BD Rebuilder Beta - Bug Reports Only

[jdobbs]

I have updated the first post of this thread with a link to the latest version of BD-RB (v0.62.11). Changes for this release:

- Corrected an issue in which encoding can
  fail while using the QSVENC (Intel) GPU 
  encoder if encoding 720x480 interlaced 
  sources and deinterlacing is enabled.
- Fixed an error in which VC-1 sources were
  failing to encode while using the QSVENC
  (Intel) GPU encoder.
- Other minor corrections and cosmetic fixes.

[Blurayhd]

Originally Posted by jdobbs

I have updated the first post of this thread with a link to the latest version of BD-RB (v0.62.11). Changes for this release:

- Corrected an issue in which encoding can
  fail while using the QSVENC (Intel) GPU 
  encoder if encoding 720x480 interlaced 
  sources and deinterlacing is enabled.
- Fixed an error in which VC-1 sources were
  failing to encode while using the QSVENC
  (Intel) GPU encoder.
- Other minor corrections and cosmetic fixes.

Executable cant modify, cant run

[jdobbs]

Originally Posted by Blurayhd

Executable cant modify, cant run

Huh? It works here.

Anybody else having issues?

[zinoswink]

Executes just fine, no issue.

[dietachi]

Originally Posted by jdobbs

Huh? It works here.

Anybody else having issues?

There are no issues here.

VC1 and 480i work fine in the latest version as well for QSVEnc.

[jdobbs]

Originally Posted by zinoswink

Executes just fine, no issue.

Originally Posted by dietachi

There are no issues here.

VC1 and 480i work fine in the latest version as well for QSVEnc.

Thanks.

[Intel Iris the Great]

Works very well, thanks for the quick fix.

[jdobbs]

Originally Posted by Intel Iris the Great

Works very well, thanks for the quick fix.

Cool

[dietachi]

@jdobbs

VC1 and 480i have problems with VCEEnc.

For VC1:

"G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avhw  --fps 23.976 -i "K:\BTTF_BD\BDMV\STREAM\00300.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 9624 --sar 1:1 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 24 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00300.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00300.AVS.264
--------------------------------------------------------------------------------
avvce: codec vc1(yuv420p) unable to decode by vce.
failed to initialize file reader(s).
Finished with error in VCEEncC.

I switched to "--avsw" and it works

"G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avsw  --fps 23.976 -i "K:\BTTF_BD\BDMV\STREAM\00300.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 9624 --sar 1:1 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 24 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00300.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00300.AVS.264
--------------------------------------------------------------------------------
storage->SetProperty(BPicturesDeltaQP)=6 failed: invalid param..
storage->SetProperty(ReferenceBPicturesDeltaQP)=4 failed: invalid param..
VCEEnc (x86) 8.23 (r1710) by rigaya, Jul 18 2024 12:57:30 (VC 1940/Win)
OS:            Windows 11 x64 (22631) [UTF-8]
CPU:           AMD Ryzen 9 5950X 16-Core Processor [4.99GHz] (16C/32T)
GPU:           AMD Radeon RX 7600 [32.0.11029.1008]
AMF:           Runtime 1.4.34 / SDK 1.4.34
Input Info:    avsw: vc1(yv12)->nv12 [AVX2], 1920x1080, 24000/1001 fps

For the 480i the option "--vpp-deinterlace" does not exist in VCEEnc, only the following:

--vpp-afs
--vpp-nnedi
--vpp-yadif
--vpp-decomb

I tested it with "--vpp-decomb" and it works

G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avhw  --fps 29.97 --interlace tff --vpp-decomb -i "K:\BTTF_BD\BDMV\STREAM\00314.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 2021 --sar 40:33 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 30 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264
--------------------------------------------------------------------------------
storage->SetProperty(BPicturesDeltaQP)=6 failed: invalid param..
storage->SetProperty(ReferenceBPicturesDeltaQP)=4 failed: invalid param..
VCEEnc (x86) 8.23 (r1710) by rigaya, Jul 18 2024 12:57:30 (VC 1940/Win)
OS:            Windows 11 x64 (22631) [UTF-8]
CPU:           AMD Ryzen 9 5950X 16-Core Processor [4.96GHz] (16C/32T)
GPU:           AMD Radeon RX 7600 [32.0.11029.1008]
AMF:           Runtime 1.4.34 / SDK 1.4.34
Input Info:    avvce: H.264/AVC, 720x480, 30000/1001 fps
VPP            cspconv(nv12 -> yv12)
               decomb: full on, threshold 20, dthreshold 7, blend off
               cspconv(yv12 -> nv12)

Works with "--vpp-nnedi"

