x265 low performance and suggested hardware

[ultradj83]

Description of problem or question:
Dears,
I'm facing low performance with h265 and I'm trying to understand if this is normal.
I'm also trying to understand the best hardware configuration.
I'm using the VerySlow encoder preset to encode 4K video RF 22 with Decomb, cause I've always been using this for h264 with good performances.
Speed is:

• 0.1fps on Core i7-1185G7 @ 3.00GHz, 1805 Mhz, 4 Core(s), 8 Logical Processor(s), 16GB RAM, enabled Intel QuickSync encoders support, Windows 11
  0.1fps on Nvidia RTX Virtual Workstation (AWS EC2), 4vCPUs, 16GB RAM, Nvidia RTX cloud GPU, enabled Nvidia NVEnc support, Windows Server 2022

I will soon try a AWS EC2 VT1 (Accelerated Video Transcoding Instance 1) with 24vCPUs, 48GB RAM.

Clear I'm using a "intensive" setting, but 0.1fps, overall on Nvidia RTX, looks abnormal.
We'll see the VT1.

Your thoughts?
Any suggesions?

Whould it be preferable to use Windows, Linux or Mac?

Regards


HandBrake version (e.g., 1.0.0):
1.6.1 (2023012300)



Operating system and version (e.g., Ubuntu 16.04 LTS, macOS 10.13 High Sierra, Windows 10 Creators Update):
Windows 11



HandBrake Activity Log ***required*** (see How-to get an activity log)

Code: Select all

HandBrake 1.6.1 (2023012300)
OS: Microsoft Windows NT 10.0.22621.0
CPU: 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz
Ram: 16106 MB, 
GPU Information:
  Intel(R) Iris(R) Xe Graphics - 31.0.101.3616
Screen: 1536x864
Temp Dir: C:\Users\feder\AppData\Local\Temp\
Install Dir: C:\Program Files\HandBrake\
Data Dir: C:\Users\feder\AppData\Roaming\HandBrake

