View Full Version : New Power Director 7 patch and CUDA
rica
22nd March 2009, 16:36
Hi,
This time Cyberlink has almost done with the latest patch of PowerDirector7, 2519.
Even i see 100% CPU utilisation, it is clear it is effectively using the GPU.
Mean encoding time is realtime*1.5 and the encoding quality is satisfactory:
http://img5.imageshack.us/img5/1594/pdnew.th.png (http://img5.imageshack.us/my.php?image=pdnew.png)
Dunno what Ati users say? :)
EDIT: just finished the encoding of whole BD movie; re-encoding time was 1.25 times to realtime...wow :)
http://img5.imageshack.us/img5/9297/pdnew02.th.png (http://img5.imageshack.us/my.php?image=pdnew02.png)
_ _ _ __
Vista 32 SP1 E6750 GTX260 182.08 FW
_ _ _ _ __ __
Sagekilla
22nd March 2009, 22:15
Got any pictures of what the quality is like?
rica
22nd March 2009, 22:35
Got any pictures of what the quality is like?
Sure:)
http://img5.imageshack.us/img5/8115/hm01.th.png (http://img5.imageshack.us/my.php?image=hm01.png)
_ _ _ _ _
Dark Shikari
22nd March 2009, 22:54
A picture is completely meaningless for comparing quality if you don't know the bitrate it was encoded at and don't compare it against some known benchmark encoder.
Also, only 1.25x realtime? Are they still completely utterly incapable of getting worthwhile performance from CUDA?
rica
22nd March 2009, 23:08
A picture is completely meaningless for comparing quality if you don't know the bitrate it was encoded at and don't compare it against some known benchmark encoder.
You might see the bitrate on above SCaps :)
Plus i just said "satisfactory" but it was a lie; it is "better than satisfactory"
Here is a sample if you wish to give it a go:
http://www.sendspace.com/file/n4s8wd
Also, only 1.25x realtime? Are they still completely utterly incapable of getting worthwhile performance from CUDA?
Pardon?
1920*1080 with this bitrate?
_ _ _
Dark Shikari
22nd March 2009, 23:13
Pardon?
1920*1080 with this bitrate?
_ _ _Yes, 1.25x realtime is quite slow for that. Considering that x264 can get 1.25x realtime 1080p on a decent CPU without a significant sacrifice in quality settings, I don't see how a CUDA encoder can be competitive unless it gives a very very significant performance boost over that.
rica
22nd March 2009, 23:15
Yes, 1.25x realtime is quite slow for that. Considering that x264 can get 1.25x realtime 1080p on a decent CPU without a significant sacrifice in quality settings, I don't see how a CUDA encoder can be competitive unless it gives a very very significant performance boost over that.
With what preset profile?
Excuse me; dunno any profile like this:
one-pass
VB
10000-15800
Dark Shikari
22nd March 2009, 23:32
With what preset profile?Not any in particular, but I work for a company that does real-time 10 megabit CBR High Profile x264 encoding (Television broadcast) at around ~subme=7 with quad-core CPUs... at 1080i30. Since Blu-rays are 24fps, 30fps is 1.25x that. And 1080i is significantly slower to encode than 1080p.
So 1.25x realtime doesn't seem like that much to me.
LoRd_MuldeR
22nd March 2009, 23:38
Speed alone is completely useless as long as the bitrate and the quality aren't reasonable too.
So unless somebody makes a quality comparison with "1.25x real-time speed" settings we can only speculate. Of course both encoders need to be configured with the same target bitrate.
rica
22nd March 2009, 23:41
Not any in particular, but I work for a company that does real-time 10 megabit CBR High Profile x264 encoding (Television broadcast) at around ~subme=7 with quad-core CPUs... at 1080i30. Since Blu-rays are 24fps, 30fps is 1.25x that. And 1080i is significantly slower to encode than 1080p.
So 1.25x realtime doesn't seem like that much to me.
There is no custom selection in the application as of right now.
If the original is 23.97 BD you have to select 29.97 i.
As a second step i used eac3to to get 23.97 :)
rica
22nd March 2009, 23:43
Speed alone is completely useless as long as the bitrate and the quality aren't reasonable too.
So unless somebody makes a quality comparison with "1.25x real-time speed" settings we can only speculate. Of course both encoders need to be configured with the same target bitrate.
This is what i was trying to say. :)
Additionally under the same circumstances; same CPU, same GPU, same OS.
