Hi
We are having a lot of problems rendering heavy scenes with Solaris and Redshift - a big part of the rendered frames show artefacts or "glitches" which make them unusable. These seem completely random and on re-rendering usually they will finish. So far we have been talking to Redshift about those, but the following might seem like a houdini/husk issue:
The glitch I am talking about is attached. The beauty (rgba) pass will look completely fine, but all other passes show this streaky behavior, what I believe is the bottom line of the frame stretched across the whole image. The file size is also significantly smaller than good frames.
During rendering Redshift/Husk will save *.exr.part files. When I load these into nuke they are fine - all channels. But the final EXR still can show this strange behavior nevertheless.
Does anybody have an idea what this could be?
We are working with Houdini 18.0.
Cheers
Marco
Full Version: frames ok during rendering, turn bad on save
Only juuuust getting started with rendering in lops with other renderers myself, but just curious, are you able to move to 18.5? Lops in 18.0 was pretty raw, might be worth testing in a more recent build.
Also you might get more answers if you can reduce the issue down to simple repro hip. If nothing else you'll be more likely to get 'me too' posts which can help build your case with support.
-matt
Also you might get more answers if you can reduce the issue down to simple repro hip. If nothing else you'll be more likely to get 'me too' posts which can help build your case with support.
-matt
Thanks for your answer.
We are looking into 18.5 but can't change during the project - and we were using 3rd party builds that weren't available for 18.5 when we started. But yeah, we are getting this a lot everywhere : D
There is not really more to go on unfortunately. The error is so random. The same frame will get rendered fine when requeued - meaning from the same usd file. We can't trace it to a few bad farm nodes. The error happens on both servers as on workstations. This scene (from the one I write about) is the first one that reliably produces this error which prevents us from just requeuing over and over again (right now we cannot render this shot). But we have no clue where it comes from. I can provide the 5GB usd file for this scene - I have to Redshift already if you think that's helpful, but as it contains all the shot content it could only be to a secure sideFX server.
But to me and our IT team it does seem like a Houdini/Husk file write issue when the part exr is fine but the final one shows this glitch.
This is just one of many errors and glitches that we get. So far I thought it was redshift, but this one seemed like it could be Houdini so I was wondering if someone experienced the same without using a third party renderer.
We are looking into 18.5 but can't change during the project - and we were using 3rd party builds that weren't available for 18.5 when we started. But yeah, we are getting this a lot everywhere : D
There is not really more to go on unfortunately. The error is so random. The same frame will get rendered fine when requeued - meaning from the same usd file. We can't trace it to a few bad farm nodes. The error happens on both servers as on workstations. This scene (from the one I write about) is the first one that reliably produces this error which prevents us from just requeuing over and over again (right now we cannot render this shot). But we have no clue where it comes from. I can provide the 5GB usd file for this scene - I have to Redshift already if you think that's helpful, but as it contains all the shot content it could only be to a secure sideFX server.
But to me and our IT team it does seem like a Houdini/Husk file write issue when the part exr is fine but the final one shows this glitch.
This is just one of many errors and glitches that we get. So far I thought it was redshift, but this one seemed like it could be Houdini so I was wondering if someone experienced the same without using a third party renderer.