-------------------------------------------

 # Starting Encode ...

[16:56:47] base preset: Fast 1080p30 (Modified)
[16:56:47] Remote Process started with Process ID: 29808 using port: 8037. Max Allowed Instances: 1
[16:56:47] Worker: Starting HandBrake Engine ...
[16:56:47] Worker: Starting Web Server on port 8037 ...
[16:56:48] Worker: Disconnected worker monitoring enabled!
[16:56:48] Compile-time hardening features are enabled
[16:56:49] qsv: is available on this system
[16:56:49] vcn: not available on this system
Cannot load nvEncodeAPI64.dll
[16:56:49] hb_init: starting libhb thread
[16:56:49] Starting work at: Wed Apr 19 16:56:49 2023
[16:56:49] 1 job(s) to process
[16:56:49] json job:
{
  "Audio": {
    "AudioList": [
      {
        "DRC": 0,
        "Encoder": "copy:opus",
        "Gain": 0,
        "Mixdown": -1,
        "NormalizeMixLevel": false,
        "Samplerate": 0,
        "Track": 0,
        "DitherMethod": 0
      }
    ],
    "CopyMask": [
      "copy:aac"
    ]
  },
  "Destination": {
    "ChapterList": [
      {
        "Name": "Chapter 1"
      }
    ],
    "ChapterMarkers": true,
    "AlignAVStart": false,
    "File": "\\\\ULTRANAS\\Download\\[...].mkv",
    "Mp4Options": {
      "IpodAtom": false,
      "Mp4Optimize": false
    },
    "Mux": "av_mkv"
  },
  "Filters": {
    "FilterList": [
      {
        "ID": 4,
        "Settings": {
          "mode": "7"
        }
      },
      {
        "ID": 3,
        "Settings": {
          "block-height": "16",
          "block-thresh": "40",
          "block-width": "16",
          "filter-mode": "2",
          "mode": "3",
          "motion-thresh": "1",
          "spatial-metric": "2",
          "spatial-thresh": "1"
        }
      },
      {
        "ID": 14,
        "Settings": {
          "crop-bottom": "232",
          "crop-left": "416",
          "crop-right": "414",
          "crop-top": "234",
          "height": "2160",
          "width": "3840"
        }
      },
      {
        "ID": 7,
        "Settings": {
          "mode": "0"
        }
      }
    ]
  },
  "PAR": {
    "Num": 1,
    "Den": 1
  },
  "Metadata": {
  [...]
  },
  "SequenceID": 0,
  "Source": {
    "Angle": 1,
    "Range": {
      "Type": "chapter",
      "Start": 1,
      "End": 1
    },
    "Title": 1,
    "Path": "\\\\ULTRANAS\\Download\\[...].webm"
  },
  "Subtitle": {
    "Search": {
      "Burn": false,
      "Default": false,
      "Enable": false,
      "Forced": false
    },
    "SubtitleList": []
  },
  "Video": {
    "Encoder": "x265",
    "Level": "auto",
    "TwoPass": false,
    "Turbo": false,
    "ColorMatrixCode": 0,
    "Options": "",
    "Preset": "veryslow",
    "Profile": "auto",
    "Quality": 22,
    "QSV": {
      "Decode": false
    },
    "HardwareDecode": 0
  }
}
[16:56:49] CPU: 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz
[16:56:49]  - Intel microarchitecture Tiger Lake
[16:56:49]  - logical processor count: 8
[16:56:49] Intel Quick Sync Video support: yes, gpu list: 0
[16:56:49] Intel Quick Sync Video integrated adapter with index 0
[16:56:49] Impl mfx-gen library path: C:\WINDOWS\System32\DriverStore\FileRepository\iigd_dch.inf_amd64_f4e2087fb68fa9c1\libmfx64-gen.dll
[16:56:49]  - Intel Media SDK hardware: API 2.7 (minimum: 1.3)
[16:56:49]  - Decode support:  h264 hevc (8bit: yes, 10bit: yes) av1 (8bit: yes, 10bit: yes)
[16:56:49]  - H.264 encoder: yes
[16:56:49]     - preferred implementation: hardware (1) via D3D11
[16:56:49]     - capabilities (hardware):  lowpower breftype icq vsinfo chromalocinfo opt1 opt2+mbbrc+extbrc+trellis+repeatpps+ib_adapt+nmpslice
[16:56:49]  - H.265 encoder: yes (8bit: yes, 10bit: yes)
[16:56:49]     - preferred implementation: hardware (1) via D3D11
[16:56:49]     - capabilities (hardware):  lowpower bpyramid icq vsinfo masteringinfo cllinfo opt1
[16:56:49]  - AV1 encoder: no
[16:56:49] hb_scan: path=\\ULTRANAS\Download\[...].webm, title_index=1
udfread ERROR: ECMA 167 Volume Recognition failed
src/libbluray/disc/disc.c:333: failed opening UDF image \\ULTRANAS\Download\[...].webm
src/libbluray/disc/disc.c:437: error opening file BDMV\index.bdmv
src/libbluray/disc/disc.c:437: error opening file BDMV\BACKUP\index.bdmv
[16:56:50] bd: not a bd - trying as a stream/file instead
libdvdread: Encrypted DVD support unavailable.
libdvdread: libdvdread: DVDOpenFileUDF:UDFFindFile /VIDEO_TS/VIDEO_TS.IFO failed
Can't open file VIDEO_TS.IFO.
libdvdnav: vm: vm: failed to read VIDEO_TS.IFO
[16:56:50] dvd: not a dvd - trying as a stream/file instead
Input #0, matroska,webm, from '\\ULTRANAS\Download\[...].webm':
  Metadata:
  [...]
[16:56:50] scan: decoding previews for title 1
[16:56:50] scan: audio 0x1: opus, rate=48000Hz, bitrate=1 English (Opus) (2.0 ch)
[16:56:51] scan: 10 previews, 3840x2160, 25.000 fps, autocrop = 0/0/416/414, aspect 16:9, PAR 1:1, color profile: 1-1-1, chroma location: unspecified
[16:56:51] libhb: scan thread found 1 valid title(s)
[16:56:51] qsv: integrated qsv adapter with index 0 has been selected
[16:56:51] Starting Task: Encoding Pass
[16:56:51] Skipping vfr filter
[16:56:51] work: only 1 chapter, disabling chapter markers
[16:56:51] job configuration:
[16:56:51]  * source
[16:56:51]    + \\ULTRANAS\Download\[...].webm
[16:56:51]    + title 1, chapter(s) 1 to 1
[16:56:51]    + container: matroska,webm
[16:56:51]    + data rate: 10642 kbps
[16:56:51]  * destination
[16:56:51]    + \\ULTRANAS\Download\[...].mkv
[16:56:51]    + container: Matroska (libavformat)
[16:56:51]  * video track
[16:56:51]    + decoder: vp9 8-bit (yuv420p)
[16:56:51]    + filters
[16:56:51]      + Comb Detect (mode=3:spatial-metric=2:motion-thresh=1:spatial-thresh=1:filter-mode=2:block-thresh=40:block-width=16:block-height=16)
[16:56:51]      + Decomb (mode=39)
[16:56:51]      + Crop and Scale (width=3840:height=2160:crop-top=234:crop-bottom=232:crop-left=416:crop-right=414)
[16:56:51]        + source: 3840 * 2160, crop (234/232/416/414): 3010 * 1694, scale: 3840 * 2160
[16:56:51]    + Output geometry
[16:56:51]      + storage dimensions: 3840 x 2160
[16:56:51]      + pixel aspect ratio: 1 : 1
[16:56:51]      + display dimensions: 3840 x 2160
[16:56:51]    + encoder: H.265 (libx265)
[16:56:51]      + preset:  veryslow
[16:56:51]      + profile: auto
[16:56:51]      + level:   auto
[16:56:51]      + quality: 22.00 (RF)
[16:56:51]      + color profile: 1-1-1
[16:56:51]      + chroma location: unspecified
[16:56:51]  * audio track 1
[16:56:51]    + decoder: English (Opus) (2.0 ch) (track 1, id 0x1)
[16:56:51]      + samplerate: 48000 Hz
[16:56:51]    + Opus Passthru
[16:56:51] sync: expecting 6167 video frames
x265 [info]: HEVC encoder version 3.5+1-f0c1022b6
x265 [info]: build info [Windows][GCC 10.2.0][64 bit] 8bit+10bit+12bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
x265 [info]: Main profile, Level-5 (Main tier)
x265 [info]: Thread pool created using 8 threads
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 3 / wpp(34 rows)
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: Residual QT: max TU size, max depth : 32 / 3 inter / 3 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 4 / 5
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00
x265 [info]: Lookahead / bframes / badapt        : 40 / 8 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 1
x265 [info]: References / ref-limit  cu / depth  : 5 / off / off
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-22.0 / 0.60
x265 [info]: tools: rect amp rd=6 psy-rd=2.00 rdoq=2 psy-rdoq=1.00 rskip mode=1
x265 [info]: tools: signhide tmvp b-intra strong-intra-smoothing deblock sao
[16:56:51] sync: first pts audio 0x1 is 0
[16:56:51] sync: first pts video is 675
[16:56:51] sync: "Chapter 1" (1) at frame 1 time 675

