CyberLink Community Forum
where the experts meet
| Advanced Search >
Issues With The Latest Update
Reply to this topic
StrikeHD [Avatar]
Newbie Private Message Joined: Sep 11, 2018 18:30 Messages: 4 Offline
[Post New]
Hi there!

Been using Powerdirector all the way back since Powerdirector 12 and I'm really loving the new features and quality of life improvements in the latest version. However, I've noticed something very strange, I use hardware acceleration for my renders and in Powerdirector 18 - and all versions before that - my GPU would put in almost all of the work, my GPU would normally be at 60% usage whereas my CPU would be at about 30%. With the exact same settings in PD19, my usage tends to be 70% for CPU, and around 40-50% for GPU. Not only that, my average render time has almost doubled. I've already made sure my drivers are up to date, not sure what else could help. Would really appreciate any advice about this - Thanks!

I'll list my specs here, in case that helps:

Motherboard: Gigabyte Z390 AORUS PRO
Processor: Intel Core i9 Eight Core Processor i9-9900k (3.6GHz) 16mb cache
GPU: 8GB ASUS ROG STRIX GEFORCE RTX 2080
RAM: 32GB Corsair Vengeance RGB DDR4 3000MHz
Storage Drives: 250gb Seagate Barracuda 2.5" SSD + 4TB Seagate Barracuda Pro 3.5" 7200RPM Hard Drive
OS: Windows 10 Pro
PSU: Corsair 850W RMx Series Modular 80 Plus Gold Ultra Quiet
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
I don't have any benchmarks between the two PD versions, and my i9-9900K system died the day after PD19 was released and I replaced it with a 16 core/32 thread AMD CPU. I've only done one production run and both my CPU and nVidia GPU were in the 30% range producing to 4K 60p AVC.

It might help to see more relevent details about your system from the DxDiag test as described in the Read Me Before Posting guide, and also which output profile you're producing to.

YouTube/optodata


DS365 | Win10 Pro | Ryzen 9 3950X | RTX 2070 | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
Reply
StrikeHD [Avatar]
Newbie Private Message Joined: Sep 11, 2018 18:30 Messages: 4 Offline
[Post New]
Quote I don't have any benchmarks between the two PD versions, and my i9-9900K system died the day after PD19 was released and I replaced it with a 16 core/32 thread AMD CPU. I've only done one production run and both my CPU and nVidia GPU were in the 30% range producing to 4K 60p AVC.

It might help to see more relevent details about your system from the DxDiag test as described in the Read Me Before Posting guide, and also which output profile you're producing to.


Hi!

Thanks for your response - I've attached the dxdiag test, as well as a screenshot of my custom H.264 AVC profile. Getting late in my country now so I'll be heading off for the night, but I will be able to answer additional questions once I wake up.

Hope it helps,
[Thumb - PD19 Profile Video.png]
 Filename
PD19 Profile Video.png
[Disk]
 Description
 Filesize
15 Kbytes
 Downloaded:
8 time(s)
 Filename
DxDiag 21-09-2020.txt
[Disk]
 Description
 Filesize
85 Kbytes
 Downloaded:
13 time(s)
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
Thanks for the details. Everything looks good so I'd suggest trying two things to see if one or both might make a difference.

In your custom profile, try turning off Use dynamic GOP and see if that changes anything.

You may also want to check which version of the nVidia driver you have installed. If you have the Game Ready version, try doing a clean install of the Studio Driver.

YouTube/optodata


DS365 | Win10 Pro | Ryzen 9 3950X | RTX 2070 | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
Reply
StrikeHD [Avatar]
Newbie Private Message Joined: Sep 11, 2018 18:30 Messages: 4 Offline
[Post New]
Quote Thanks for the details. Everything looks good so I'd suggest trying two things to see if one or both might make a difference.

In your custom profile, try turning off Use dynamic GOP and see if that changes anything.

You may also want to check which version of the nVidia driver you have installed. If you have the Game Ready version, try doing a clean install of the Studio Driver.


Thank you! Turning of dynamic GOP worked perfectly and if anything its faster than ever before, really appreciate the help!
Reply
StrikeHD [Avatar]
Newbie Private Message Joined: Sep 11, 2018 18:30 Messages: 4 Offline
[Post New]
Quote


Thank you! Turning of dynamic GOP worked perfectly and if anything its faster than ever before, really appreciate the help!


Hmm, strange. Scratch that, same issue is back again. Only the first render I did with the new settings was free of the issue. I guess for now we'll have to assume its an issue with the new edition of Powerdirector. Hopefully performance enhancing patches will be on the way soon enough
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
Any performance patches aren't likely to materialize all by themselves, they'll need to be created in response to problems submitted to the official support channel that can be replicated and understood in their test lab.

There are actually many kinds of edits that will slow producing down dramatically, such as applying color look-up tables (CLUTs).