"G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avhw  --fps 29.97 --interlace tff --vpp-nnedi -i "K:\BTTF_BD\BDMV\STREAM\00314.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 2021 --sar 40:33 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 30 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264
--------------------------------------------------------------------------------
storage->SetProperty(BPicturesDeltaQP)=6 failed: invalid param..
storage->SetProperty(ReferenceBPicturesDeltaQP)=4 failed: invalid param..
VCEEnc (x86) 8.23 (r1710) by rigaya, Jul 18 2024 12:57:30 (VC 1940/Win)
OS:            Windows 11 x64 (22631) [UTF-8]
CPU:           AMD Ryzen 9 5950X 16-Core Processor [4.92GHz] (16C/32T)
GPU:           AMD Radeon RX 7600 [32.0.11029.1008]
AMF:           Runtime 1.4.34 / SDK 1.4.34
Input Info:    avvce: H.264/AVC, 720x480, 30000/1001 fps
VPP            cspconv(nv12 -> yv12)
               nnedi: field auto, nns 32, nsize 32x4, quality fast, prec fp32
                       pre_screen new_block, errortype abs, weight "internal"

It works for "--vpp-yadif mode=auto"

"G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avhw  --fps 29.97 --interlace tff --vpp-yadif mode=auto -i "K:\BTTF_BD\BDMV\STREAM\00314.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 2021 --sar 40:33 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 30 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264
--------------------------------------------------------------------------------
storage->SetProperty(BPicturesDeltaQP)=6 failed: invalid param..
storage->SetProperty(ReferenceBPicturesDeltaQP)=4 failed: invalid param..
VCEEnc (x86) 8.23 (r1710) by rigaya, Jul 18 2024 12:57:30 (VC 1940/Win)
OS:            Windows 11 x64 (22631) [UTF-8]
CPU:           AMD Ryzen 9 5950X 16-Core Processor [4.74GHz] (16C/32T)
GPU:           AMD Radeon RX 7600 [32.0.11029.1008]
AMF:           Runtime 1.4.34 / SDK 1.4.34
Input Info:    avvce: H.264/AVC, 720x480, 30000/1001 fps
VPP            cspconv(nv12 -> yv12)
               yadif: mode auto

It works with "--vpp-afs preset=default"

"G:\BD_Rebuilder\tools\vceenc\vceencc.exe" --avhw  --fps 29.97 --interlace tff --vpp-afs preset=default -i "K:\BTTF_BD\BDMV\STREAM\00314.m2ts" --codec h264 --preset fast --qp-min 0 --vbr 2021 --sar 40:33 --vbv-bufsize 30000 --max-bitrate 35000 --gop-len 30 --bframes 3 -o "G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264"
--------------------------------------------------------------------------------
G:\BD_REBUILDER\WORK\WORKFILES\VID_00314.AVS.264
--------------------------------------------------------------------------------
storage->SetProperty(BPicturesDeltaQP)=6 failed: invalid param..
storage->SetProperty(ReferenceBPicturesDeltaQP)=4 failed: invalid param..
VCEEnc (x86) 8.23 (r1710) by rigaya, Jul 18 2024 12:57:30 (VC 1940/Win)
OS:            Windows 11 x64 (22631) [UTF-8]
CPU:           AMD Ryzen 9 5950X 16-Core Processor [4.83GHz] (16C/32T)
GPU:           AMD Radeon RX 7600 [32.0.11029.1008]
AMF:           Runtime 1.4.34 / SDK 1.4.34
Input Info:    avvce: H.264/AVC, 720x480, 30000/1001 fps
VPP            cspconv(nv12 -> yv12)
               afs: clip(T 16, B 16, L 32, R 32), switch 0, coeff_shift 192
                    thre(shift 128, deint 48, Ymotion 112, Cmotion 224)
                    level 3, shift on, drop off, smooth off, force24 off
                    tune off, tb_order 1(tff), rff on, timecode off, log off
               cspconv(yv12 -> nv12)

For "--vpp-afs preset=XYZ" I did not check all the modes.

My apologies. I didn't had enough time to test this at the same time as the QSVenc.
Let me know if you need more info.

[jdobbs]

@dietachi

I'll work on it. As you know, I don't have a VCEEnc GPU for testing. Of the options available to replace "--vpp-deinterlace", which do you think generates the best looking output?

I'm inclined to go with "-vpp-decomb" since the others appear to throw away one field and regenerate it from the kept field. But I have seen instances where that actually looks better...

[dietachi]

Originally Posted by jdobbs

@dietachi

I'll work on it. As you know, I don't have a VCEEnc GPU for testing. Of the options available to replace "--vpp-deinterlace", which do you think generates the best looking output?

I'm inclined to go with "-vpp-decomb" since the others appear to throw away one field and regenerate it from the kept field. But I have seen instances where that actually looks better...

I will try to run some tests, but I can only do so after this Thursday.

When I had a look at the output videos after the encodes.
For this video "--vpp-decomb" was good.

