Intel Xeon Gold 6230 - slow x265 encoding

[Efre]

Hello,
I am looking for someone who can help me figure out, why it is my encoding so slow. Just 2fps at max.

I have a DGINdexNV, indexed 1080p 2h long file. Tried "Simple x264/x265 Launcher 64bit" and encoding to x265 10bit 1080p with avisynth file as source where is basically nothing.
LoadPlugin(.... and
DGSource("C:\ENCODE\testfile.dgi", fieldop=0)

Output is x265 hevc (64bit), 10bit, crf20 slower, profile main 10
Custom Encode Parameters:
--sar 1:1 --colorprim bt709 --transfer bt709 --colormatrix bt709 --pme --pmode

or what other encoder you would recommend? Thank you.

[KarMa]

CPU usage and how many cores are being used?

[jagabo]

2 fps seems low for that CPU, but not horribly so. With ffmpeg at those settings I'm getting close to 4 fps (100 percent CPU usage) on an i9 9900k (4.68 GHz while encoding).

ffmpeg -y -hwaccel dxva2 -i %1 -pix_fmt yuv420p10le ^
    -c:v libx265 -preset slower -profile main10 ^
    -x265-params "crf=20:colorprim=bt709:transfer=bt709:colormatrix=bt709:pme=1:-pmode=1" ^
    -c:a copy "%~n1.x265.m10.mkv"

I have less that half as many cores but they are running more than twice as fast. Do you really need the Slower preset? Switching to Slow or Medium will be much faster. For example, the Medium preset gets about 35 fps on my computer.

[Zero-11]

I tested the Turing encoder (starts with the 1650 super) with the Matrix Bluray and it had 80 fps (h.265 slow setting in Handbrake) which is equivalent to x265 medium except for the background which is still more pixelated
or try these scripts: https://www.dropbox.com/sh/rebtkv4wtd9i7ra/AAD9FM4IVM5-_T5TgWBRhnDia?dl=0
rename .txt to .bat and execute after you put the file and the ffmpeg exe in the same folder
bitrate or crf can be edited in the .txt pretty easily
the x265 version is slower of course

[babygdav]

The cpu has no built in Quicksync support for hardware accelerated encoding and you're encoding to intensive 265 10-bit format, so it'll suck (vs accelerated Quicksync /nvenc encodes).

That said: https://software.intel.com/en-us/articles/video-transcoding-on-intel-xeon-scalable-pro...or-with-ffmpeg

Try a test using the same software, setup. Should be able to get around real time encodes with 20 cores

[sophisticles]

I'm guessing the x265 binary you are using wasn't built with SIMD optimizations enabled.

https://handbrake.fr/docs/en/1.3.0/technical/performance.html

Your cpu is a 20C/40T Cascade Lake with dual FMA AVX512 running at a base clock of 2.1 GHZ and max turbo of 3.9 GHZ, the cpu in Hanbrake's test is a 22C/44T Broadwell with a base clock of 2.2 GHZ and max turbo of 3.6 GHZ with no AVX512.

In one of the Handbrake tests they took a 4k source, cropped and resized it to 1080p, using x265 CRF 24 slower preset they saw 3.2 fps, your cpu has the dual FMA AVX-512 unit, a faster max turbo, more memory channels (6 vs 4) and a newer overall architecture but their cpu has significantly more cache (55MB vs 27.5 MB) 2 extra cores and a slightly higher base clock. They were also creating a bottleneck by cropping and resizing a 4k source to 1080p

Based on this I would estimate that you should be seeing about 10fps.

If the binary you are using was built with SIMD, look for decoding or I/O bottlenecks.