Announcement: Our new CyberLink Feedback Forum has arrived! Please transfer to our new forum to provide your feedback or to start a new discussion. The content on this CyberLink Community forum is now read only, but will continue to be available as a user resource. Thanks!
CyberLink Community Forum
where the experts meet
| Advanced Search >
Nvidia GTX650 PD13 Produce Anomaly with Hardware Video Encoding
JL_JL [Avatar]
Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline
[Post New]
I was trying to do some frame accurate position in a video and came upon this anomaly. I was working with Canon 1920x1080 24Mbps source but could replicate the same issue with the basic sample file Boats.wmv. This issue is, if one “Produces” the video with an Nvidia card (GTX650) with “Hardware video encoder” selected on the “Produce” page with a M2TS container H.264, 1920x1080 24Mbps or even the 16Mbps profile, randomly PD13 produces what I’ll call a short duration clip. My simple test procedure to show the random anomaly is as follows:

1) NTSC timeline
2) Put 1 Boats.wmv in the timeline
3) Produce to H.264, M2TS container, 1920x1080, 24Mbps, United States (NTSC)
4) Select Edit, Produce and simply produce again changing nothing. I did this loop 20 times to show the random variability.

For the M2TS container, the results are shown in pic1. As you can see, repetitive “Produce” of the exact same timeline results in various duration of produced file. From the nearly correct 13:00 duration to 12:21, nearly 2.5% shorter. That’s rather significant when one needs duration accuracy. Also note out of these 20 loops of the identical “Produce” operation, only 4 produce operations are nearly correct at the original Boats.wmv duration of 13:01. So 20% of the time PD13 can render the Boats.wmv successfully to a H.264 M2TS container.

If the above is repeated for the MKV container, maintaining the exact same produce settings, PD13 consistently renders a 12:29 duration file. Pretty good results as shown in pic2.

If one does essentially the same steps as above but actually uses the “Create Disc” module to create a m2ts file within a BD file strcture with the same H.264, 1920x1080 24Mbps settings and enables the hardware video encoder during the “Create a folder” burn process and then loads all 20 of the created 00000.m2ts into the PD13 media library one notices that the duration are identical and all 13:01 as shown in pic3.

Bottom line, I’d say PD13 has some random anomaly in the “Produce” operation when a Nvidia GTX650 GPU is used with Hardware video encoder that makes frame accurate editing and positioning difficult. If I switch to CPU encoding or use a GTX950 or GTX970 the issue does not occur. Maybe it’s Kepler based GPU’s, 600, 700 series or maybe even more precisely the GK107 Kepler as in my GTX650 and PD13. Oddly, the “Create Disc” operation with Hardware video encoder” is always correct duration so it’s not just a GTX650 issue within PD13. To me looks like a PD13 "Produce" function issue.

PD13 V3130
Nvidia 355.82 driver although tried several recent releases
Win7/64

Same test in a Win10/64 with same GTX650 consistently produced the 12:21 duration for 40 tries of the M2TS container, no random correct durations of 13:00, why such short duration for this container during “Produce”, I’m unsure. Everything else with Win10/64 repeated as above.

Jeff
[Thumb - mkv.png]
 Filename
mkv.png
[Disk]
 Description
pic2: MKV container "Produce" file durations
 Filesize
522 Kbytes
 Downloaded:
101 time(s)
[Thumb - m2ts.png]
 Filename
m2ts.png
[Disk]
 Description
pic1: M2TS container "Produce" file durations
 Filesize
485 Kbytes
 Downloaded:
102 time(s)
[Thumb - BD.png]
 Filename
BD.png
[Disk]
 Description
pic3: BD created m2ts file duration results
 Filesize
544 Kbytes
 Downloaded:
99 time(s)
Dafydd B [Avatar]
Senior Contributor Joined: Aug 26, 2006 08:20 Messages: 11973 Offline
[Post New]
Hi JL_JL,
I have forwarded your observations to CyberLink and asked for your findings to be replicated/tested. I'll post a followup when I have more information.
Dafydd
JL_JL [Avatar]
Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline
[Post New]
Quote: Hi JL_JL,
I have forwarded your observations to CyberLink and asked for your findings to be replicated/tested. I'll post a followup when I have more information.
Dafydd


Thanks. Curious what CL or others find.

Attached is one of the "short" duration PD13 produced H.264 M2TS container added to the timeline and compared to Boats.wmv in the timeline. I've also added the more correct MKV produced file for comparison. One can see the obvious short duration. One can also see some shift in the audio relative to original Boats.wmv.

Jeff
[Thumb - timeline.png]
 Filename
timeline.png
[Disk]
 Description
PD13 "Produce" m2ts container in timeline
 Filesize
627 Kbytes
 Downloaded:
98 time(s)
ynotfish
Senior Contributor Location: N.S.W. Australia Joined: May 08, 2009 02:06 Messages: 9977 Offline
[Post New]
Hi Jeff -

Just chiming in to confirm your observations. Same process - GTX680

With M2TS, the produced files were mostly 12:21, with random occurences of 13:00

MKV gave me a consistent 12:29 & MPEG-4 a consistent 13:01

Screenshot similar to Jeff's attached.

