to raytrace, see if it's any faster...

#6 Erik_JE

Erik_JE

    Illusionist

  • Members
  • PipPipPip
  • 377 posts
  • Joined: 12-January 10
  • Location:Stockholm, Sweden
  • Name:Erik Johansson-Evegård

Posted 23 May 2012 - 07:25 AM

View PostChristianW, on 23 May 2012 - 06:30 AM, said:

Swap to raytrace, see if it's any faster...

It will be. At least to start the render but not necessarily finishing it. But as eetu said check the Displacement Bounds it can make a huge difference.
This is either a really smart move or by far the stupidest thing that we have ever tried.

#7 jim c

jim c

    Illusionist

  • Members
  • PipPipPip
  • 385 posts
  • Joined: 01-November 08
  • Name:Jim Crafton

Posted 23 May 2012 - 03:33 PM

Thanks guys! eetu, thanks that was exactly it. I had the scale at 1, and the default setting for the bounds was to follow the scale. Deleting the channel and changing it to 0.05 seemed to work, any smaller and I could see cracks in the render. There appears to be no difference in rendering in the original camera with 1 or 0.05 for the bounds, so what exactly is it doing? Is there really any reason you would want it at 1?

#8 Erik_JE

Erik_JE

    Illusionist

  • Members
  • PipPipPip
  • 377 posts
  • Joined: 12-January 10
  • Location:Stockholm, Sweden
  • Name:Erik Johansson-Evegård

Posted 23 May 2012 - 11:26 PM

View Postjim c, on 23 May 2012 - 03:33 PM, said:

Thanks guys! eetu, thanks that was exactly it. I had the scale at 1, and the default setting for the bounds was to follow the scale. Deleting the channel and changing it to 0.05 seemed to work, any smaller and I could see cracks in the render. There appears to be no difference in rendering in the original camera with 1 or 0.05 for the bounds, so what exactly is it doing? Is there really any reason you would want it at 1?

If your displacement had a higher amount or amplitude you would need a bigger displacement bound.
This is either a really smart move or by far the stupidest thing that we have ever tried.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users