Testing Nvidia Nvenc encoding
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below: [hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command prime-select on-demand and verified that both the Intel and Nvidia drivers were available inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends. ============== Test h264_nvenc.mp4 ------------------- OK ./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05 Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default) ** rendered 5972 frames in 27.391 secs, 218.028 fps ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default) ------- Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !? Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default) =================== Test h265_nvenc.mp4 ------------------- Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library [hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished. =================== GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new Tested hevc_nvenc full transcoding with FFmpeg 7.1 ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4 frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890 ------------------ ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
Cingg then should supports NVENC via some dynamic linked external Nvidia lib
Nvenc supported in ffmpeg, but you need to add nvidia headers and add a flag during ffmpeg configuration step. The friend of mine tested davinci and ffmpeg encoding on nvidia card on windows. The davinci is 7 times faster compared to ffmpeg. In both cases nvenc is used and card used by 100%. I don't know the reason of this. Best regards, Andrey пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video
acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
Den 20.12.2024 17:27, skrev Андрей Спицын via Cin:
Cingg then should supports NVENC via some dynamic linked external Nvidia lib
Nvenc supported in ffmpeg, but you need to add nvidia headers and add a flag during ffmpeg configuration step.
The friend of mine tested davinci and ffmpeg encoding on nvidia card on windows. The davinci is 7 times faster compared to ffmpeg. In both cases nvenc is used and card used by 100%. I don't know the reason of this.
Not sure why h264_nvenc works and not h265_nvenc in Cingg. Quality/speed may differ with different profiles/presets in use, though 7x was much .
пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video
acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
====== for hevc nvenc test may be try to set mjpeg colorspace instead of jpeg?
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
Den 20.12.2024 17:40, skrev Andrew Randrianasulu:
пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
======
for hevc nvenc test may be try to set mjpeg colorspace instead of jpeg?
Tried the other option MPEG, and also BT.709, which both works for h264_nvenc, but has no effect with regards to the hevc errors. Tried also to comment out each and both of the h265_nvenc preset's preset medium profile main but with no effect. So far it looks for me that there is something wrong with Cingg's hevc_nvenc encoding.
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
пт, 20 дек. 2024 г., 21:58 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 20.12.2024 17:40, skrev Andrew Randrianasulu:
пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video
acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
======
for hevc nvenc test may be try to set mjpeg colorspace instead of jpeg?
Tried the other option MPEG, and also BT.709, which both works for h264_nvenc, but has no effect with regards to the hevc errors.
Tried also to comment out each and both of the h265_nvenc preset's preset medium profile main
but with no effect.
So far it looks for me that there is something wrong with Cingg's hevc_nvenc encoding.
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime. you can try to install something like nv-codec-headers and then add --disable-ffnvcodec configure switch to cingg (disables our copy of nv headers from 2020) and see if ffmpeg compilation picked up system's headers. BUT you also can test progressive (not interlaced) file, like you did with qsv.
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
Den 21.12.2024 03:17, skrev Andrew Randrianasulu:
пт, 20 дек. 2024 г., 21:58 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 20.12.2024 17:40, skrev Andrew Randrianasulu:
пт, 20 дек. 2024 г., 19:20 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
Continued and extracted from [Cin] Video accelerators support, builds and use https://lists.cinelerra-gg.org/pipermail/cin/2024-December/009226.html
Den 12.12.2024 00:54, skrev Terje J. Hanssen:
Den 11.12.2024 23:56, skrev Andrew Randrianasulu:
чт, 12 дек. 2024 г., 00:43 Terje J. Hanssen via Cin <cin@lists.cinelerra-gg.org>:
To clarify some pieces once again, I put up some basic statements or questions:
..........snip
So what happened when adding oneVPL (qsv) support to the build system; dynamic linked to system or static added embedded into the build?
dynamic
.....snip
Could in principle similar methods be extended to include broader video acceleration support for AMD/amf and NVIDIA/nvenc?
nvenc already supported, I think? At some point I tried it with GF710 on livedvd and it was working for me. try to test it if you have proprietary nvidia drivers.
Apparently Cingg then should supports NVENC via some dynamic linked external Nvidia lib, equivalent to what we did for for QSV via oneVPL support. While h264_nvenc.mp4 rendering works, h265_nvenc.mp4 fails with the following errors, see detail below:
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library
Yes, seemingly: Cin/bin/ffmpeg/video> ls *nvenc* h264_nvenc.mp4 h264_nvenc.qt h265_nvenc.mp4
NVENC Presets
cat h264_nvenc.mp4 mp4 h264_nvenc # encode for nvidia graphics hw only preset medium profile main
cat h265_nvenc.mp4 mp4 hevc_nvenc # encode for nvidia graphics hw only preset medium profile main
Maybe I can do an attempt later, if I get life in the old GeForce GTX 960 in my Skylake workstation.