I noticed in the past that "--vpp-afs" can cause some problems depending on the source file.

[jdobbs]

Originally Posted by dietachi

I will try to run some tests, but I can only do so after this Thursday.

When I had a look at the output videos after the encodes.
For this video "--vpp-decomb" was good.

I noticed in the past that "--vpp-afs" can cause some problems depending on the source file.

I'll just go ahead and replace it with "--vpp-decomb". If it appears to have issues later, we can always address it again.

[jdobbs]

I have updated the first post of this thread with a link to the latest version of BD-RB (v0.62.12). Changes for this release:

- Corrected an issue in which encoding can
  fail while using the VCEENCC (AMD) GPU 
  encoder if encoding 720x480 interlaced 
  sources and deinterlacing is enabled.
- Fixed an error in which VC-1 sources were
  failing to encode while using the VCEENCC
  (AMD) GPU encoder.
- Other minor corrections and cosmetic fixes.

[jdobbs]

Originally Posted by dietachi

I will try to run some tests, but I can only do so after this Thursday.

When I had a look at the output videos after the encodes.
For this video "--vpp-decomb" was good.

I noticed in the past that "--vpp-afs" can cause some problems depending on the source file.

I just thought I'd share something I just discovered. If you are encoding a source that is VC-1 using QSVEnc (Intel GPU), the speed of the encode is slower than what you might typically see because of the change to to "--avsw".

I found that forcing the encoder to use AVISYNTH and "--avs" as a source is actually faster -- almost twice as fast on my system.

There is a hidden setting NVENC_FORCE_AVS=1 that can be used. While it appears (by the name) to be applied only to NVEnc, it just so happens that it is applied across all three GPU encoders. So if you have a heavily VC-1 source, and are in a hurry, you might save yourself some encoding time by adding it to BDREBUILDER.INI's "[Options]" area just before the encode.

I'm not sure if the speed increase is also the case when using VCEEnc, as I'm not able to test it. It would be interesting to find out.

Of course you'd also have to remember to remove or disable that hidden setting when you are doing non-VC-1 sources -- as "--avhw" is likely significantly faster than using AVS mode.

[dietachi]

Originally Posted by jdobbs

I just thought I'd share something I just discovered. If you are encoding a source that is VC-1 using QSVEnc (Intel GPU), the speed of the encode is slower than what you might typically see because of the change to to "--avsw".

I found that forcing the encoder to use AVISYNTH and "--avs" as a source is actually faster -- almost twice as fast on my system.

There is a hidden setting NVENC_FORCE_AVS=1 that can be used. While it appears (by the name) to be applied only to NVEnc, it just so happens that it is applied across all three GPU encoders. So if you have a heavily VC-1 source, and are in a hurry, you might save yourself some encoding time by adding it to BDREBUILDER.INI's "[Options]" area just before the encode.

I'm not sure if the speed increase is also the case when using VCEEnc, as I'm not able to test it. It would be interesting to find out.

Of course you'd also have to remember to remove or disable that hidden setting when you are doing non-VC-1 sources -- as "--avhw" is likely significantly faster than using AVS mode.

Thank you for the information.
I will test this with VCEEnc and QSVEnc where both GPUs are in the same machine.
The disc I have to test is all mixed codec and interlaced and progressive for extras etc, while the main title is AVC.
Once done I'll report back.

[Intel Iris the Great]

Originally Posted by jdobbs

I just thought I'd share something I just discovered. If you are encoding a source that is VC-1 using QSVEnc (Intel GPU), the speed of the encode is slower than what you might typically see because of the change to to "--avsw".

I found that forcing the encoder to use AVISYNTH and "--avs" as a source is actually faster -- almost twice as fast on my system.

There is a hidden setting NVENC_FORCE_AVS=1 that can be used. While it appears (by the name) to be applied only to NVEnc, it just so happens that it is applied across all three GPU encoders. So if you have a heavily VC-1 source, and are in a hurry, you might save yourself some encoding time by adding it to BDREBUILDER.INI's "[Options]" area just before the encode.

I'm not sure if the speed increase is also the case when using VCEEnc, as I'm not able to test it. It would be interesting to find out.

Of course you'd also have to remember to remove or disable that hidden setting when you are doing non-VC-1 sources -- as "--avhw" is likely significantly faster than using AVS mode.

Yeah I noticed too that the VC-1 encode was much slower than an MPEG-4 encode using QSVEnc. Well it's already done now, but for now I won't be using VC-1 anytime soon. Thanks for the clarification though.

As for speed on my system, the QSVEnc GPU encoder runs over 3x faster than DirectShow CPU encoder for MPEG-4 encodes, so that's quite a significant increase !!