 # Encode Paused


 # Encode Resumed

[02:22:15] reader: done. 1 scr changes
[02:39:09] work: average encoding speed for job is 0.180499 fps
[02:39:09] comb detect: heavy 14 | light 86 | uncombed 6066 | total 6166
[02:39:09] decomb: deinterlaced 14 | blended 86 | unfiltered 6066 | total 6166
[02:39:09] opus-decoder done: 12334 frames, 0 decoder errors
[02:39:09] vp9-decoder done: 6166 frames, 0 decoder errors
[02:39:09] sync: got 6166 frames, 6167 expected
[02:39:09] sync: framerate min 25.000 fps, max 25.000 fps, avg 25.000 fps
x265 [info]: frame I:     41, Avg QP:21.03  kb/s: 18285.00
x265 [info]: frame P:   1645, Avg QP:22.59  kb/s: 12011.35
x265 [info]: frame B:   4480, Avg QP:26.84  kb/s: 3698.64
x265 [info]: Weighted P-Frames: Y:2.7% UV:1.9%
x265 [info]: Weighted B-Frames: Y:2.3% UV:1.2%
encoded 6166 frames in 34937.65s (0.18 fps), 6013.34 kb/s, Avg QP:25.66
[02:39:10] mux: track 0, 6166 frames, 185415876 bytes, 6012.19 kbps, fifo 512
[02:39:10] mux: track 1, 12334 frames, 3930273 bytes, 127.44 kbps, fifo 2048
[02:39:10] Finished work at: Thu Apr 20 02:39:10 2023
[02:39:10] libhb: work result = 0

 # Job Completed!

[rollin_eng]

I would start by turning off any filters you don’t need, decomb etc.