Setup ---------- Booted BIOS and checked that the integrated Intel gpu and multimonitor setup were enabled. Installed the suse-prime package for nvidia/intel gpu selection (for NVIDIA optimus laptops with bbswitch support) and could then run the command
prime-select on-demand
and verified that both the Intel and Nvidia drivers were available
inxi -G Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 Display: wayland server: X.org v: 1.21.1.14 with: Xwayland v: 24.1.4 compositor: gnome-shell v: 47.2 driver: X: loaded: modesetting,nvidia unloaded: fbdev,vesa dri: iris gpu: i915 resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: intel mesa v: 24.3.1 renderer: Mesa Intel HD Graphics 530 (SKL GT2) API: EGL Message: EGL data requires eglinfo. Check --recommends.
==============
Test h264_nvenc.mp4 ------------------- OK
./CinGG-20241120-x86_64.AppImage Cinelerra Infinity - built: Nov 20 2024 22:06:05
Loads hdv09_04.m27 Shift-R mp4 h264_nvenc (video preset) pixels: yuv420p (default)
** rendered 5972 frames in 27.391 secs, 218.028 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_yuv420p.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2156 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
-------
Tried also h264_nvenc with pixels: nv12 which for slowed down to 45% speed !?
Render::render_single: Session finished. ** rendered 5972 frames in 60.368 secs, 98.927 fps
ffprobe -hide_banner hdv09_04_h264_nvenc_nv12.mp4 Stream #0:0[0x1](und): Video: h264 (Main) (avc1 / 0x31637661), yuvj420p(pc, smpte170m/unknown/unknown), 1440x1080 [SAR 4:3 DAR 16:9], 2154 kb/s, 25 fps, 25 tbr, 12800 tbn (default)
======
for hevc nvenc test may be try to set mjpeg colorspace instead of jpeg?
Tried the other option MPEG, and also BT.709, which both works for h264_nvenc, but has no effect with regards to the hevc errors.
Tried also to comment out each and both of the h265_nvenc preset's preset medium profile main
but with no effect.
So far it looks for me that there is something wrong with Cingg's hevc_nvenc encoding.
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
you can try to install something like nv-codec-headers and then add
--disable-ffnvcodec
configure switch to cingg (disables our copy of nv headers from 2020) and see if ffmpeg compilation picked up system's headers.
BUT you also can test progressive (not interlaced) file, like you did with qsv.
Yeah, hevc_nvenc turns out to have just the same tff interlace issue as we experienced with qsv. At first I just tried to change Cingg Setting | Fomat Video to "Not interlaced" and the interlaced hdv file rendered OK. Tested also that a deinterlaced (progressive) version also rendered ok. Loaded also a bff interlaced SD.dv file that Cingg Setting recognized as "Unknown", did render ok without manual change. Maybe also h265_nvenc preset at opportunity also better should be renamed to hevc_nvenc in accordance with ffmpeg's use.
===================
Test h265_nvenc.mp4 -------------------
Did not render, neither with pixels yuv420p nor nv12: FFMPEG::open_encoder err: Generic error in an external library
[hevc_nvenc @ 0x7ff79c015640] No capable devices found FFMPEG::open_encoder err: Generic error in an external library int FFMPEG::open_encoder(const char*, const char*): open failed hevc_nvenc:/run/media/terje/video/hdv09_04_h265_nvenc_yuv420p.mp4 Render::render_single: Session finished.
===================
GeForce GTX 960 (GM206) should be capable to manage NVENC h265/hevc YUV 420 encoding https://en.wikipedia.org/wiki/Nvidia_NVENC#Third_generation,_Maxwell_GM20x https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new
Tested hevc_nvenc full transcoding with FFmpeg 7.1
ffmpeg -hide_banner -hwaccel cuda -hwaccel_output_format cuda -i hdv09_04.m2t -c:v hevc_nvenc hdv09_04_ff_hevc_nvenc.mp4
frame= 5963 fps=270 q=34.0 Lsize= 64316KiB time=00:03:58.77 bitrate=2206.6kbits/s speed=10.8x [aac @ 0x55a38d4c4d40] Qavg: 488.890
------------------
ffprobe -hide_banner hdv09_04_ff_hevc_nvenc.mp4 Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], 2074 kb/s, 24.99 fps, 25 tbr, 12800 tbn (default)
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
you can try to install something like nv-codec-headers and then add
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 22.12.2024 00:04, skrev Andrew Randrianasulu:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Andrew, As far as there might be a workaround also for the nvenc tff interlace issue, I didn't do more about the latter than searching the most similar package ffnvcodec-devel (FFmpeg version of NVIDIA codec API headers) Additional I think an AppImage built successful of my dynamic build ffmpeg-7.1 with sh ./bld_appimage.sh bin_use_system_ffmpeg-71 It works on the build-machine, and I will test it on the older machines too. Phyllis, I was about to send a little comment to your first News version, regarding relative "new" Intel hardware. The SkyLake/ KabyLake test machines are from 2015/ 2016 respectively :) Of course they have lesser codecs support than the relative new bult-machine.
