I have recently setup Cinelerra-gg on a new PC (Ryzen 7 5700G with 32gb), with Kubuntu 21.04 and the latest Appimage of Cinelerra.
When I render a video, CPU usage mainly stays around 25%, sometimes briefly going above 50%.
If I transcode with Handbrake, CPU usage goes up to 98%. Why is this? Have I inadvertently changed a setting that has caused this?
I have to say that rendering does seem very slow and is not a huge improvement over my 8 year old i7 machine.
It is possible ffmpeg is not configured.
you should increase the number of threads in the ffmpeg configuration
threads=4
it depends if you have threads available.
@glitterball3 I have seen this sometimes as well. It does not always use full CPU. Which video format are you encoding to? Are you using HW acceleration in encoding? E.g. one of the encoding with vaapi?
What seems to help it setting up a render farm on the same machine. Basically you set up and start say 8 render clients. The rendering is split in blocks over the 8 instances, and the results are combined. Read the manual on how to set it up.
As already mentioned the solution in these cases is the RenderFarm. See also on Bug Tracker, where a good user has created a script that works together with the renderfarm:
https://cinelerra-gg.org/bugtracker/view.php?id=575
Thanks for the replies.
Can anyone explain why it takes 4x longer to render the video, when the video will playback at the full framerate in the timeline with effects applied?
I havent managed to get the render farm to work yet - it throws up failed to start errors.
I am not using any HW acceleration, as I was able to get a faster playback framerate using none.
I am encoding to ProRes, but encoding to h264 is the same (i.e. does not use anywhere near full CPU).
@sparkill Thanks for the tip: I didnt even know about that setting (the default was 1), however changing it does not make any difference to the render time or CPU load.
@sparkill I also think its because of CinGGs internal ffmpeg; but changing the threads=… parameter never leads to any change in rendering. From my tests doing the same rendering with CinGG and with external ffmpeg shows a very big gap in CPU threads usage and rendering time, all in favor of external ffmpeg.
It is possible to compile CinGG to work with external ffmpeg, but Ive never been able to.
Can anyone explain why it takes 4x longer to render the video, when the video will playback at the full framerate in the timeline with effects applied?
Is the Video Driver set to X11-OpenGL? If so, then it will play a lot faster. Rendering is done solely with software so it does not take advantage of the graphics card.
BTW: my laptop (with 8 CPUs) renders faster than the desktop Epyc with 128 CPUs. But if the desktop takes advantage of the render farm, a 1 hour video can be done in less than 10 minutes.
@phylsmith2004 I have the video driver set to X11-XV. It doesnt make sense to me - and makes me think that there must be a software-imposed bottleneck in the rendering process somewhere.