Right?
rica
23rd March 2009, 00:27
Really, i've been wondering any comment on the sample file?
LoRd_MuldeR
23rd March 2009, 00:34
Additionally under the same circumstances; same CPU, same GPU, same OS.
Right?
Not necessarily. Using a system with a high-end multi-core CPU (or even several CPU's) and a low-end GPU would be biased towards x264.
And using a system with a high-end GPU of the latest generation and a medium (or low-end) CPU would be biased towards the CUDA encoder.
Therefore both encoders should be measured under their "preferred" circumstances. But still some "realistic" setup should be used...
rica
23rd March 2009, 00:43
But still some "realistic" setup should be used...
High-end CPU plus High-end GPU together would be more realistic setup then?
Prerequirement must have been the same profile/bitrate settings?
LoRd_MuldeR
23rd March 2009, 00:58
High-end CPU plus High-end GPU together would be more realistic setup then?
Prerequirement must have been the same profile/bitrate settings?
Again not necessarily. Somebody with a relatively weak CPU may tend to "upgrade" his system with a high-end GPU. Such people would probably benefit most from a CUDA enabled encoder. At the same time somebody targeting for pure software encoding would probably get the best CPU he can afford (remember that 16 core Opteron server?), but may very well use the "onboard" GPU - these are sufficient for HD playback nowadays. So I think the encoders should be measured on the type of system where they would benefit the user most...
rica
23rd March 2009, 01:06
Again not necessarily. Somebody with a relatively weak CPU may tend to "upgrade" his system with a high-end GPU. Such people would probably benefit most from a CUDA enabled encoder. At the same time somebody targeting for pure software encoding would probably get the best CPU he can afford (remember that 16 core Opteron server?), but may very well use the "onboard" GPU. So I think the encoders should be measured on the type of system where they would benefit the user most...
This is completely right; i got it.
But what i am saying how to compare them under the same circumstances?
Even me have never believed GPU based encoders would do the job as much as SW encoders(remember Badaboom). But anymore, i'm not sure so much.:confused:
deekey777
23rd March 2009, 02:10
Not any in particular, but I work for a company that does real-time 10 megabit CBR High Profile x264 encoding (Television broadcast) at around ~subme=7 with quad-core CPUs... at 1080i30. Since Blu-rays are 24fps, 30fps is 1.25x that. And 1080i is significantly slower to encode than 1080p.
So 1.25x realtime doesn't seem like that much to me.
What speed do you expect for GPU encoding?
There is an article in the german magazine c't about "Supercomputers at home". And a part of this article is video-transcoding on GPU (PowerDirector 7 (both IHVs), Badaboom). And there are some words of you in this article about problems of GPU based video-transcoding (parallism, integer performance etc).
Nvidia says, that the current version of the Badaboom uses 64 SPs only, so can we be sure, that their own encoder is better? Or AMD's AVIVO Converter?
LoRd_MuldeR
23rd March 2009, 19:14
There is an article in the german magazine c't about "Supercomputers at home". And a part of this article is video-transcoding on GPU (PowerDirector 7 (both IHVs), Badaboom). And there are some words of you in this article about problems of GPU based video-transcoding (parallism, integer performance etc).
Yeah, I read that article too. And I could remember Dark Shikari's post that was cited in the article. It seems the c't authors are following the x264 discussion on Doom9 :D
Also interesting that they attested "Avidemux" to scale excellent on a multi-core CPU without even mentioning that this is thanks to "x264" ;)
(Anyway, I noticed a mistake in that article: When they talked about different Containers/Formats, they didn't distinguish between "Codec" and "Format" properly. For example they listed "DivX" as one of the Formats that usually are stored in the AVI container. They also often use "MPEG-4" as a synonym for "MPEG-4 Part-2 ASP", which is inaccurate at least...)
Dark Shikari
24th March 2009, 00:36
What speed do you expect for GPU encoding?
There is an article in the german magazine c't about "Supercomputers at home". And a part of this article is video-transcoding on GPU (PowerDirector 7 (both IHVs), Badaboom). And there are some words of you in this article about problems of GPU based video-transcoding (parallism, integer performance etc).
