CyberLink Community Forum
where the experts meet
| Advanced Search >
CD4 vs PD14 Color Correction Encode Performance
Reply to this topic
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4541 Offline
[Post New]
A fair warning, this write-up contains a fairly detailed look into color correction performance difference betwen PD14 and CD4. So for those that are simply glad they got a "Produced" file of their project timeline, this discussion probably not for you. A few others may have some interest to continue reading the attached pdf.

Jeff
 Filename
CD4_PD14_Color.pdf
[Disk]
 Description
CD4 vs PD14 Encode Performance
 Filesize
965 Kbytes
 Downloaded:
288 time(s)
Reply
Dafydd B [Avatar]
Senior Contributor Private Message Joined: Aug 26, 2006 08:20 Messages: 11973 Offline
[Post New]
An interesting post Jeff,
I have passed the report onto CyberLink.
Dafydd
Reply
CS2014
Senior Contributor Private Message Location: USA-Eastern Time Zone Joined: Sep 16, 2014 16:44 Messages: 629 Offline
[Post New]
Looks like very nice analytical work to me Jeff - nicely done!

You used 4k source material for the analysis - would you think there would be similar results with clips that are just 1920x1080 HD sort of stuff?

CS

This message was edited 1 time. Last update was at May 04. 2016 14:04

PD13 Ultimate - Build 3516, WIN 8.1, 64 Bit, 16G RAM, Intel Core i5 4460, CPU @ 3.2GHz, NVIDIA GeForce GT720, Graphics Memory(total avail.)-4093MB
LG WH14NS40 Blu-Ray Drive
Reply
Richmond Dan
Senior Contributor Private Message Location: Richmond, VA Joined: Aug 07, 2014 17:17 Messages: 639 Offline
[Post New]
Jeff,

Just a guess to answer your last question in the PDF (why two different approaches?): perhaps CL purchased the CD operation from another software company, then repackaged it under the CL brand, rather than developing their own from scratch.

I have no facts to support this, just top-of-the-head guess.



(Great work, by the way. I'll never need that level of quality, but I'm always pleased when I can follow an explanation. Someone who can explain complex issues clearly is very gifted.)

This message was edited 1 time. Last update was at May 04. 2016 15:25

Regards,
Dan

Power Director 14-Ultimate v.2820, Win-10 64-bit, Intel Core i7-4790 processor (8M Cache, 3.6 GHz),
8GB DDR3 RAM, 1TB 7200 RPM SATA HDD 6.0 Gb/s, NVIDIA GeForce GT 720 ( v 375.95 ) 1GB DDR3

Panasonic HC-V750

[url=http://www.pass
Reply
ynotfish
Senior Contributor Private Message Location: N.S.W. Australia Joined: May 08, 2009 02:06 Messages: 9810 Offline
[Post New]
Yes - excellent work Jeff!

I'd noticed something similar a while back, but never stopped to try to analyse why it was so. I have, however, just repaired an old wheelbarrow & got it functional again - an indication of my level of technical prowess embarassed

As far as I understood the contents of your PDF comparing rendering times between PDR & CDR, I did a similar test here. The results are consistent with yours, though I didn't go into CPU/GPU loads etc.

Original UHD clip - 3840X2160 @ 50Mbps - trimmed to 1 minute & produced to UHD & 1080p. Those two clips were my "source" material. Followed your procedure of adjustment in CDR4 (Exposure - 0.5) then rendering in PDR14 vs. adjustment & rendering directly in CDR4.



CDR took about the same time to render each file, even though one was being resized from UHD to 1080.

That, I think, is consistent with your findings. CDR & PDR go about it in very different ways. One way may be preferable over the other but, viewed on HD monitor/4K monitor/4K TV, these old eyes can't spot any significant difference (I even duplicated & renamed the files so I wouldn't know which one I was watching).

Cheers - Tony

This message was edited 1 time. Last update was at May 04. 2016 22:39


Visit PDtoots. PowerDirector Tutorials, tips, free resources & more. Subscribe!
Full linked Tutorial Catalog
PDtoots happily supports fellow PowerDirector users!
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4541 Offline
[Post New]
Thanks for the confirmation Tony, yes, very similar overall character and totaly consistent findings.

Someone else mentioned to me that my observation was because of differing source bitrates. I thought I'd post a table that I did not include in the PDF because it was simply irrelevant overload. I did consider if bitrate played a role and it's not significant as shown in the attached pic. Basically having source material of the same frame size with bitrates of 20Mbps to 50Mbps had no influence on the encode times to create a color corrected 1920x1080 output.