Den 22.12.2024 01:07, skrev Terje J. Hanssen:
Den 22.12.2024 00:04, skrev Andrew Randrianasulu:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
ffmpeg 7.1 itself is apparently capable to do AV1 accelerated encodings: ffmpeg -hide_banner -encoders | egrep -i 'av1_nvenc|av1_qsv|av1_vaapi|av1_vulkan' V....D av1_nvenc NVIDIA NVENC av1 encoder (codec av1) V..... av1_qsv AV1 (Intel Quick Sync Video acceleration) (codec av1) V....D av1_vaapi AV1 (VAAPI) (codec av1) and awaiting for the upcoming av1_vulkan next ......... 😉
you can try to install something like nv-codec-headers and then add
Andrew, As far as there might be a workaround also for the nvenc tff interlace issue, I didn't do more about the latter than searching the most similar package ffnvcodec-devel (FFmpeg version of NVIDIA codec API headers)
Additional I think an AppImage built successful of my dynamic build ffmpeg-7.1 with sh ./bld_appimage.sh bin_use_system_ffmpeg-71 It works on the build-machine, and I will test it on the older machines too.
Tested this appimage with system FFmpeg 7.1 on SkyLake w/dual dGPU NV GTX 960 + iGPU Intel HD 530: h265_nvenc with workaround switching Cingg Format tff interlaced to Not interlaced ** rendered 5972 frames in 62.890 secs, 94.959 fps hevc_qsv rendered ok directly w/audio ** rendered 5972 frames in 151.656 secs, 39.379 fps ------------------------ Tested on KabyLake (Dell XPS-13/9370) with system FFmpeg 7.1 and iGPU Intel UHD 620: hevc_qsv rendered ok directly w/audio ** rendered 5972 frames in 91.944 secs, 64.953 fps audio0 pad 64 0 (64)
Phyllis, I was about to send a little comment to your first News version, regarding relative "new" Intel hardware. The SkyLake/ KabyLake test machines are from 2015/ 2016 respectively :) Of course they have lesser codecs support than the relative new bult-machine.
Terje, thanks for reviewing the News in the CinGG website. It is sort of vague but If anyone is interested in really getting the most benefit out of the absolutely newest hardware/software, they really need to build CinGG themselves and hopefully they will read the Archived mail on how you did that. I would have liked to document this in the manual but without having all of the latest hardware, I just would not get it correct. And anyway there is always new and better software/hardware coming out and changing things. Thanks for all that you do ! ...Phyllis Phyllis,
I was about to send a little comment to your first News version, regarding relative "new" Intel hardware. The SkyLake/ KabyLake test machines are from 2015/ 2016 respectively :) Of course they have lesser codecs support than the relative new bult-machine.
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later. On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers /FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/ As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll: Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work. On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Great so it will be included in the next release. Thank you very much. On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally). Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date? Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710. https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2*
Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits
Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
> most likely our nv headers drifted from that ffmpeg-7.0/proprietary > driver assumes at runtime. > I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
> you can try to install something like nv-codec-headers and then add > >
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw. Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;) https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/ yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2*
Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits
Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
> Andrew, > >> most likely our nv headers drifted from that ffmpeg-7.0/proprietary >> driver assumes at runtime. >> > I have been wondering about nv-codec-headers as we are at: > > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 > but I am unsure about updating to: > > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 > because if you look at: > https://github.com/FFmpeg/nv-codec-headers/releases/ > the release versions go from 12.xx to 8.x and it is really weird AND > there is no year on the release dates but just day and month. > Since it is such an important part of ffmpeg inside CinGG, I am > concerned but will at least try the 12.2.72.0 just to see what it does. >
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
> >> you can try to install something like nv-codec-headers and then add >> >>
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation) https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2* Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 27.12.2024 21:37, skrev Terje J. Hanssen:
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation) https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
And then, about 6 months ago, I had the following drivers for GT 730 enabled on Leap15.5 and Slowroll respectively: Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.256.02 Device-1: NVIDIA GK208B [GeForce GT 730] driver: nouveau v: kernel
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2* Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
пт, 27 дек. 2024 г., 23:56 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 27.12.2024 21:37, skrev Terje J. Hanssen:
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation)
https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
And then, about 6 months ago, I had the following drivers for GT 730 enabled on Leap15.5 and Slowroll respectively:
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.256.02
isn't 470 < 550 ?
