Selur's Little Message Board

Could elaborate more and give very detailed description how to do that. I really appreciate your time

---

Hello  Im on page 2

What is unclear about what I wrote?

Quote:Create a file with the Vapoursynth script that would be used during encoding and test how fast that would be processed.To do this you would (before creating the job for processing) look at the Vapoursynth Preview (to make sure any used indexing file exists and to check that the script works) and the Vapoursynth Script Preview. Then you would create a test.vpy file with the content of the script shown in the Vapoursynth Script Preview and call:"PATH TO HYBRID/64bit/Vapoursynth/VSPipe.exe" -p "PATH to the test.vpy" NUL in a Windows Command Line prompt.This will do the decoding and filtering of the source without the decoding.While the decoding is running check the cpu utilisation.If the utilisation is low then the used script is the issue. (which is either the fault of one of the filters used, the source filter or the script can't get the source faster)
If the utlilisation is high the speed you see is the max speed you can get and the bottleneck is the encoder.

I suspect its the filtering bottleneck in both vaporsynth and avisynth when doing 60 frames interpolation, so question is: whats the way rounds this then to solve the problem and uncap 20 percent and reach 100 percent CPU usage?

Without knowing what exactly is causing the bottleneck there is nothing I could suggest that makes sense.
Also increasing the cpu usage doesn't mean that the processing will go higher.
Depending on the exact scripts used and the system changing the used cache and threads or using other filters might help. Bu then the issue might not be the filtering or the encoding, could be the decoding, memroy speed, the file access times, some tool interfering. Could also be an issue with the efficiency core setting of the cpu,..

Cu Selur

Here's an example on my system,....

Decoding speed with 4k Video on my system (Rypezn 9 3950X, Geforce GTX 1070ti):
I'm using a Vappoursynth script and just VSPipe, so no encoding is happening, just the decoding and filtering.
(source parameters, format: avc, color sampling: YUV420P8, bit depth: 8, resolution: 4096x2160, fps: 25, color matrix: 2020ncl, yuv luminance scale: limited, scanorder: progressive)

Just decoding:
----------------------
Using:
LWLibavSource + hardware decoding:
-> Output 6263 frames in 46.48 seconds (134.76 fps)

LWLibavSource + software decoding:
-> Output 6263 frames in 56.70 seconds (110.46 fps)

LibavSMASHSource:
-> Output 6263 frames in 55.06 seconds (113.76 fps)

FFMS2Source:
-> Output 6262 frames in 42.96 seconds (145.75 fps)

DGDecNV:
-> Output 6263 frames in 38.71 seconds (161.78 fps)

Note that:
a. different source might show different speeds
b. decoding speeds can wary from run to run about 10%
c. depending on the system software decoding speeds can vary a lot
d. this is all without encoding
e. cpu usage can vary a lot

Just decoding + SVP:
----------------------
DGDecNV + SVP: (software, preset: medium, GPU: off)
-> Output 15029 frames in 656.85 seconds (22.88 fps) (~64-85% cpu usage of VSPipe)

DGDecNV + SVP: (software, preset: medium, GPU: on)
-> Output 15029 frames in 1129.21 seconds (13.31 fps) (~51-80% cpu usage of VSPipe)

FFMS2Source + SVP: (software, preset: medium, GPU: on)
=> Output 15027 frames in 201.62 seconds (74.53 fps) (~30-40% cpu usage of VSPipe)

So software decoding and using SVP through the gpu gives the best speed on my system, but the cpu usage isn't that high.

Note:
1. When encoding is added things are changed up!
2. high cpu usage does not necessarily mean high speed
3. different GPU and CPU the numbers can be totally different

=> There is no: "Use this and your cpu usage and speed will be top"
Only thing you can do is check what works best on your system, by identifing bottlenecks.

Cu Selur

Hello Selur

SOLUTION: I reinstalled from Windows 10 to Windows 11 and it seemed to have solved the problem but thank you anyway for the help and time you spent explaining things.

Thank you

Regards

So it wasn't Hybrid that was causing the issue, but something that was messed up with your system settings.
-> Happy you solved the issue.

Cu Selur

Hello again,

It seems however that CPU is only used 100 percent when converting 1080p videos but when it comes to converting to 4K@60 fps it is only used at around 30 percent and is underutilized - conversion speed is only 2 FPS per second  . I am not sure what is problem. also when converting 4K the whole PC becomes very laggy.

Guess something wrong with you computer.

I just checked h264 1920x1080 24 video converted to 48 fps using Interframe/SVP to FFV1 codec MKV.
CPU 6 core/12 threads i7-8700K overclocked to 4.6ghz
GPU Radeon RX 580


GPU disabled = 88fps
GPU enabled = 79fps

Have no idea why but speed is slower when GPU enabled. Must be something not too optimized in this plugin for GPU. Enabled GPU in other plugins usually makes processing 3-5 times faster.
CPU usage in both situations is about 50% Seems plugin use all cores but can't use all threads.

@4K@60FPS:

Quote:also when converting 4K the whole PC becomes very laggy.

Hmm,... system should not get laggy, since Hybrid usualy calls everything with 'idle' priority.
you might want to check:
a. to input, temp, output folders lie on different drives?
b. is there something else using the disk?

About cpu usage with 1080p:
How is the cpu usage distributed between (vspipe,encoder) when you encode 1080p and 4k?

Assuming SVP is the bottleneck, you you try tweaking the SVP parameters, read: https://www.svp-team.com/wiki/Manual:SVPflow

Cu Selur