I am currently having a very annoying problem with Shotcut while previewing a video and while exporting. When I apply a lot of filters, my CPU isn’t utilized as much as it should, resulting in choppy audio and playback. For example, if I use the “Sketch” filter in conjunction with the “Rotate and Scale” filter, it only uses 4 threads. I’m not sure if this has something to do with the filters, as when I have a Vertigo effect on the video, more cores appear to be utilized in the Task Manager. Also, while exporting, the CPU starts at around 50% usage, but then it drops down to 10% after a few seconds, drastically slowing down the export speed. I tried different codecs, but they all had the same result: started out fast, but then slowed to a crawl. Changing the number of export threads didn’t seem to do anything. My CPU is an Intel Xeon E5-2680 V4, a 14-core CPU with an all-core turbo boost speed of 2.9Ghz, so it should have no problem rendering videos. The video I am editing in question is only a few minutes long, and I believe that it is 1080p @ 30fps.
Huananzhi X99-8M Motherboard
Intel Xeon E5-2680 V4
32GB Registered ECC DDR4
AMD Firepro S10000 12GB Passive Edition
Zorin OS Lite Version 15.3
Some filters(I’m not sure about the ones you mentioned specifically) are single threaded, also “Rotate and Scale” sounds like an older version of shotcut(it was merged recently into a size/position/rotate filter) Can you confirm what version of shotcut you’re using?
On an unrelated note how are you liking the Huananzhi board? I’ve been considering doing a dual E5-2690 build and had looked at the HUANANZHI X79 16D
You can try turning on parallel processing in Export > Advanced > Video but there is a low risk that it will result in some incorrectly rendered frames or go slower if the source files do not seek quickly.
I see. Thank you very much for the information. Is there a way to optimize it the most I possibly can? The video I am editing was shot on an iPhone 6, and it is stored on my computer as a .mov file. Shotcut says the codec is “H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10,” the resolution is 1080p @ 30fps, and the format is “yuv420p.” I am exporting the video file as a .mp4.
By “Rotate and Scale,” I meant to put “Size, Position, and Rotate,” which is what is being used in the project. I also found something interesting. If I put the “Sketch” filter on top of the “Size, Position and Rotate” filter, the preview is very laggy with lots of dropped frames. The odd thing is that it is not maxing out any cores when it does this, it just has very low CPU usage across all cores with a few being higher than the others. However, if I move the Sketch filter below the other filter, the video plays back much more smoothly and CPU usage is much higher across all cores. I also found that the low CPU usage also occurs when I use the “Halftone” filter by itself.
I do have parallel processing checked, and I also noticed that when I exported the video without any filters, CPU usage was very high across all cores, and it went very fast. However, when I applied the Halftone filter and exported it, the difference was extremely substantial, as barely any CPU cores were being used.
I am curious as to why certain filters would be able to utilize more resources than other filters. Is it a limitation of the program itself, or is it an optimization problem?
The version of Shotcut I am using is 20.09.13.
As for the motherboard question, the Huananzhi X99-8M has been holding up well. I only got it about a month ago, but I’ve never had any issues with it. My only complaint is that the audio quality isn’t the greatest, but other than that, it’s been a good motherboard.
That is very interesting; I thought that all of the filters were written by the same group of people. Now it makes more sense how the filters work together in weird ways. So is there really nothing else I can do besides checking parallel processing to increase multicore usage?
Another thing worth checking is if you’re using hardware acceleration(your gpu) since some filters are(and some aren’t) gpu accelerated as that can cause odd looking cpu usage(low cpu because the gpu is fully loaded) as well.
Thank you for the information. Since the FirePro S10000 I have is completely passive, I can easily tell when it is at full load based solely on the temperature. However, the temperature stayed virtually constant when playing back the video, so the GPU wasn’t being bottlenecked at all. The one thing I did find is that playback can either be smooth or choppy depending on the order I put the filters in, so I will experiment with that and see how it goes.
Also, on an unrelated note regarding your question about the Huananzhi board, there is one last thing that I forgot to mention: If you choose to build the dual E5-2690 system, be sure that the RAM you use is compatible with the motherboard. I bought normal desktop DDR4 (Mushkin Silverline 2400MHz DDR4) thinking that it would work. Unfortunately, the system would not POST. I thought that the motherboard was dead, but I read in online forums that those motherboards will only support RAM modules with a certain amount of memory banks. In fact, in almost every single online video of X99 and X79 Huananzhi motherboards, people were using ECC RAM instead of normal desktop memory. I ordered 16GB of 2133Mhz Hynix REG ECC DDR4 to be on the safe side, and the system finally posted without any issues with the new RAM installed.
Good luck with the filters, I wish I could help more but “sketch” isn’t one i’ve used at all. H264 MOV files should edit fine but it might be worth trying a different codec, I generally make HUFFYUV files for editing myself, it may also be worth trying a different OS if you can this could be a linux specific issue
As far as the x79 I’m not worried on the memory front, I have 192Gb of DDR3 ECC ram(and a spare 2690) already, i’d be gutting my dual x5690 T7900 for parts(otherwise i’d go with x99 but since I have so much ECC DDR3 I plan to use it) working with xeons is part of my dayjob