But I do not think opening desktop case just for this test actually good idea ... do not worry, ff nvidia headers quite small tarball, I always can fish old one from one of my backups .... Device-1: NVIDIA GK208B [GeForce GT 730] driver: nouveau v: kernel
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2*
Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits
Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
> > > вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>: > >> Andrew, >> >>> most likely our nv headers drifted from that >>> ffmpeg-7.0/proprietary driver assumes at runtime. >>> >> I have been wondering about nv-codec-headers as we are at: >> >> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 >> but I am unsure about updating to: >> >> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 >> because if you look at: >> https://github.com/FFmpeg/nv-codec-headers/releases/ >> the release versions go from 12.xx to 8.x and it is really weird >> AND there is no year on the release dates but just day and month. >> Since it is such an important part of ffmpeg inside CinGG, I am >> concerned but will at least try the 12.2.72.0 just to see what it does. >> > > in theory it should give users of new nvidia hardware av1 encoding > ...... but not sure how it will work with older drivers and hardware. > > > > >> >>> you can try to install something like nv-codec-headers and then >>> add >>> >>>
Den 27.12.2024 22:25, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 23:56 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 27.12.2024 21:37, skrev Terje J. Hanssen:
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation) https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
And then, about 6 months ago, I had the following drivers for GT 730 enabled on Leap15.5 and Slowroll respectively:
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.256.02
isn't 470 < 550 ?
Clearly, yes. The opensuse SDB:NVIDIA lists the following driver series for different NV devices/families https://en.opensuse.org/SDB:NVIDIA_drivers#Determination_of_driver_version Note: NVIDIA driver versions are mapped to the following naming convention listed below. You will need this information when you are ready to install via commandline/zypper. * G03 = driver v340 = legacy driver (via a community user <https://software.opensuse.org/search?baseproject=ALL&q=g03>) * G04 = driver v390 = legacy driver (Fermi devices) * G05 = driver v470 = driver for Kepler devices * G06 = driver v550 = driver for Maxwell, Pascal devices and up Note: Some OEMs use different chip familys (Kepler, Fermi) for the same product, as example GeForce 730. *Warning:* Legacy drivers G03 and G04 won't work with the actual kernel 6.6.x series. Use free nouveau drivers instead!
But I do not think opening desktop case just for this test actually good idea ...
do not worry, ff nvidia headers quite small tarball, I always can fish old one from one of my backups ....
Yeah, I don't use the gt730 anymore, as gtx960 was some more powerful.
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nouveau v: kernel
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2* Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 28.12.2024 01:51, skrev Terje J. Hanssen:
Den 27.12.2024 22:25, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 23:56 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 27.12.2024 21:37, skrev Terje J. Hanssen:
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation) https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
And then, about 6 months ago, I had the following drivers for GT 730 enabled on Leap15.5 and Slowroll respectively:
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.256.02
isn't 470 < 550 ?
Clearly, yes.
Just to verify by searching for gt730 (or gt710), GF 700 series, GF730, Linux 64-bit : https://www.nvidia.com/en-us/drivers/ Driver Version:470.256.02 Release Date:Tue Jun 04, 2024 If Linux 32-bit ARM: Driver Version:390.157 Release Date:Tue Nov 22, 2022
The opensuse SDB:NVIDIA lists the following driver series for different NV devices/families https://en.opensuse.org/SDB:NVIDIA_drivers#Determination_of_driver_version
Note: NVIDIA driver versions are mapped to the following naming convention listed below. You will need this information when you are ready to install via commandline/zypper.
* G03 = driver v340 = legacy driver (via a community user <https://software.opensuse.org/search?baseproject=ALL&q=g03>) * G04 = driver v390 = legacy driver (Fermi devices) * G05 = driver v470 = driver for Kepler devices * G06 = driver v550 = driver for Maxwell, Pascal devices and up
Note: Some OEMs use different chip familys (Kepler, Fermi) for the same product, as example GeForce 730.
*Warning:* Legacy drivers G03 and G04 won't work with the actual kernel 6.6.x series. Use free nouveau drivers instead!
But I do not think opening desktop case just for this test actually good idea ...
do not worry, ff nvidia headers quite small tarball, I always can fish old one from one of my backups ....
Yeah, I don't use the gt730 anymore, as gtx960 was some more powerful.