It might help to see if it's something in your second project that might be causing the slower producing, and you might want to produce the first project again as a verification. Even a producing just a few percent will probably let you know if PD is slower with the first project the second time through or not.
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4493 Offline
[Post New]
Quote Been using Powerdirector all the way back since Powerdirector 12 and I'm really loving the new features and quality of life improvements in the latest version. However, I've noticed something very strange, I use hardware acceleration for my renders and in Powerdirector 18 - and all versions before that - my GPU would put in almost all of the work, my GPU would normally be at 60% usage whereas my CPU would be at about 30%. With the exact same settings in PD19, my usage tends to be 70% for CPU, and around 40-50% for GPU. Not only that, my average render time has almost doubled. I've already made sure my drivers are up to date, not sure what else could help. Would really appreciate any advice about this - Thanks!

PD19 has a big encoding performance deficit relative to PD18 when CPU decoding of the timeline is done. The CPU doing the decoding is also in line with your % increased load observations. Depending on timeline content, you might make sure hardware decoding is checked in pref > Hardware Acceleration and see if encoding performance improves. When GPU decoding of timeline can be done as well as GPU encoding, performance with PD18 is the same. When CPU decoding of the timeline is used, encoding performance can be significantly worse than PD18, 1.5-2X worse not uncommon.

Jeff
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
Quote When CPU decoding of the timeline is used, encoding performance can be significantly worse than PD18, 1.5-2X worse not uncommon.

Can you share the testing data for this?
Reply
marshm777 [Avatar]
Newbie Private Message Joined: Jul 01, 2020 16:21 Messages: 4 Offline
[Post New]
Quote Hi there!

Been using Powerdirector all the way back since Powerdirector 12 and I'm really loving the new features and quality of life improvements in the latest version. However, I've noticed something very strange, I use hardware acceleration for my renders and in Powerdirector 18 - and all versions before that - my GPU would put in almost all of the work, my GPU would normally be at 60% usage whereas my CPU would be at about 30%. With the exact same settings in PD19, my usage tends to be 70% for CPU, and around 40-50% for GPU. Not only that, my average render time has almost doubled. I've already made sure my drivers are up to date, not sure what else could help. Would really appreciate any advice about this - Thanks!

I'll list my specs here, in case that helps:

Motherboard: Gigabyte Z390 AORUS PRO
Processor: Intel Core i9 Eight Core Processor i9-9900k (3.6GHz) 16mb cache
GPU: 8GB ASUS ROG STRIX GEFORCE RTX 2080
RAM: 32GB Corsair Vengeance RGB DDR4 3000MHz
Storage Drives: 250gb Seagate Barracuda 2.5" SSD + 4TB Seagate Barracuda Pro 3.5" 7200RPM Hard Drive
OS: Windows 10 Pro
PSU: Corsair 850W RMx Series Modular 80 Plus Gold Ultra Quiet


Aloha,
I'm getting all kinds of glitches with the latest verson and sometimes it stops responding all together. Didn't have that issue before!
Anyone got ideas??
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4493 Offline
[Post New]
Quote Can you share the testing data for this?

One can evaluate with about anything as long as you still have PD18 available.

For a very basic test that could even be replicated by CL perhaps, I took the 3 skateboard files, Skateboard01.mp4, 02, and 03 and duplicated in the timeline 5 times to create some working footage. I then produced this to a AVC MP4, default 1920x1080/60p 40Mbps profile, again just to create some higher complexity working footage with reasonable decode load with the canned source files.

1) place sport 01.jpg in the timeline (needed because of another continued PD anomaly/bug)
2) place produced footage above in timeline (see timeline pic)
3) produce to same profile for simplicity with various decode and encode device options but not utilizing SVRT, we want to actually encode the timeline
4) do the same with PD18 and PD19

Results shown in attached table. Basically, for this simple transcode case, anytime the CPU is used for decoding, PD19 is nearly 1.5X slower encoding than PD18. Results will vary based on hardware but PD19 a significant step backward on encode performance when CPU needs to do the decode tasks which is often needed to prevent other anomalies often shown in prior versions.

So what often is referenced as the staple, CPU decode and encode, simply got significantly slower with PD19 for the same perceived quality.

Jeff
[Thumb - PD19_encode_vs_PD18.png]
 Filename
PD19_encode_vs_PD18.png
[Disk]
 Description
 Filesize
11 Kbytes
 Downloaded:
18 time(s)
[Thumb - PD19_Encode.png]
 Filename
PD19_Encode.png
[Disk]
 Description
 Filesize
913 Kbytes
 Downloaded:
9 time(s)
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
This is very helpful. Thanks very much for spelling out your methodology and results!
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
I just tried out this approach on my desktop, and with a 16-core CPU there is no meaningful difference between PD18 and PD19 as far as producing specs are concerned.