Nvidia says, that the current version of the Badaboom uses 64 SPs only, so can we be sure, that their own encoder is better? Or AMD's AVIVO Converter?From my analysis of this PowerDirector CUDA encoder (from the sample file), it is much better than Badaboom. It does full analysis (all partition modes) and has some basic psy opts of some sort. It's probably nowhere near x264, but I can't guarantee that statement unless I see a real comparison at lower bitrates between it and x264. But it's definitely a huge step above Badaboom.
I don't know what speed to expect from GPU encoders, but I expect that if they are to be useful, they should be significantly faster than CPU encoders (to make up for the fact that they require an expensive add-on card). Of course, there's a speed-quality tradeoff; if the encoder gets relatively good quality, it can justify lower performance.
rica
24th March 2009, 02:24
if the encoder gets relatively good quality, it can justify lower performance.
Seems to me that common belief will be a nostalgia in a very close feature :eek:
Too much optimistic?
Cyber-Mav
30th March 2009, 22:37
so how is this power director panning out to be? worth buying badaboom over it?
CruNcher
31st March 2009, 05:53
Rica the Encoder is definitely tuned for High Bitrate same as Cyberlinks own so AVCHD Scenario just try Parkrun @ 3 Mbit and see a complete different world ;)
and in the last tests i did i couldn't be really sure if the GPU was used @ all , because as you said 100% Cpu utilization not used from Badaboom of that High one :D
Dark Shikari
31st March 2009, 08:08
Rica the Encoder is definitely tuned for High Bitrate"Tuned for high bitrate" is a euphemism for "sucks."
Sagittaire
31st March 2009, 15:07
"Tuned for high bitrate" is a euphemism for "sucks."
Well it's not false ... lol
For me it's "don't work very well at low bitrate"
rica
1st April 2009, 01:59
and in the last tests i did i couldn't be really sure if the GPU was used @ all , because as you said 100% Cpu utilization not used from Badaboom of that High one :D
CruNcher, i have to admit i've never seen any SW based encoder which has 1.25 real-time encoding speed at that bitrate with my poor core2-duo.
CruNcher
4th April 2009, 05:32
It seems Nvidia is changing something :)
there is now a nvcuvenc.dll (NVIDIA CUDA Video Encoder, Version 185.66 ) in the latest Cuda 2.2 driver (185.66 beta) not the nvpvenc.ax ("NVIDIA Pure Video Encoder") So Nvidia renamed it from Pure Video Encoder to Cuda Video Encoder, it still seems to be the same Directshow filter with some more options now :)
I wonder if they finally gonna make it accessible to everyone with the Final Cuda 2.2 SDK like they did with nvcuvid.dll (NVIDIA CUDA Video Decoder) http://forums.nvidia.com/index.php?showtopic=92416 nothing about the change here it seems only for 3rd Party ISVs currently @ all
Nvidia Pure Video Encoder: 1.253.376 Bytes Nvidia Cuda Video Encoder: 1.310.720 Bytes
http://s10.directupload.net/images/090404/ulqbtjo9.png http://s10.directupload.net/images/090404/anqpx9nw.png
i guess quality wise though not much has changed gonna test if it still works with PowerDirector
hackaro70
4th April 2009, 14:02
Hi all,
I tried Power Director v.7 build 2519 myself with an Intel Core 2-Duo E8600 (3.33GHz) and an ATI HD 4850 (1GB ddr5, not OC) as on the Cyberlink site they state that there is an acceleration for ATI graphic video cards too.
A full HD (1080p , 25 fps .m2ts stream from a BD) stream of 1:10 (hour:min) duration has been converted to AVC H.264 in less than 3 hours, which is very good in my experience. For doing the same file by the way of x.264 converter my pc took around 10 hours!
the dummy question is: Power Director 7 produces as AVC264 output an .m2ts stream , how can I convert it to .mp4 or .mkv WITHOUT recoding it another time!?
Thanks to all! :-)
LoRd_MuldeR
4th April 2009, 16:28
A full HD (1080p , 25 fps .m2ts stream from a BD) stream of 1:10 (hour:min) duration has been converted to AVC H.264 in less than 3 hours, which is very good in my experience. For doing the same file by the way of x.264 converter my pc took around 10 hours!
This doesn't say anything. You would have some useful result, if you visually compared x264 to Power Director 7 with settings that give the same speed for both encoders.
