My machine is an Hp Z440 Workstation. I’m basically happy with using Shotcut, but I’d like to speed up some processes. For example, playing the timeline to make it more continuous.
Cpu: Intel Xeon E5 2697 v3
Gpu: Nvidia Quadro K5000
Ram: Sk Hynix HMA41GR7MFR8N-TF DDR4 (4*8 Gb)
Storage: Kingstone SA400S37240G (225Gb)
Kingstone SNV2S250G (233Gb)
Operating System: Win10 pro x64
During PC monitoring, the maximum usage is cpu 35%, gpu 30%, ram 50% (of which shotcut 35%).
Videos are edited in 4K resolution.What kind of hardware development could be used to make the timeline continuous, perhaps within a setting program?
Your RAM and hard drive are fine. CPU is a tad slow on base frequency (2.6 GHz) compared to modern 3.4 to 4.2 GHz processors. Are you using Preview Scaling or proxy editing within Shotcut?
Even more important than base frequency is the single thread performance.
For example, compare this Xeon processor to a meager (but modern) i5 processor with a lower base clock speed:
The Xeon processor single thread rating is 43% lower than the i5 processor. Unfortunately, Shotcut has little use for the 28 threads provided by the Xeon processor. In this comparison, I would take the i5 over the Xeon any day.
Hi Brian! I’m curious about the areas where you find single-thread performance to be useful. In my workflow with proxy and preview scaling, I’m often seeing 8 to 12 out of 24 available cores working from Shotcut during preview. I think I remember preview being capped at 4 cores, so I assume the rest are used by decoders. I’m aware of some filters being limited to one thread, such as the wavelet denoiser, but I don’t run into those filters very often. In my case, since I pretty much always get 4+ cores active, I’m not sure I am benefitting from a higher boost frequency. Do you see a different usage pattern for other people? Just wondering if I’m missing an opportunity here. Thanks!
Shotcut will try to use all the cores available. But as a practical matter, it can’t because some threads have to wait until other threads are done before they can work on the next step. My personal experience is similar to yours - Shotcut can usually use 8-12 cores during preview. Maybe more during export.
I do not think the single thread rating is entirely driven by the turbo frequency. It also has to do with cache size and other processing efficiencies. Just look at the overall benchmark of the two CPUs I compared. The i5 has an overall higher performance rating with fewer than half as many cores. Surely those cores are more productive. Shotcut has built in optimizations to use the latest instructions sets when they are available - like AVX-512. The Xeon processor doesn’t support those newer instructions so Shotcut has to use less efficient instructions instead on older CPUs.
There is no preview scaling, it does not work with proxy files.
Thanks for the explanation. I can certainly see instruction set and cache size playing a role.