Optimizing DR

[NicoC]

Hello

I have a question about DR and scene loading. To summarize our needs, we most of time render between 10 - 40 images for a single project, a single scene. Depending on the urgency, we either submitt all the images to BB to get each rendernode rendering one image, or we submitt all the image to only one BB Server which will use the others' spawners using DR. In this second option, would there be a way add an option to DR, so that DR nodes keep the same scene for each rendering, but somehow "receive the information" that the only thing that changes is the camera ? I don't know if my explanations are clear :P

Any help would be appreciated

Thanks !

[tashko.zashev]

Hi,

Could you share some more details about your workflow in order to be in a better assistance.
Is your camera animated or you need to render different camera views only?

[NicoC]

Hello

No animated camera, only multiple still ones. I guess we could use only one camera and animate it from frame to frame, to get our multiple point of views, but it turned out it wasn't really flexible for us so ...

[Neilg]

You could pre calculate the irradiance map & light cache, reusing it for all the views.
40 separate images is more a job for backburner rather than DR though. DR is when you want all your power on one image, network rendering is to spread power between as many separate tasks as possible.

[tashko.zashev]

Hello

No animated camera, only multiple still ones. I guess we could use only one camera and animate it from frame to frame, to get our multiple point of views, but it turned out it wasn't really flexible for us so ...

As Neilg mentioned above, such number of render frames are more suitable for network render instead of DR one. But there is a lot of hand work to setup different jobs with all scene cameras.
Animating one camera to cover all views is an option too, but again it requires a lot of adjustments and additional work.

Perhaps one good approach is to use batch render to setup all render views and send a backburner job from there. Then you will get a queue with all render views, that could be rendered on multiple machines or single one with DR enabled.

[NicoC]

You could pre calculate the irradiance map & light cache, reusing it for all the views.
40 separate images is more a job for backburner rather than DR though. DR is when you want all your power on one image, network rendering is to spread power between as many separate tasks as possible.

Hello Neilg
Yes, sure, distributing all the images over the BB Servers is an option we use regularly. But depending on the urgency, we switch to BB + DR configuration, thus helping us to get rendered images quicker, so that we can start psotproduction work ASAP.

[NicoC]

As Neilg mentioned above, such number of render frames are more suitable for network render instead of DR one. But there is a lot of hand work to setup different jobs with all scene cameras.
Animating one camera to cover all views is an option too, but again it requires a lot of adjustments and additional work.

Perhaps one good approach is to use batch render to setup all render views and send a backburner job from there. Then you will get a queue with all render views, that could be rendered on multiple machines or single one with DR enabled.

Hello Tashko,

Thanks for your post. Obviously, with the amount of work we deal with, we want/need to minize handwork. That's why we use (the crappy) batch render intensively to submitt all our shots en-masse. Oh, and btw, if Chaos Group has some free dev time, a TRUE batch rendering system would be a killer just kidding ... or not !

Thanks.