Here are my results (with 2 added lines to show actual PD18 + PD19 produce times):



I've also created a test project in PD18 if anyone else wants to try, and you can download it from this OneDrive folder (no need to sign in).

I ran these tests with the latest nVidia Studio Driver (456.38) and I closed/reopened PD after changing the Hardware decode option. Bitrates and file sizes were obtained using MediInfo v20.08.

This message was edited 1 time. Last update was at Sep 25. 2020 14:46



YouTube/optodata


DS365 | Win10 Pro | Ryzen 9 3950X | RTX 2070 | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4493 Offline
[Post New]
Probably have to wait for a few other users to see what happens for them if they understand the settings options. Would be interesting to see if it affected the OP issue too. Many posts starting to appear on poor encode performance so potentially many things at play, like scrolling titles yet again.

Two possible differences in this small test case, 365 vs perpetual and AMD vs Intel CPU. I did try 456.38 with no significant change as expected, I doubt the issue is on the GPU side. I also tried an older GTX 1070 which shed no additional insight as the issue occurs with pure CPU encode and decode.

Jeff
Reply
Chris_T53 [Avatar]
Newbie Private Message Joined: Apr 05, 2020 05:53 Messages: 11 Offline
[Post New]
Quote Probably have to wait for a few other users to see what happens for them if they understand the settings options. Would be interesting to see if it affected the OP issue too. Many posts starting to appear on poor encode performance so potentially many things at play, like scrolling titles yet again.

Two possible differences in this small test case, 365 vs perpetual and AMD vs Intel CPU. I did try 456.38 with no significant change as expected, I doubt the issue is on the GPU side. I also tried an older GTX 1070 which shed no additional insight as the issue occurs with pure CPU encode and decode.

Jeff


Scrolling titles appear to be my problem as well. Rendering crawls from the point where they start. Didn't have the problem with PD18.
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
Quote Scrolling titles appear to be my problem as well. Rendering crawls from the point where they start. Didn't have the problem with PD18.

Unfortunately I think there's a serious issue when working with titles in PD19, especially scrolling ones.

I see the same issue as you when using your shared project from the other discussion, but that project has been saved in PD19 and can no longer be opened in PD18 for comparison so I took the PD18 project I created here and added a single scrolling title with the same font and shadow settings as in yours. The new project is in the same shared OneDrive folder.

In PD18, I can produce the Forum Test Project + title clip in 1:54 using the same configuration as the 4th column in my table above, which is about an 85% increase in production time to deal with the animation and shading of the scrolling text. When using the nVidia GPU for encoding and decoding (2nd column) it took 1:36, or about 48% longer then the simple video project.

With PD19, however, the project becomes unusable.

Normally, projects load within a few seconds on my desktop, but this project takes almost a minute to load and about another 30 sec or so to bring up the Produce screen. The only difference is a single title with dozens of lines of text copied from the Falling Down default title with the effect changed to Scroll Up.

Clicking on Start can take another half minute to see PD take action, and when producing finally starts it's at an absolute crawl: 4 min 10 sec to produce the first 3 seconds of the clip to the HD MP4 60p profile used above. That's 3.5 hours to produce a 2.5 min HD clip with scrolling text vs. under 2 min for the same thing in PD18.

During this time, the GPU load is zero because it's only encoding one frame every couple of seconds instead of the dozens or hundreds/sec it's capable of. The CPU is running along at 40% utilization but I have no idea what's its spending all that energy doing. Someone at CL will need to take a deep look at this.

I believe there's been a serious issue with titles since at least PD17 when I wrote the Death By Titles post, but the issue wasn't widespread and it wasn't ever resolved back then. Maybe this project will allow the issue to be explored more fully so the underlying cause(s) can finally be addressed.
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4493 Offline
[Post New]
Quote Unfortunately I think there's a serious issue when working with titles in PD19, especially scrolling ones.

I see the same issue as you when using your shared project from the other discussion, but that project has been saved in PD19 and can no longer be opened in PD18 for comparison so I took the PD18 project I created here and added a single scrolling title with the same font and shadow settings as in yours. The new project is in the same shared OneDrive folder.

In PD18, I can produce the Forum Test Project + title clip in 1:54 using the same configuration as the 4th column in my table above, which is about an 85% increase in production time to deal with the animation and shading of the scrolling text. When using the nVidia GPU for encoding and decoding (2nd column) it took 1:36, or about 48% longer then the simple video project.

With PD19, however, the project becomes unusable.

Normally, projects load within a few seconds on my desktop, but this project takes almost a minute to load and about another 30 sec or so to bring up the Produce screen. The only difference is a single title with dozens of lines of text copied from the Falling Down default title with the effect changed to Scroll Up.