You could also try moving your source and destination files to local drives to rule out any network speed issues.

[ultradj83]

I would start by turning off any filters you don’t need, decomb etc.

You could also try moving your source and destination files to local drives to rule out any network speed issues.

Well, yes, I'm testing on local drives.
About decomb, shouldn't it perform deinterlacing only when interlaced video is detected?

Regards

[mduell]

I mean, you picked a very slow setting on a slow encoder with high resolution video... so it's going to be slow.

But I think the problem here may be that you paused the encode for 9+ hours, which may throw off the math.

[ultradj83]

I mean, you picked a very slow setting on a slow encoder with high resolution video... so it's going to be slow.

But I think the problem here may be that you paused the encode for 9+ hours, which may throw off the math.

Will i know it's extreme, but I never paused the encode...

[rollin_eng]

[16:56:51] sync: first pts video is 675
[16:56:51] sync: "Chapter 1" (1) at frame 1 time 675

# Encode Paused


# Encode Resumed

[02:22:15] reader: done. 1 scr changes

Like I said, try it without filters and with local drives, don’t pause it and repost your log.

[ultradj83]

[16:56:51] sync: first pts video is 675
[16:56:51] sync: "Chapter 1" (1) at frame 1 time 675

# Encode Paused


# Encode Resumed

[02:22:15] reader: done. 1 scr changes

Like I said, try it without filters and with local drives, don’t pause it and repost your log.

I'm pretty sure I was sleeping at 2am

[rollin_eng]

Well maybe you have ghosts

[Deleted User 11865]

x264 veryslow falls somewhere between x265 medium and x265 slow, IIRC; so yes, x265 is significantly slower than x264 (and, more generally, HEVC encoders are slower than H.264 encoders). Some of the features that improve compression efficiency do result in additional time spent encoding the video.

[ultradj83]

x264 veryslow falls somewhere between x265 medium and x265 slow, IIRC; so yes, x265 is significantly slower than x264 (and, more generally, HEVC encoders are slower than H.264 encoders). Some of the features that improve compression efficiency do result in additional time spent encoding the video.

Thanks for comparison, I'll set medium when I'll have a full movie to encode, otherwise it will wake weeks...
I see h265 is heavier, of course, being more efficient.
I'm just surprised that the RTX was completely useless, it's reported as the fastest hardware encoder, but in the truth there's no improvement compared to a regular software encoder on Core i7. I'm still not sure this is normal...
I'm still waiting for the AWS support to allow me 24vCPUs for my VT instance, that would be close to the best hardware I can have, let's see what happens...

Regards

[s55]

I'm just surprised that the RTX was completely useless, it's reported as the fastest hardware encoder, but in the truth there's no improvement compared to a regular software encoder on Core i7. I'm still not sure this is normal...

The log above showed you used x265(software encoder), not nvenc.

NVEnc can be bottlenecked pretty easily if you don't have the entire pipeline on the GPU which won't be the case when you have software filters running.

Also, try avoid using UNC shares. " \\ULTRANAS\"

[ultradj83]

I'm just surprised that the RTX was completely useless, it's reported as the fastest hardware encoder, but in the truth there's no improvement compared to a regular software encoder on Core i7. I'm still not sure this is normal...

The log above showed you used x265(software encoder), not nvenc.

NVEnc can be bottlenecked pretty easily if you don't have the entire pipeline on the GPU which won't be the case when you have software filters running.

Also, try avoid using UNC shares. " \\ULTRANAS\"

Yes, that log comes from my i7 laptop.
When I tested RTX, everything was local.
That was a Nvidia RTX Virtual Workstation (AWS EC2), 4vCPUs, 16GB RAM, Nvidia RTX cloud GPU, enabled Nvidia NVEnc support, Windows Server 2022.

Regards

[s55]

" enabled Nvidia NVEnc support" > what do you mean? Enabling in preferences is not sufficient for it to be used. Did you actually choose the NVEnc encoder?

[ultradj83]

" enabled Nvidia NVEnc support" > what do you mean? Enabling in preferences is not sufficient for it to be used. Did you actually choose the NVEnc encoder?

I'm not sure.
It was enabled in the settings.
Anything else to do?