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nouveau v: kernel
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2* Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith: > Downloaded latest version of > nv-codec-headers release and built CinGG > with it and no problems here. > Unfortunately the 4 different computers > I attempted to test on did not have the > correct Nvidia hardware or software. > Why? I do not know but will try one more > later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
> > On Sat, Dec 21, 2024 at 4:04 PM Andrew > Randrianasulu <randrianasulu@gmail.com> > wrote: > > > > вс, 22 дек. 2024 г., 01:53 Phyllis > Smith <phylsmith2017@gmail.com>: > > Andrew, > > most likely our nv headers > drifted from that > ffmpeg-7.0/proprietary > driver assumes at runtime. > > I have been wondering about > nv-codec-headers as we are at: > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 > but I am unsure about updating to: > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 > because if you look at: > https://github.com/FFmpeg/nv-codec-headers/releases/ > the release versions go from > 12.xx to 8.x and it is really > weird AND there is no year on > the release dates but just day > and month. > Since it is such an important > part of ffmpeg inside CinGG, I > am concerned but will at least > try the 12.2.72.0 just to see > what it does. > > > in theory it should give users of > new nvidia hardware av1 encoding > ...... but not sure how it will work > with older drivers and hardware. > > > > > > you can try to install > something like > nv-codec-headers and then add >
The nvidia driver version 470 has problems with latest kernel version 6.12, but patch is available. I tested it with gtx 650. Best regards, Andrey вс, 29 дек. 2024 г., 16:07 Terje J. Hanssen via Cin < cin@lists.cinelerra-gg.org>:
Den 28.12.2024 01:51, skrev Terje J. Hanssen:
Den 27.12.2024 22:25, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 23:56 Terje J. Hanssen <terjejhanssen@gmail.com>:
Den 27.12.2024 21:37, skrev Terje J. Hanssen:
Den 27.12.2024 21:07, skrev Andrew Randrianasulu:
пт, 27 дек. 2024 г., 22:55 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, are you saying that nv-codec-headers at 10.0.26.0 in the current CinGG work with your GT710? and that the newest 12.2.72.0 does not? If so, then maybe we should not upgrade. When I look at the support matrix (url below) it does not mention the GT710.
Do not worry, I only used proprietary driver from live dvd to see if nvenc ever work at all. I normally use built-in nouveau, even if it does not support encoding block on any hw.
Do not consider my oddball machine support as showstopper - I definitely can roll my own build ;)
https://gssparks.github.io/post/can-a-geforce-gt710-nvenc/
yeah, gt710 is sort of two-cards-under-same-name, mine is Kepler based so it worked ....
A side comment: Before I rebuilt my legacy SM workstation, I replaced a GT730 with GTX960 on the MSI mobo (from my upgraded Chieftec workstation)
https://www.asus.com/motherboards-components/graphics-cards/asus/gt730-sl-2g...
And then, about 6 months ago, I had the following drivers for GT 730 enabled on Leap15.5 and Slowroll respectively:
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 470.256.02
isn't 470 < 550 ?
Clearly, yes.
Just to verify by searching for gt730 (or gt710), GF 700 series, GF730, Linux 64-bit : https://www.nvidia.com/en-us/drivers/
Driver Version:470.256.02 Release Date:Tue Jun 04, 2024
If Linux 32-bit ARM:
Driver Version:390.157 Release Date:Tue Nov 22, 2022
The opensuse SDB:NVIDIA lists the following driver series for different NV devices/families https://en.opensuse.org/SDB:NVIDIA_drivers#Determination_of_driver_version
Note: NVIDIA driver versions are mapped to the following naming convention listed below. You will need this information when you are ready to install via commandline/zypper.
- G03 = driver v340 = legacy driver (via a community user <https://software.opensuse.org/search?baseproject=ALL&q=g03>) - G04 = driver v390 = legacy driver (Fermi devices) - G05 = driver v470 = driver for Kepler devices - G06 = driver v550 = driver for Maxwell, Pascal devices and up
Note: Some OEMs use different chip familys (Kepler, Fermi) for the same product, as example GeForce 730. *Warning:* Legacy drivers G03 and G04 won't work with the actual kernel 6.6.x series. Use free nouveau drivers instead!
But I do not think opening desktop case just for this test actually good idea ...
do not worry, ff nvidia headers quite small tarball, I always can fish old one from one of my backups ....
Yeah, I don't use the gt730 anymore, as gtx960 was some more powerful.