Jeff
[Thumb - CD_PD_Color.png]
 Filename
CD_PD_Color.png
[Disk]
 Description
 Filesize
43 Kbytes
 Downloaded:
85 time(s)

This message was edited 1 time. Last update was at May 04. 2016 23:14

Reply
CS2014
Senior Contributor Private Message Location: USA-Eastern Time Zone Joined: Sep 16, 2014 16:44 Messages: 629 Offline
[Post New]
Tony,

Looks like with regard to your table that the source material of 1920x1080 - there was not as much of a time savings in rendering compared to the 2160 material. Also looks like in Jeff's table - PD didn't utilize the GPU in it's processing whereas in the CD4 - loads were observed in the GPU and CPU! This would be why it's processing so much faster wouldn't it?

You used the same duration clip for both sources in your table correct? Just curious what was the duration of the clip? Even a 30% gain in less time rendered is significant, at least to me. I think this would be something to use on longer projects - perhaps not the short 5 minute YT videos, but hour long projects - that time savings is nice to have.

Great topic/discussion!

CS

This message was edited 1 time. Last update was at May 05. 2016 06:56

PD13 Ultimate - Build 3516, WIN 8.1, 64 Bit, 16G RAM, Intel Core i5 4460, CPU @ 3.2GHz, NVIDIA GeForce GT720, Graphics Memory(total avail.)-4093MB
LG WH14NS40 Blu-Ray Drive
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4541 Offline
[Post New]
Quote:
You used 4k source material for the analysis - would you think there would be similar results with clips that are just 1920x1080 HD sort of stuff?

That was covered in the PDF. A range of simulated source footage was used and included the more standard 1920x1080 16Mbps and 24Mbps. The benefit of using CD4 is only marginal since in the test the source and target videos have the same frame size. Shown in Table 2 was ~20% benefit. If on the other hand you wanted to down scale your 1920x1080 source to DVD quality, a much smaller frame size output, CD4 would then be very benefical again, like 4-5 times faster than PD14.

You can also see the same effect in Tony's chart. He went from both a source and target of 1920x1080 50Mbps for the color correction and CD4 was a modest ~30% faster, but not like 4x.

Jeff
Reply
CS2014
Senior Contributor Private Message Location: USA-Eastern Time Zone Joined: Sep 16, 2014 16:44 Messages: 629 Offline
[Post New]
Ah,.... I see that now. I had thought you had used all 2048 or higher source clips. Clearly in your table 2 is 1920x1080 source clips also.

Do you think this more efficient rendering is due to the GPU getting involved with the processing?

CS PD13 Ultimate - Build 3516, WIN 8.1, 64 Bit, 16G RAM, Intel Core i5 4460, CPU @ 3.2GHz, NVIDIA GeForce GT720, Graphics Memory(total avail.)-4093MB
LG WH14NS40 Blu-Ray Drive
Reply
ynotfish
Senior Contributor Private Message Location: N.S.W. Australia Joined: May 08, 2009 02:06 Messages: 9810 Offline
[Post New]
Hi CS -

About your first question: You used 4k source material for the analysis - would you think there would be similar results with clips that are just 1920x1080 HD sort of stuff? The difference in rendering times is greatest when PDR/CDR are downscaling. If the source clip & target resolution are both 1080p, the difference between CDR & PDR is much less significant, as you would have seen in Jeff's tables & my little example above.

Example - if the source video & target resolution are both 3840x2160, the encode time ratio is about 1.3



You used the same duration clip for both sources in your table correct? Just curious what was the duration of the clip? Both clips in my example were 1 minute in duration. Yes - up/downscaling exaggerates the difference in encoding times. If it was a longer project (say 60-120 minutes), it wouldn't be too difficult to adjust workflow to minimise time wastage.



I'm still testing things here, trying to determine the significance of differences in visual quality of the final produced files, using more extreme colour adjustments. After all, the bottom line is what you're viewing - not how long it took to encode it! laughing

Cheers - Tony
Visit PDtoots. PowerDirector Tutorials, tips, free resources & more. Subscribe!
Full linked Tutorial Catalog
PDtoots happily supports fellow PowerDirector users!
Reply
ynotfish
Senior Contributor Private Message Location: N.S.W. Australia Joined: May 08, 2009 02:06 Messages: 9810 Offline
[Post New]
Following on from Jeff's original investigation, I wondered whether the levels of adjustments made might have some impact on (a) encoding time ratio & (b) visual quality.

