Chaos Corona for Cinema 4D > [C4D] I need help!

Render times increasing every frame?

<< < (2/5) > >>

TomG:
Sorry for the delay in writing back - we've been looking into this (using another scene that gets slower with repeated renders - but of course we can't confirm it's the same issue as you have reported, since unfortunately we can't access your scene). For our reference, (Internal ID=427033657)

Some questions for you, if you don't mind
- Which version of Cinema 4D, which version of Corona, which OS?
- What are the Corona Team Render settings? Are they default, or did you modify them at all?
- What speed is your internal network, 1GB or 10GB?
- Does the scene have high geometry, e.g. something like grass that isn't using instancing at all, or similar?


We understand you may not have time to do any testing, but here are some tests that would provide useful information for us

- Does the same scene show this slowdown if it is rendered locally, and not using Team Render? Or does the slowdown only occur using Team Render?

- If it is Team Render related, it could be that data is "backing up" on the network, which might explain the bell curve (first renders are fine, as the backlog increases it takes longer to have a frame marked as complete, and at the end that backlog begins to catch up again). Since these are large resolutions, that's not impossible. You could try swapping to "Abritrated (experimental)" which is intended to help with any network congestion. If that alone doesn't help, you could try leaving it set to Arbitrated and increasing both the client update interval and the Maximum size of packet (so that larger packets of data are sent, but less often) - for larger resolutions, packet sizes of 256 or 512 may be useful, and update intervals of 20 or 30s may be useful

- If you have any information on what part takes longer, that would be great, e.g. is the parsing time for the scene getting longer with each render? Or is that staying constant, but the rendering stage is taking longer? This could be seen in the VFB stats tab.

Also, the VFB render stamp (which can be added as a multipass render element in the latest versions of Corona 5, and saved separately by enabling saving of all passes in the C4D save set up) can be used to show parsing time, render time, memory use, number of primitives being rendered, and so on. In case you are able to look into that, here's the render stamp I use in my tests that includes the relevant info, you can copy and paste:
Corona Renderer | Corona Mem %mcp | Corona + C4D Mem %mpp | Parsing Time %ptp | Noise %pe | Time: %pt | Passes: %pp | Primitives: %si | Rays/s: %pr

------------------

If you are able to replicate the issue in a scene that is safe to share with us privately, that would of course be awesome. We appreciate your patience, and any additional information you can give (and if you can do any specific tests, very much appreciated!)

rojharris:
Hi Tom,

I mentioned a bit more further down the post but to answer what I can right now:

Cinema 4D R20, Corona 5RC2, Mojave and Win 10... Same problem on both machines. Mac is intel i9, Win is Threadripper. both 64GB RAM

Same thing happens with and without team render... but only on animation... stills are fine.
Same happens on Rebus farm too!

Scene is a fairly simple product model.. Not polygon heavy and only a few instances. I can share it once the product is launched but that may not be for a while.


I've started using Arbitrated for TR but as I say, it makes no difference whether I use TR or not. Just rendering on local machine (either mac or PC) gives same problem.

At the moment I don't know which part of the render is taking the time. I'll do some tests as you suggest later in the week if I get time.

Many thanks,
Roger

TomG:
Thank you Roger for that information, very useful! Especially knowing it happens whether TR is used or not - that increases the odds that it's the same issue as we are seeing in this other scene that also gets slower the more frames that are rendered (that scene, it's the parsing time that gets slower - memory use and primitives stay the same). We do understand that testing is often impossible, as you have all the production work to take care of, so we truly appreciate any that you are able to do. Meantime, this information is very useful, and as I say, it sounds more like the same issue we have a test scene for, so if we find a cause/solution there, we can see if it helps for your scene too.

Thanks again for your patience!

rojharris:
Hi Tom,

Here's a little movie showing the render stamp of the first 20 frames of the animation (cropped from 4k). As you can see, although the parsing does increase by a small amount, it seems to be the rays/sec calculation that's slowing down considerably. Even after 20 frames we're at triple the initial frame time. What's a bit freaky is that the final 20 frames will speed up again in the opposite to this...

Hope it helps.

TomG:
Hi Roger! Again, thanks for that, awesome information! I've grabbed the movie and added it to our internal tracking. We'll let you know as soon as we have any further information, and you can give us a holler if you think of anything else that might be a pointer to what is happening!

Navigation

[0] Message Index

[#] Next page

[*] Previous page