Cheers - Tony
[Thumb - PDR13 Boats M2TS.png]
 Filename
PDR13 Boats M2TS.png
[Disk]
 Description
 Filesize
63 Kbytes
 Downloaded:
103 time(s)

Visit PDtoots. PowerDirector Tutorials, tips, free resources & more. Subscribe!
Full linked Tutorial Catalog
PDtoots happily supports fellow PowerDirector users!
optodata
Senior Contributor Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 8630 Offline
[Post New]
I don't see the random issue after 20 runs of the 24Mbps 60i profile with a 780Ti on Win10



But you'll see that the 3 original produced files (at top) have totally different lengths depending on the profile I used. This table was identical runs of the 24Mbps 60i clip produced at 24Mbps 60i. I'll run the 16Mbps 24p clip next since it has the shortest result, and I'll also try it to different output formats to see if there's a combination that triggers the issue on my machine

YouTube/optodata


DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
JL_JL [Avatar]
Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline
[Post New]
Quote: I don't see the random issue after 20 runs of the 24Mbps 60i profile with a 780Ti on Win10


Yep, as I had quoted in OP, no randomness on WIN10, just the wrong produced duration. Looks like the 700 series has the same issue.

Jeff
optodata
Senior Contributor Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 8630 Offline
[Post New]
OK, this is very interesting. I tried producing the 60i 24Mbps clip to 24p 16Mbps and got a length of 12:06. I then produced to 60p 28Mbps and PD is insisting that these are audio clips only! (I had to attach these because I apparently broke Imgur with my last clip )

Here's a link to a *sample* What does PD13 do when you import it?


EDIT: I can play the clip normally in VLC, but PD13 only shows it as an audio-only track even in the timeline!

EDIT #2: PD12 plays it normally
[Thumb - I broke Imgur.jpg]
 Filename
I broke Imgur.jpg
[Disk]
 Description
Sorry!
 Filesize
57 Kbytes
 Downloaded:
97 time(s)
[Thumb - MediaInfo.jpg]
 Filename
MediaInfo.jpg
[Disk]
 Description
Details from clip
 Filesize
61 Kbytes
 Downloaded:
96 time(s)
[Thumb - Houston problem.jpg]
 Filename
Houston problem.jpg
[Disk]
 Description
Messed up file type
 Filesize
189 Kbytes
 Downloaded:
89 time(s)

This message was edited 4 times. Last update was at Sep 07. 2015 22:43



YouTube/optodata


DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
optodata
Senior Contributor Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 8630 Offline
[Post New]
Maybe there's something with my system. Just take the normal boats.wmv and produce it to 1920x1080 60p 28Mbps. PD13 is showing that as an audio-only file here. If you produce it, do you get different details in MediaInfo than my version of the clip?

Also very strange, the clip listed as "boats anomaly 60p" in the screenshot I posted earlier was accidentally produced as 1440x1080 60p - and it's recognized as a video file by PD13. However, the file properties in the media library show it as 1920x1080 29.97 fps interlaced! MediaInfo shows it as 1440x1080 60p 24.4Mbps, but again, it plays normally in PD13 whereas the actual 1920x1080 60p clip does not.

I just added this bizarre clip to the OneDrive folder I just shared.

I had removed the K-Lite MEGA codec pack fix a few days ago so this is with "pure" Win10. I'll reinstall that and see if that has any effect on PD13's ability to produce and/or play these clips

This message was edited 1 time. Last update was at Sep 07. 2015 23:02



YouTube/optodata


DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
optodata
Senior Contributor Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 8630 Offline
[Post New]
I don't want to hijack this thread, so I'll stop with this update.

I've added two more versions of boats.wmv produced with the K-LITE codec pack to the OneDrive folder. The one ending with "CPU" was produced without HA and it plays normally in PD13. The other K-LITE clip was produced with HA and only shows as an audio file.

To make things more confusing, MediaInfo shows identical specs for both clips, with only minor bit-rate, file size and duration differences - and yet PD13 can't even see the video in the HA-produced clip.

This message was edited 1 time. Last update was at Sep 07. 2015 23:44



YouTube/optodata


DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
JL_JL [Avatar]
Senior Contributor Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 6091 Offline
[Post New]
Yes, many minefields with PD13 when trying to use hardware encoding. 28Mbps has issues as well again in this release as you also found. This and other issues have been brought to the attention in other threads. The 28Mbps initial issue was it would "Produce" but then for any format setting after that one had the unusual "Production Unsuccessful" msg and you needed to shutdown PD13 and try again. Every new release brings changes, sometimes a correction.

In this 3130 release try 2k..4k 60p as its not properly supported either (at least for me), which it was previously. It's difficult to know what to document as each release simply does something different, not necessarily corrected. Release notes are never very specific on particular issues being addressed. Others issues are, 2k..4k, interlace not supported. This was referred and follow-up from mod indicated some issue was experienced during development so now feature is not available for hardware encoding. Wrong frame size vs profile settings was addressed in release 2604 now this release produces a “music” file. Other items too.

This thread had a pdf attached of a bunch of hardware encoding anomalies http://forum.cyberlink.com/forum/posts/list/25/41925.page . The test was simple to potentially show issues, one would think simple things like this are added to a regression suite that needs to be passed prior to release of a version. Apparently not, more of a release and see what consumers find approach.

This thread dealt with the single anomaly which was posted in hopes that a small specific example could be passed to CL and yield resolution.

Jeff
Powered by JForum 2.1.8 © JForum Team