Device-1: NVIDIA GK208B [GeForce GT 730] driver: nouveau v: kernel
https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new BUT when I do a compare of the old one we have currently and the latest at this website:
https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 I get the following differences which DO NOT say anything about dropping support for any previous Nvidia graphics boards. Below list the comparison which shows what was changed from the older version to the latest
*Bump for (in-dev) 10.0.26.1* Fix macOS build Add linker functions and types Fix CUdeviceptr definition for ARM64 CPU *Bump for (in-dev) 10.0.26.2* Update headers from Video SDK 11.0 *Bump for (in-dev) 11.0.10.1* Add cuModuleGetGlobal Add texture alignment device attributes add cuMemcpy
*Bump for (in-dev) 11.0.10.2*
Fix UB lshift in nvenc struct version macros Add more memory functions Update headers from Video SDK 11.1 Add timeline semaphore type Add CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_TIMELINE_SEMAPHORE_WIN32 Add a subset of CUDA EGLStreams interop API Update forgotten version number in README Fix codestyle Add D3D11 interop functions *Release 11.1.5.2* Prepare for SDK 12.0 update Preliminary update of nvEncodeAPI to SDK 12.0
Add missing SDK 12.0 bits
Update for final release of Video Codec SDK 12.0.16 Remove AV1 High Profile GUID Update README Add cuArrayCreate *Bump for (in-dev) 12.0.16.1* Add functions to identify devices in more ways Add missing cuStreamWaitEvent function Update headers from Video SDK 12.1 Add cuDriverGetVersion Add cuCtxGetCurrent function *Bump for (in-dev) 12.1.14.1* Update headers from Video SDK 12.2
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
> > > > Den 24.12.2024 13:29, skrev Terje J. Hanssen: > > > > > Den 24.12.2024 00:58, skrev Phyllis Smith: > > Downloaded latest version of nv-codec-headers release and built > CinGG with it and no problems here. Unfortunately the 4 different > computers I attempted to test on did not have the correct Nvidia hardware > or software. Why? I do not know but will try one more later. > > > According to the > *Readme *https://github.com/FFmpeg/nv-codec-headers > > *FFmpeg version of headers required to interface with Nvidias codec > APIs.* > *Corresponds to Video Codec SDK version 12.0.16.* > *Minimum required driver versions:* > *Linux: 550.54.14 or newer* > > As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs > with the following nv driver version on Slowroll: > > Graphics: > Device-1: Intel HD Graphics 530 driver: i915 v: kernel > Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135 > > > > If you want and have a test download, I can give it a try on my > legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) > works as previously. > > > On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < > randrianasulu@gmail.com> wrote: > >> >> >> вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>: >> >>> Andrew, >>> >>>> most likely our nv headers drifted from that >>>> ffmpeg-7.0/proprietary driver assumes at runtime. >>>> >>> I have been wondering about nv-codec-headers as we are at: >>> >>> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 >>> but I am unsure about updating to: >>> >>> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 >>> because if you look at: >>> https://github.com/FFmpeg/nv-codec-headers/releases/ >>> the release versions go from 12.xx to 8.x and it is really weird >>> AND there is no year on the release dates but just day and month. >>> Since it is such an important part of ffmpeg inside CinGG, I am >>> concerned but will at least try the 12.2.72.0 just to see what it does. >>> >> >> in theory it should give users of new nvidia hardware av1 encoding >> ...... but not sure how it will work with older drivers and hardware. >> >> >> >> >>> >>>> you can try to install something like nv-codec-headers and then >>>> add >>>> >>>> > >
-- Cin mailing list Cin@lists.cinelerra-gg.org https://lists.cinelerra-gg.org/mailman/listinfo/cin
Andrew, I am having difficulty interpreting the statement "min driver note". According to: https://forums.developer.nvidia.com/t/h264-nvenc-the-minimum-required-nvidia... the minimum nvidia driver is 520-56-06. Is that what you mean should be included in the release notes? On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
> most likely our nv headers drifted from that ffmpeg-7.0/proprietary > driver assumes at runtime. > I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
> you can try to install something like nv-codec-headers and then add > >
сб, 28 дек. 2024 г., 01:22 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, I am having difficulty interpreting the statement "min driver note". According to:
https://forums.developer.nvidia.com/t/h264-nvenc-the-minimum-required-nvidia...
the minimum nvidia driver is 520-56-06.
Is that what you mean should be included in the release notes?
I think this is official supported minimum driver version , yes. But in slackware for example you can find older nvidia-legacy driver I used for my older card. Remember, NVIDIA, like all other companies, literally in business selling you new hardware! So they "dark pattern" their info for 'nudging' users into using new hardware .....