[dietachi]

Originally Posted by jdobbs

I just thought I'd share something I just discovered. If you are encoding a source that is VC-1 using QSVEnc (Intel GPU), the speed of the encode is slower than what you might typically see because of the change to to "--avsw".

I found that forcing the encoder to use AVISYNTH and "--avs" as a source is actually faster -- almost twice as fast on my system.

There is a hidden setting NVENC_FORCE_AVS=1 that can be used. While it appears (by the name) to be applied only to NVEnc, it just so happens that it is applied across all three GPU encoders. So if you have a heavily VC-1 source, and are in a hurry, you might save yourself some encoding time by adding it to BDREBUILDER.INI's "[Options]" area just before the encode.

I'm not sure if the speed increase is also the case when using VCEEnc, as I'm not able to test it. It would be interesting to find out.

Of course you'd also have to remember to remove or disable that hidden setting when you are doing non-VC-1 sources -- as "--avhw" is likely significantly faster than using AVS mode.

AVS vs. AVSW Encoding Performance Comparison

This comparison evaluates the encoding performance difference between using the --avs option and the --avsw option with two different GPU setups, using specific quality and preset settings for each encoder.

QSVEnc Settings: --quality fastest
VCEEnc Settings: --preset fast

Test Setup 1: QSVEnc with Intel Arc A770 Graphics
CPU: AMD Ryzen 9 5950X 16-Core Processor [4.99GHz] (16C/32T)
GPU Info: Intel Arc A770 Graphics (512EU) 300-2400MHz (32.0.101.5768)

Video Source: VC-1 - 1920x1080

With --avs: Encoded 15,456 frames at 435.87 fps, 5616.79 kbps, 431.64 MB.

With --avsw: Encoded 15,456 frames at 145.93 fps, 5616.78 kbps, 431.64 MB.

Result: The --avs option was approximately 3 times faster than --avsw .

Test Setup 2: VCEEnc with AMD Radeon RX 7600
CPU: AMD Ryzen 9 5950X 16-Core Processor [4.99GHz] (16C/32T)
GPU: AMD Radeon RX 7600 [32.0.11029.1008]

Video Source: VC-1 - 1920x1080

With --avs: Encoded 15,456 frames at 304.35 fps, 9550.10 kbps, 733.90 MB.

With --avsw: Encoded 15,456 frames at 121.79 fps, 9550.10 kbps, 733.90 MB.

Result: The --avs option was approximately 2.5 times faster than --avsw.

In summary, for VC-1 1920x1080 video decoding, the --avs option is notably faster than --avsw, achieving up to 3 times the speed depending on the system configuration.

[dietachi]

Originally Posted by jdobbs

I have updated the first post of this thread with a link to the latest version of BD-RB (v0.62.12). Changes for this release:

- Corrected an issue in which encoding can
  fail while using the VCEENCC (AMD) GPU 
  encoder if encoding 720x480 interlaced 
  sources and deinterlacing is enabled.
- Fixed an error in which VC-1 sources were
  failing to encode while using the VCEENCC
  (AMD) GPU encoder.
- Other minor corrections and cosmetic fixes.

I have tested both fixes and can confirm that they are working fine.

[dietachi]

AVHW vs. AVS Encoding Performance Comparison

This comparison evaluates the encoding performance of --avhw versus --avs for an AVC 1920x1080 video source across two different GPU configurations.

Test Setup 1: VCEEnc with AMD Radeon RX 7600
CPU: AMD Ryzen 9 5950X 16-Core Processor [4.92GHz] (16C/32T)
GPU: AMD Radeon RX 7600 [32.0.11029.1008]

With --avhw: Encoded 16,001 frames at 285.68 fps, 19526.57 kbps, 1553.48 MB.

With --avs: Encoded 16,001 frames at 190.26 fps, 19526.57 kbps, 1553.48 MB.

Result: The --avhw option was around 1.5 times faster than --avs on the AMD Radeon RX 7600.

Test Setup 2: QSVEnc with Intel Arc A770 Graphics
CPU: AMD Ryzen 9 5950X 16-Core Processor [4.92GHz] (16C/32T)
GPU: Intel Arc A770 Graphics (512EU) 300-2400MHz (32.0.101.5768)

With --avhw: Encoded 16,001 frames at 479.16 fps, 19960.55 kbps, 1588.01 MB.

With --avs: Encoded 16,001 frames at 304.42 fps, 19966.33 kbps, 1588.47 MB.

Result: The --avhw option was around 1.6 times faster than --avs on the Intel Arc A770 Graphics.

[jdobbs]

@dietachi

Thanks for the testing and feedback. It's good to know that my experience wasn't an anomaly.

I think I may go in and change the default encoding to "--avs" for QSVEnc and VCEEnc when VC-1 is the source. That way the user won't have to add and remove "NVENC_FORCE_AVS=1" from their BDREBUILDER.INI file to get the faster speeds.