In the table below, the source video is UHD 3840x2160 & target resolution is HD 1920x1080. The same procedure was followed comparing adjustment in CDR & production in PDR with adjust/produce in CDR.



My thinking was wrong. Though, of course, it takes longer to render a more heavily adjusted clip, the encode time ratio is fairly constant. Nor have I been able to detect much different in the visual quality of the output files. CDR settings for "*5" and "**10" are shown in the attached pic.

Cheers - Tony
[Thumb - CDR Settings.png]
 Filename
CDR Settings.png
[Disk]
 Description
 Filesize
52 Kbytes
 Downloaded:
88 time(s)

This message was edited 1 time. Last update was at May 06. 2016 04:16


Visit PDtoots. PowerDirector Tutorials, tips, free resources & more. Subscribe!
Full linked Tutorial Catalog
PDtoots happily supports fellow PowerDirector users!
Reply
JL_JL [Avatar]
Senior Contributor Private Message Location: Arizona, USA Joined: Oct 01, 2006 20:01 Messages: 4541 Offline
[Post New]
Quote: Do you think this more efficient rendering is due to the GPU getting involved with the processing?

More than likely, that was mentioned in the PDF as well. While I thought perhaps OpenCL utilization at the time, after a little more effort it looks like it might simply be DirectCompute. This would probably explain the lack of OpenCL icon credentials on the CD About page. MS DirectCompute/Nvidia CUDA/OpenCL/ATI Stream, all just various methods of using the GPU as a general purpose compute device.

Jeff
Reply
CS2014
Senior Contributor Private Message Location: USA-Eastern Time Zone Joined: Sep 16, 2014 16:44 Messages: 629 Offline
[Post New]
Tony, Jeff,

Thanks for the response - Tony I pretty much understand your responses... Jeff.... I do apologize but I'm going to have to do a bit more reading to understand your response - please.. no offense is intended. I am certainly not 'up to snuff' as you two are. But I can gather that CDR is benefiting when reducing from 4k to 1080 is more significant than if just going from 1080 to 1080.

(I hope I am not making a complete fool of myself as I have tried to not portray my understanding as more than what I have communicated in the past).

Hey in my defense... I"m just about to embark ... going off to the 'fishing expedition' that I look forward to each year. This is certainly no excuse but, after a lil wine n Scotch... well......

Will have some connectivity over the next two weeks but hopefully will have more opportunity to 'capture' some footage.

take care guys!

CS

This message was edited 2 times. Last update was at May 07. 2016 21:28

PD13 Ultimate - Build 3516, WIN 8.1, 64 Bit, 16G RAM, Intel Core i5 4460, CPU @ 3.2GHz, NVIDIA GeForce GT720, Graphics Memory(total avail.)-4093MB
LG WH14NS40 Blu-Ray Drive
Reply
[Post New]
That result is consistent with PD being a Hodgepodge of various pieces of software, not always interacting in the most benefic way.
Reply
OffTheMark [Avatar]
Member Private Message Joined: Jun 12, 2016 10:39 Messages: 114 Offline
[Post New]
Quote That result is consistent with PD being a Hodgepodge of various pieces of software, not always interacting in the most benefic way.


So after all is said and done... is there software out there that is more effecient than powerdirector when downsampling footage from 4K to 1080p (with color adjustments / grading)?

Or are other programs even SLOWER than PD / CD???
Reply
ynotfish
Senior Contributor Private Message Location: N.S.W. Australia Joined: May 08, 2009 02:06 Messages: 9810 Offline
[Post New]
Quote So after all is said and done... is there software out there that is more effecient than powerdirector when downsampling footage from 4K to 1080p (with color adjustments / grading)? Or are other programs even SLOWER than PD / CD???


Hi OffTheMark -

I won't try to answer your question because I've done no comparisons... & this thread is 6 months old. Similar tests to those above have been repeated with PDR15 & CDR5 - http://forum.cyberlink.com/forum/posts/list/50679.page#266184

My observation in that test is that both PDR15 & CDR5 have improved performance over previous versions.

Cheers - Tony
Visit PDtoots. PowerDirector Tutorials, tips, free resources & more. Subscribe!
Full linked Tutorial Catalog
PDtoots happily supports fellow PowerDirector users!
Reply
Reply to this topic
Powered by JForum 2.1.8 © JForum Team