Hi I wrote about this concept a long ago. Now I am back again to use Shotcut to make some works. I learned by myself to build a compiled ffmpeg installation supporting CUDA cores. I wrote that NVIDIA support increase a lot the export speed. I run the consolle to see the video engine at work. Otherwise the decode is still software. The CPU cores are working at 60% and the CUDA engine at 40%. This is great and I am so happy! But now I tryed to compile ffmpeg with the CUDA support for DECODE too. The ffmpeg can read video files and can use CUDA cores to load video in the Nvidia cards memory! I am a miner, so I have CUDA cores for all people living in my village (LoL). I noticed that with the support of the hardware decoding and then the hardware encoding support, we can get: a cpu completely free; the system ram completely free; the cpu fan more relaxed; the silence in our room; and the speed friends! The speed is absolutely umbelieavable! When 2000 CUDA cores run at 100% there is nothin more orgasmic when you are at pc! And I noticed that the NVIDIA temperature is not high! This because the conversion is not an hard work for it. I suggest again, to the developers, that the full support of GPUs is more requested that an usefull interface. For this reason I will use ffmpeg commands to convert my videos. There are some other reason that take to me to the ffmpeg command line: multiple GPU support. In the command line of ffmpeg they developed the support to send jobs to multiple GPU. So if you have to convert 2 video tapes (both 2 hours) you have 4 hours of videos!! You can run two jobs at once without CPU loading!! I have a lot of tapes to convert with denoise filters and it is impossible to make the job with CPU!! Shotcut remains the ONLY software I love!!! But I suggest to develope more nvidia support.
(I am creating an ARM cluster with a lot of Mali GPUs. Ffmpeg supports that gpus too!!)
This topic was automatically closed after 90 days. New replies are no longer allowed.