Note that x264 can run very fast, if you only configure it accordingly. You will loose some quality, but maybe not enough to loose against Power Director ;)
the dummy question is: Power Director 7 produces as AVC264 output an .m2ts stream , how can I convert it to .mp4 or .mkv WITHOUT recoding it another time!?
eac3to.exe "foobar.m2ts" "foobar.mkv"
STaRGaZeR
4th April 2009, 16:38
A full HD (1080p , 25 fps .m2ts stream from a BD) stream...
Interesting, 1080p25 is not in the BD spec. Are you sure about that?
Interesting, 1080p25 is not in the BD spec. Are you sure about that?
1080i25 (or 1080i50, if you want to use that notation) is in the spec. If such video is progressive, it is effectively 1080p25.
CruNcher
4th April 2009, 18:04
Btw that's also a funny thing Cyberlinks Encoder Produces Progressive Streams and Nvidias Interlaced with Pal 25 selected ;)
hackaro70
4th April 2009, 18:48
This doesn't say anything. You would have some useful result, if you visually compared x264 to Power Director 7 with settings that give the same speed for both encoders.
Note that x264 can run very fast, if you only configure it accordingly. You will loose some quality, but maybe not enough to loose against Power Director ;)
eac3to.exe "foobar.m2ts" "foobar.mkv"
thanks for the hint. I will try immediately.;)
As for the visually comparable x.264 and Power Director 7 files, I used the Blue-ray standalone profile for x.264 (8000 mb/s as AVBR) because I looked for the highest quality, while with Power Director I used an higher bitrate 10000-15000, if I don't remember wrong; anyway you cannot choose it chooses for you. I'm a really a newbie ... I don't know any much more!
hackaro70
4th April 2009, 18:49
Interesting, 1080p25 is not in the BD spec. Are you sure about that?
Yes, I'm pretty sure. In EU countries 25 fps is the standard and all the BD discs are encoded like that! :)
LoRd_MuldeR
4th April 2009, 18:58
As for the visually comparable x.264 and Power Director 7 files, I used the Blue-ray standalone profile for x.264 (8000 mb/s as AVBR) because I looked for the highest quality, while with Power Director I used an higher bitrate 10000-15000, if I don't remember wrong; anyway you cannot choose it chooses for you. I'm a really a newbie ... I don't know any much more!
For a useful comparison you must configure both encoders to the same target bitrate, of course.
hackaro70
4th April 2009, 19:13
eac3to.exe "foobar.m2ts" "foobar.mkv"
it doesn't work :(
here it's the log:
--
eac3to v3.15
command line: eac3to m:\geisha_p2.m2ts m:\geisha_p2.mkv
TS, 1 video track, 1 audio track, 1:09:38, 50i
1: h264/AVC, 1080i50 (16:9)
2: AC3, 2.0 channels, 256kbps, 48khz, 8ms
[v01] Extracting video track number 1...
[v01] Muxing video to Matroska...
[a02] Extracting audio track number 2...
[a02] A remaining delay of +8ms could not be fixed.
[a02] Creating file "m:\geisha_p2 - 2 - AC3, 2.0 channels, 256kbps, 48khz.ac3"...
Added fps value to MKV header.
Video track 1 contains 104451 frames.
eac3to processing took 6 minutes, 32 seconds.
Done.
----
but then I cannot open the MKV, Nero tells me "error in the file" and exit. Same with WMP. :confused:
And now what!? :-) gee ... it's so damn difficult to convert a video file!?!?!?
LoRd_MuldeR
4th April 2009, 19:21
The re-muxing was successful, it seems. So throw away Nero/WMP and get a player that actually works instead of annoying the user!
I suggest MPlayer -or- VLC Player -or- MediaPlayerClassic + Haali's Media Splitter + ffdshow ;)
rica
4th April 2009, 19:21
Can you upload a cut?
eac3to m:\geisha_p2.m2ts m:\geisha_p2.mkv -50MB
STaRGaZeR
4th April 2009, 19:30
Yes, I'm pretty sure. In EU countries 25 fps is the standard and all the BD discs are encoded like that! :)
Nope, they aren't. Almost all of them are 1080p24. In your own log is the answer:
TS, 1 video track, 1 audio track, 1:09:38, 50i
1: h264/AVC, 1080i50 (16:9)
2: AC3, 2.0 channels, 256kbps, 48khz, 8ms
1080i50 is not 1080p25. 1080p25 is not BD compliant ;)
Can you upload a sample?
hackaro70
4th April 2009, 19:34
Can you upload a cut?