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
> Andrew, > >> most likely our nv headers drifted from that ffmpeg-7.0/proprietary >> driver assumes at runtime. >> > I have been wondering about nv-codec-headers as we are at: > > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 > but I am unsure about updating to: > > https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 > because if you look at: > https://github.com/FFmpeg/nv-codec-headers/releases/ > the release versions go from 12.xx to 8.x and it is really weird AND > there is no year on the release dates but just day and month. > Since it is such an important part of ffmpeg inside CinGG, I am > concerned but will at least try the 12.2.72.0 just to see what it does. >
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
> >> you can try to install something like nv-codec-headers and then add >> >>
Updated nv-codec-headers from 10.0.26.0 to 12.2.72.0 and checked into GIT. Andrey's builds should be automatically going now for newer O/S package creation. 550.54.14 is "supposedly" the minimum required driver version for Linux according to the README associated with the release. And this is the same as what was written for the previous version. So whatever worked before should still work. The previous information from a forum post said that the minimum nvidia driver is 520-56-06 BUT I have since reread that and determined that that was November 2022 - so out of date. On Fri, Dec 27, 2024 at 3:37 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
сб, 28 дек. 2024 г., 01:22 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, I am having difficulty interpreting the statement "min driver note". According to:
https://forums.developer.nvidia.com/t/h264-nvenc-the-minimum-required-nvidia...
the minimum nvidia driver is 520-56-06.
Is that what you mean should be included in the release notes?
I think this is official supported minimum driver version , yes. But in slackware for example you can find older nvidia-legacy driver I used for my older card.
Remember, NVIDIA, like all other companies, literally in business selling you new hardware! So they "dark pattern" their info for 'nudging' users into using new hardware .....
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg.
https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020.
https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen < terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the *Readme *https://github.com/FFmpeg/nv-codec-headers
*FFmpeg version of headers required to interface with Nvidias codec APIs.* *Corresponds to Video Codec SDK version 12.0.16.* *Minimum required driver versions:* *Linux: 550.54.14 or newer*
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
> > > вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>: > >> Andrew, >> >>> most likely our nv headers drifted from that >>> ffmpeg-7.0/proprietary driver assumes at runtime. >>> >> I have been wondering about nv-codec-headers as we are at: >> >> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 >> but I am unsure about updating to: >> >> https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 >> because if you look at: >> https://github.com/FFmpeg/nv-codec-headers/releases/ >> the release versions go from 12.xx to 8.x and it is really weird >> AND there is no year on the release dates but just day and month. >> Since it is such an important part of ffmpeg inside CinGG, I am >> concerned but will at least try the 12.2.72.0 just to see what it does. >> > > in theory it should give users of new nvidia hardware av1 encoding > ...... but not sure how it will work with older drivers and hardware. > > > > >> >>> you can try to install something like nv-codec-headers and then >>> add >>> >>>
Den 28.12.2024 21:37, skrev Phyllis Smith:
Updated nv-codec-headers from 10.0.26.0 to 12.2.72.0 and checked into GIT. Andrey's builds should be automatically going now for newer O/S package creation.
550.54.14 is "supposedly" the minimum required driver version for Linux according to the README associated with the release. And this is the same as what was written for the previous version. So whatever worked before should still work. The previous information from a forum post said that the minimum nvidia driver is 520-56-06 BUT I have since reread that and determined that that was November 2022 - so out of date.
"Supported Products" for the latest Linux driver v. 550.142 is available here https://www.nvidia.com/en-us/drivers/details/237853/
On Fri, Dec 27, 2024 at 3:37 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
сб, 28 дек. 2024 г., 01:22 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, I am having difficulty interpreting the statement "min driver note". According to:
https://forums.developer.nvidia.com/t/h264-nvenc-the-minimum-required-nvidia...
the minimum nvidia driver is 520-56-06.
Is that what you mean should be included in the release notes?
I think this is official supported minimum driver version , yes. But in slackware for example you can find older nvidia-legacy driver I used for my older card.
Remember, NVIDIA, like all other companies, literally in business selling you new hardware! So they "dark pattern" their info for 'nudging' users into using new hardware .....
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Den 28.12.2024 21:37, skrev Phyllis Smith:
Updated nv-codec-headers from 10.0.26.0 to 12.2.72.0 and checked into GIT. Andrey's builds should be automatically going now for newer O/S package creation.
I just tested the RPM on Leap 15.5 and can confirm that h264_nvenc.mp4 and h265_nvenc.mp4 works fine on GTX960. Thanks to Phyllis and Andrey
550.54.14 is "supposedly" the minimum required driver version for Linux according to the README associated with the release. And this is the same as what was written for the previous version. So whatever worked before should still work. The previous information from a forum post said that the minimum nvidia driver is 520-56-06 BUT I have since reread that and determined that that was November 2022 - so out of date.
On Fri, Dec 27, 2024 at 3:37 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
сб, 28 дек. 2024 г., 01:22 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, I am having difficulty interpreting the statement "min driver note". According to:
https://forums.developer.nvidia.com/t/h264-nvenc-the-minimum-required-nvidia...
the minimum nvidia driver is 520-56-06.