Clicking on Start can take another half minute to see PD take action, and when producing finally starts it's at an absolute crawl: 4 min 10 sec to produce the first 3 seconds of the clip to the HD MP4 60p profile used above. That's 3.5 hours to produce a 2.5 min HD clip with scrolling text vs. under 2 min for the same thing in PD18.

During this time, the GPU load is zero because it's only encoding one frame every couple of seconds instead of the dozens or hundreds/sec it's capable of. The CPU is running along at 40% utilization but I have no idea what's its spending all that energy doing. Someone at CL will need to take a deep look at this.

I believe there's been a serious issue with titles since at least PD17 when I wrote the Death By Titles post, but the issue wasn't widespread and it wasn't ever resolved back then. Maybe this project will allow the issue to be explored more fully so the underlying cause(s) can finally be addressed.


Long discussion appears vastly off topic for OP, but yes, as I had mentioned, scrolling title issues yet again but your PD17 reference a little too kind. Users have been putting up with title performance issues since at least PD12 if not prior, that's the level of focused resolution it's received by CL, 7+yrs no proper solution, https://forum.cyberlink.com/forum/posts/list/30455.page#post_box_167232

Jeff
Reply
prevaljo [Avatar]
Member Private Message Joined: Oct 01, 2017 22:19 Messages: 51 Offline
[Post New]
Interesting thread. I haven't tried producing anything in PD 19 but thought I'd take a look at the performance issue identified. Currently I have both PD18 and PD19 running on my desktop.

Note this is a non scientific post..

PD19

1. I created a simple video on the timeline as suggested in the thread Skateboard 1, 2, 3, replicated 5 times to give a simple video about 3 mins long. Added a default scrolling title at the beginning. Tried to preview it, it was very slow as in unwatchable in the preview, so I altered the preview resolution from the default to a lower resolution, the clip previewed OK.

2. Tried to produce the clip as an 1920 x 1080 25fps MP4, the production was incredibly slow, the CPU was at approx 90% and the GPU pretty close to 0.

Not what I expected from my desktop. I then repeated the excercise with PD18..

PD18


  1. Once again I created a simple video on the timeline using Skateboard replicated a number of times to give a simple video about 3 mins long. Added a default scrolling title at the beginning. It previewed OK in a higher resolution than PD19.



2. Tried to produce the clip as an 1920 x 1080 25fps MP4, the production was fast as expected and have previously experienced, the CPU was at approx 1-2% and the GPU pretty close to 80%.


Given this experience I doubt that I would use PD19 to produce any video. Has anybody represented this issue to Cyberlink as it is either a bug or a configuration issue. ?
Reply
optodata
Senior Contributor Private Message Location: California, USA Joined: Sep 16, 2011 16:04 Messages: 5753 Offline
[Post New]
As the topic of this discussion is Issues With The Latest Update and I've built on the testing method you outlined here, it seemed appropriate to add my new findings as well.

I've also now tried out this test on my Surface Book 3, and it took 5 min 10 sec to produce the first 2 seconds of the scrolling title project, so whatever's going on seems to be scaled by the system's processing power. In both of my tests, producing slowed down approximately 100x over a non-scrolling-titled project

I've reported the issue on CS002226952 and since that didn't look like it went through, CS002226953.
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4493 Offline
[Post New]
Quote One can evaluate with about anything as long as you still have PD18 available.

For a very basic test that could even be replicated by CL perhaps, I took the 3 skateboard files, Skateboard01.mp4, 02, and 03 and duplicated in the timeline 5 times to create some working footage. I then produced this to a AVC MP4, default 1920x1080/60p 40Mbps profile, again just to create some higher complexity working footage with reasonable decode load with the canned source files.

1) place sport 01.jpg in the timeline (needed because of another continued PD anomaly/bug)
2) place produced footage above in timeline (see timeline pic)
3) produce to same profile for simplicity with various decode and encode device options but not utilizing SVRT, we want to actually encode the timeline
4) do the same with PD18 and PD19

Results shown in attached table. Basically, for this simple transcode case, anytime the CPU is used for decoding, PD19 is nearly 1.5X slower encoding than PD18. Results will vary based on hardware but PD19 a significant step backward on encode performance when CPU needs to do the decode tasks which is often needed to prevent other anomalies often shown in prior versions.

So what often is referenced as the staple, CPU decode and encode, simply got significantly slower with PD19 for the same perceived quality.

Jeff

I just wanted to follow-up that the issue I was seeing of PD19 being ~1.5X slower than PD18 for CPU decode/encode has been resolved. I now see nearly unity ratios for PD19/PD18 for all render configurations. For some reason the Virus Protection software I use had an anomaly with PD19. It had no such issue with PD18, 17, 16. In fact, after resolution, the PD18 encode times are identical, it only reduced PD19 encode times.

Jeff
Reply
Reply to this topic
Powered by JForum 2.1.8 © JForum Team