I have converted a large Tesco store I built in 2019 for Corona. I need to send fly-thru animation clips from this to Ranch Computing render farm and need some advice on optimisation to reduce render times. My first bit I solved, the ceiling lights (there are a lot!). I made 10 large invisible corona light panels to light the store and made a luminated bacterial for the visible lights so there are now only 10 ceiling lights instead of round a 100. All my aisle racks are all instances, does it help render times If I convert them to one solid row instead of a group of many objects and shapes? I never use ‘precomputing’ UHD Cache, only set to ‘Animation (flicker-free),. I did read some time ago about creating a ‘frame stepped’ precomp UHD, then load this in. Does this speed the renders up on a render farm? Would love to get some advice on this .I have never created a pre comp UHD Cache file and struggling to understand how to create this and use. I have attached screen grabs of my set up and some images etc to help.
always a first - make a simple shaded (viewport) animation to communicate the feel of the ‘movie’ to the client and get better understanding of your ‘poiesis’
bake lightmaps on static objects (whatever ain’t moving and/or isn’t subject to lighting changes)
skip frames (then generate inbetweens depending on preferred frame rate)
use ‘fast and soft’ AA/Image filter (can also lower GI/AA balance)
prefer Xref or instances over a number of same object(s)
prefer ‘editable’ over parametric/xpresso/script/plugin driven geo and textures/maps
prefer lower res. maps and geo (know your texel density & LOD)
randomly test parts/sequences for quality assurance
I don’t fully understand this setup/question, but if you are using instances then it’s all good. Things like “100 instances of a single rack vs 10 instances of 10 joined racks” probably wouldn’t make any difference to render times, only RAM usage, but I don’t think it’s worth the effort anyway.
There is no technical reason why this wouldn’t work on a commercial render farm, but to be sure, you would need to contact them and ask whether they support it. If your animation is just a fly-through and it has no moving objects or lights, you should be able to get away with rendering a single cache file and reusing it for all frames. It does make rendering faster as you don’t spend time on cache calculation. You can learn more here: https://support.chaos.com/hc/en-us/articles/4528607789585-How-to-use-UHD-Cache-in-Corona-for-C4D (sorry for the ugly examples, we need to update them )
Hi Guys, thank you so much for your help n advice on this! Have noticed I have quite a few PSD material maps wich I have converted to PNGs. Apart from this I think it’s looking ok. Frames are taking round 4-5min (1920x1080) which is not bad for such a large store. Will defo look at GI/AA lowering and learning the UHD Cache file method. Have a fab weekend
Me again… I have read through your UHD Cache bits. Still don’t quite understand? Do I render the mid frame at 1sec, then save as a PNG and link this as a precomp file?
Sorry, I just noticed that the C4D article is much less detailed than the Max one. We will update it ASAP.
Think of the UHD Cache file as texture map applied to the whole scene saying “here is how light bounces around”.
In all of the below cases, you should enable the “Animation (flicker-free)” preset in UHD Cache settings:
If you have moving objects/lights in your scene (regardless if your camera is static or not): you most likely need to use the default settings (calculate the cache file for each frame).
If it’s a camera fly-through and there are no moving objects or lights whatsoever:
Set “Precomputation” to “Calculate from scratch”
Set “After render” to “Save”
Pick a valid path under “Filename” - don’t specify the file extension, just enter a valid file name like “my cache”
Set render time limit to 1s
Render a single frame from the middle of the animation (it is not so important to render from the middle, however it will be “safest”)
At this point, the cache file should be saved in your specified location under the specified name
Change “Precomputation” to “Load from file”
Change “After render” to “Discard” (there is actually no need to do this, it’s just to keep things tidy)
Render your whole sequence with the desired time/pass/noise limit as you would normally do. Instead of calculating the cache for each frame, it will always be loaded from the specified location.
When and how to use the “load+append” option:
What it does is calculating the cache from scratch first, and then only adding new data where it is needed. The result is a cache file merged from multiple frames, which results in something like motion blur for the bounced light data. It is not very precise, but it can be precise enough to work in many cases, for example when there are some slowly-moving objects or lights in your scene or if the movement distance is minor.
Set “Precomputation” to “Try to load + append”
Set “After render” to “Save to file”
Specify the path to save your cache file
Set your rendering to 1 second only
Go to the first frame of your animation
Set your host application to render every n-th frame
for example, in case of a 400-frames animation, you can set it to render every 50th frame to render the UHD Cache only for 8 frames
the specific n-th frame number depends on the length of the animation and the pace at which the objects are moving - this requires a trial and error approach
the more frames are rendered, the better the quality of the cache will be
Render - this will render n-th frame for only 1 second so almost no actual rendering will be done, but each time new data will be calculated and added to the existing cache file
Once the process of calculating the cache for all frames is done, set Precomputation to “Load from file”
Set your desired render limit (for example 5 minutes per frame, 100 passes, or 2% noise limit)
Set your rendering to full sequence and disable the option to render every n-th frame (be sure to render all frames)
Render
Your full sequence will be rendered using the cache file generated from multiple frames.