Is that what you mean should be included in the release notes?
I think this is official supported minimum driver version , yes. But in slackware for example you can find older nvidia-legacy driver I used for my older card.
Remember, NVIDIA, like all other companies, literally in business selling you new hardware! So they "dark pattern" their info for 'nudging' users into using new hardware .....
On Thu, Dec 26, 2024 at 7:20 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
пт, 27 дек. 2024 г., 05:09 Phyllis Smith <phylsmith2017@gmail.com>:
Great so it will be included in the next release. Thank you very much.
Can you also please copypaste min driver notes into our release notes so it will be easier for me to track it in case my older gt710 still stuck at older driver version (not a blocker, I surely can swap headers back to 2020 version locally).
Thanks and sorry for relative inactivity. I hope to pick up some steam ..... may he after formal New Year date?
Have (good) holidays!
On Thu, Dec 26, 2024, 17:10 Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 26.12.2024 23:15, skrev Phyllis Smith:
Here is the AppImage with the latest nvencoders for Nvidia graphics that goes with ffmpeg. https://cinelerra-gg.org/download/testing/cin-x86_64_nvencupdate.AppImage For comparison, here is the appimage containing the original nvencoders from 2020. https://cinelerra-gg.org/download/images/CinGG-20241031-x86_64.AppImage
I've did a quick test with h64_nvenc.mp4 and h265_nvenc.mp4 (w/setting format=Not Interlaced) Both AppImages works with similar fps speeds on my Nvidia GF GTX 960.
I am not sure if encoding with formats h264_nvenc.mp4, h265_nvenc.mp4, or h264_nvenc.qt actually work from an AppImage that was created on a computer that does not have an Nvidea graphics card or its software. But for some reason, I do not think that that matters and it should work.
Yes, nvenc works (as it correspondingly was expected to do with onevpl for Intel qsv/hw).
On Thu, Dec 26, 2024 at 2:11 PM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 13:29, skrev Terje J. Hanssen:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
According to the /Readme /https://github.com/FFmpeg/nv-codec-headers
/FFmpeg version of headers required to interface with Nvidias codec APIs./ /Corresponds to Video Codec SDK version 12.0.16./ /Minimum required driver versions:/ /Linux: 550.54.14 or newer/
As shown in my previous post, the legacy (2015) NV-GF GTX 960 runs with the following nv driver version on Slowroll:
Graphics: Device-1: Intel HD Graphics 530 driver: i915 v: kernel Device-2: NVIDIA GM206 [GeForce GTX 960] driver: nvidia v: 550.135
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
Terje, I am building an AppImage now and will let you know when it is available. I also read your next email of your test results. But if you have the time, I would still appreciate you testing the appimage I will be downloading since it is the default one that will go out to the users on Dec. 31. did not have the correct Nvidia hardware or software. Why? I do not know
but will try one more later.
The above is my problem also and the one computer that I did the previous test on and is quoted in the manual, was updated without the Nvidia drivers. More later. Thanks for the help. On Tue, Dec 24, 2024 at 5:29 AM Terje J. Hanssen <terjejhanssen@gmail.com> wrote:
Den 24.12.2024 00:58, skrev Phyllis Smith:
Downloaded latest version of nv-codec-headers release and built CinGG with it and no problems here. Unfortunately the 4 different computers I attempted to test on did not have the correct Nvidia hardware or software. Why? I do not know but will try one more later.
If you want and have a test download, I can give it a try on my legacy GF GTX 960 SkyLake workstation, to see if your new AppImage (still) works as previously.
On Sat, Dec 21, 2024 at 4:04 PM Andrew Randrianasulu < randrianasulu@gmail.com> wrote:
вс, 22 дек. 2024 г., 01:53 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew,
most likely our nv headers drifted from that ffmpeg-7.0/proprietary driver assumes at runtime.
I have been wondering about nv-codec-headers as we are at: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n10.0.26.0 but I am unsure about updating to: https://github.com/FFmpeg/nv-codec-headers/releases/tag/n12.2.72.0 because if you look at: https://github.com/FFmpeg/nv-codec-headers/releases/ the release versions go from 12.xx to 8.x and it is really weird AND there is no year on the release dates but just day and month. Since it is such an important part of ffmpeg inside CinGG, I am concerned but will at least try the 12.2.72.0 just to see what it does.
in theory it should give users of new nvidia hardware av1 encoding ...... but not sure how it will work with older drivers and hardware.
you can try to install something like nv-codec-headers and then add
participants (4)
-
Andrew Randrianasulu -
Phyllis Smith -
Terje J. Hanssen -
Андрей Спицын