Regards

[s55]

You need to choose the NVenc encoder on the video tab. x265 is a software encoder that implements the h265 standard. NVENC (H265) is a hardware encoder that implements the same standard. They are not the same. Using NVENC, your trading worse quality/files for, usually, faster encodes.

[ultradj83]

You need to choose the NVenc encoder on the video tab. x265 is a software encoder that implements the h265 standard. NVENC (H265) is a hardware encoder that implements the same standard. They are not the same. Using NVENC, your trading worse quality/files for, usually, faster encodes.

Ok, I lost a bit
I think the same is valid for Intel QSV...
I'll retest both, by the way if they return worse quality I think they are pretty useless...

Regards

[s55]

> I'll retest both, by the way if they return worse quality I think they are pretty useless...

It depends. For most, it's close enough that it doesn't matter, even if the file is a bit bigger to compensate but if you want the best quality, stick with software and just tune the settings for more tolerable speed.

[wyliec2]

Software encoding with H265 at Slow, Slower or Very Slow will be SLOW.

I use a 5950X (16 cores/32 threads) for Handbrake. I usually do 4K encodes at RF 19 and it uses > 90% CPU. Speed will vary with content, but I typically see < 2 fps on Slower and < 5 fps on Slow. I have never tried Very Slow. Most encodes at Slower run 24 - 40 hours.

This machine is largely a 'hobby' machine that runs in my basement dedicated to Handbrake encodes or other performance experiments - the long runs don't impact me other than the electricity cost..!!

[ultradj83]

Software encoding with H265 at Slow, Slower or Very Slow will be SLOW.

I use a 5950X (16 cores/32 threads) for Handbrake. I usually do 4K encodes at RF 19 and it uses > 90% CPU. Speed will vary with content, but I typically see < 2 fps on Slower and < 5 fps on Slow. I have never tried Very Slow. Most encodes at Slower run 24 - 40 hours.

This machine is largely a 'hobby' machine that runs in my basement dedicated to Handbrake encodes or other performance experiments - the long runs don't impact me other than the electricity cost..!!

Mmmh... I have something similar, but it's beginning old...
Could you detail your hardware and related cost?
I'm thinking if to buy a new server but I'm not persuaded I'll reach needed performance, that's why I'm exploring cloud computing...

Regards

[wyliec2]

My build is becoming dated as well - mostly 3 years old...X570 Asus Crosshair VIII Formula mobo, 32 GB 3600 CL16 RAM, 2 M.2 PCIE 4 SSDs built in 2020 with a 3950X CPU and 1050Ti GPU. In 2021 I upgraded the CPU to 5950X and swapped a 3080Ti for the GPU.

Today, you could likely build this for $1500 (including PSU, case, SSDs) not counting the GPU. Of course, with software encoding, the GPU provides nothing and any minimal solution will do.

With H265 at Slow (or slower) and RF in the 18-20 range, 4K encodes will run > 90% CPU utilization and BD encodes around 60%. For my hobby experimenting, I'm primarily looking for minimizing the encode time for a single file or, at most, two concurrent encodes for BD content.

I have looked at newer platforms - 7970X and 13900K. They seem to deliver around 30% better performance in Handbrake then the 5950X. The downside is that power and heat increases are substantially greater than the performance improvement. My present 5950X platform draws 300 watts continuous while running encodes - doing this 24/7 for a month does show up on my electricity bill - for my hobby purposes, I'll likely stay where I'm at.

Much of my Handbrake use is associated with my hobby to tinker with high-performance consumer machines - while I have a couple thousand movies that were encoded for playback, I wind up re-encoding them playing with system and Handbrake settings.

Handbrake is my ultimate stress test - I use Aida64, 3DMark, Cinebench R20 and R23 as tools, but ultimately a 24+ hour encode will truly test stability and reveal power use, heat, core voltage, etc.

FWIW - my system progression has been 4770K --> 5960X --> 3950X --> 5950X. Particularly for Handbrake, the 3950X/5950X need to have PBO (precision boost overdrive) turned off and an all-core clock speed applied - this is not complicated and it reduces temperatures and voltages and substantially improves performance on heavily multi-threaded workloads.

[ultradj83]

Well I did something similar years ago.
Now I use handbrake just spot, so makes less sense to build a strong machine.
I'm still waiting to try the AWS VT1, if they will ever get rid of the error preventing me to launch the instance, it's 4 days I'm waiting for it to be fixed...