sure I can ... ;)
here (http://www.aquaedi.net/ANTONIO/geisha.zip)
hackaro70
4th April 2009, 19:45
Nope, they aren't. Almost all of them are 1080p24. In your own log is the answer:
TS, 1 video track, 1 audio track, 1:09:38, 50i
1: h264/AVC, 1080i50 (16:9)
2: AC3, 2.0 channels, 256kbps, 48khz, 8ms
1080i50 is not 1080p25. 1080p25 is not BD compliant ;)
Can you upload a sample?
if you mean a sample of the Power Director 7 .m2ts just follow the link in the my previous post. Or do you mean of the original .m2ts stream?
hackaro70
4th April 2009, 19:47
.. of questions for you all:
how can I mux more than one language audio track in the same video stream?
how can I have the chapters as in the original movie? Does MKV support them? I really don't understand the differences between MP4 and MKV. Are they just containers or there is something more!?
STaRGaZeR
4th April 2009, 19:54
if you mean a sample of the Power Director 7 .m2ts just follow the link in the my previous post. Or do you mean of the original .m2ts stream?
The original, please.
LoRd_MuldeR
4th April 2009, 19:58
sure I can ... ;)
here (http://www.aquaedi.net/ANTONIO/geisha.zip)
Plays just fine in both, MPlayer and Media Player Classic (Haali Media Splitter + CoreAVC) ;)
hackaro70
4th April 2009, 20:17
The original, please.
ok, here (http://www.aquaedi.net/ANTONIO/00000.zip)you are.
Please respond to my previous post's questions, if you have time, TIA ;)
STaRGaZeR
5th April 2009, 00:58
ok, here (http://www.aquaedi.net/ANTONIO/00000.zip)you are.
Please respond to my previous post's questions, if you have time, TIA ;)
Thanks, that sample is interesting. To bypass the no 25fps progressive content support in the Blu-ray spec they've speedup the original 24fps to 25fps and then flag it as interlaced despite being progressive.
About your questions:
Matroska is a container. You can mux video streams, audio streams and subtitle streams, as many as you like (there is a limit but you will never reach it probably). You can also include chapters or even fonts for your styled subtitles, all in the same .mkv file. Use mkvmerge for all this stuff. BTW if your goal is to backup your Blu-ray's, you can just demux them and mux into Matroska. The process is lossless.
rica
5th April 2009, 01:05
To bypass the no 25fps progressive content support in the Blu-ray spec they've speedup the original 24fps to 25fps and then flag it as interlaced despite being progressive.
Correct.
This is a speedup to 25; but really it is a 25p.
STaRGaZeR
5th April 2009, 01:08
Correct.
This is a speedup to 25; but really it is a 25p.
It is at 25i, even if it's progressive material. Check the flags ;)
rica
5th April 2009, 01:23
It is at 25i, even if it's progressive material. Check the flags ;)
There is flag but it is progressive; fake interlaced.
Check with arcsoft video decoder and select "deinterlace" as "none"; do you see any interlaced artifacts?
Thanks, that sample is interesting. To bypass the no 25fps progressive content support in the Blu-ray spec they've speedup the original 24fps to 25fps and then flag it as interlaced despite being progressive.
This is just a continuation of the 50-year tradition of 2:2 pulldown. There seem to be a few cases where European Blu-ray studios have done this to a 24p source for some reason: http://www.avsforum.com/avs-vb/showthread.php?t=1117941 (some movies in that list have been originally shot as 25i/p, others as 24p).
STaRGaZeR
5th April 2009, 14:15
There is flag but it is progressive; fake interlaced.
Check with arcsoft video decoder and select "deinterlace" as "none"; do you see any interlaced artifacts?
What artifacts, if it's already progressive?
It doesn't matter how you put it, if the stream has interlaced flags any good decoder or encoder should honor those flags, unless it has some power option to override that for people like us. You can't say it's progressive when it has interlace flags. Just take a look at the Power Director output, 1080i50. It has honored the flags like it should be.
This is just a continuation of the 50-year tradition of 2:2 pulldown. There seem to be a few cases where European Blu-ray studios have done this to a 24p source for some reason: http://www.avsforum.com/avs-vb/showthread.php?t=1117941 (some movies in that list have been originally shot as 25i/p, others as 24p).
I wonder, because the whole point of not allowing 1080p25 was to forget about PAL speedups.
vBulletin® v3.8.11, Copyright ©2000-2025, vBulletin Solutions Inc.