BD-Rebuilder W/DGDecNV enabled comparisons

[jdobbs]

How do you setup bd rebuilder to use DGDecNV? Just install DGDecNV and remember where you installed it. Then within BD-RB, select DGDecNV from the SETUP dialog. BD-RB will start a browser dialog to navigate to the installation location for the two required references (the DGIndexNV.EXE and the DGDecodeNV.dll).

[Smoker25]

How do you enable 2-way with DGDecNV?


Edit: Nevermind, I realized that you need to enable the hidden Multiprocess option.

[Seajunk]

Its quick...

I7-860, GTX 670, Raid 10.

DGDecNV 45 min
[11/02/12] BD Rebuilder v0.42.03 (beta)
[11:32:27] Source:
- Input BD size: 28.31 GB
- Approximate total content: [01:43:00.799]
- Target BD size: 22.95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Auto Quality: Good (Very Fast), ABR
- Decoding/Frame serving: DGDecNV [4-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[11:32:27] PHASE ONE, Encoding
- [11:32:27] Processing: VID_50201 (1 of 1)
- [11:32:27] Extracting A/V streams [VID_50201]
- [11:38:11] Reencoding video [VID_50201]
- Source Video: VC-1, 1920x1080
- Rate/Length: 23.976fps, 148,191 frames
- Bitrate: 27,957 Kbs
- [11:38:11] Reencoding: VID_50201, Pass 1 of 1
- [12:11:18] Video Encode complete
- [12:11:18] Processing audio tracks
- Track 4352 (eng): Keeping original audio
[12:11:18]PHASE ONE complete
[12:11:18]PHASE TWO - Rebuild Started
- [12:11:18] Rebuilding BD file Structure
[12:18:33] - Encode and Rebuild complete
[12:18:33] JOB:

X264/Lavf 71 min
[11/02/12] BD Rebuilder v0.42.03 (beta)
[12:20:49] Source:
- Input BD size: 28.31 GB
- Approximate total content: [01:43:00.799]
- Target BD size: 22.95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Auto Quality: Good (Very Fast), ABR
- Decoding/Frame serving: X264/LAVF [4-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[12:20:49] PHASE ONE, Encoding
- [12:20:49] Processing: VID_50201 (1 of 1)
- [12:20:49] Extracting A/V streams [VID_50201]
- [12:26:37] Reencoding video [VID_50201]
- Source Video: VC-1, 1920x1080
- Rate/Length: 23.976fps, 148,191 frames
- Bitrate: 27,957 Kbs
- [12:26:37] Reencoding: VID_50201, Pass 1 of 1
- [13:26:21] Video Encode complete
- [13:26:22] Processing audio tracks
- Track 4352 (eng): Keeping original audio
[13:26:22]PHASE ONE complete
[13:26:22]PHASE TWO - Rebuild Started
- [13:26:22] Rebuilding BD file Structure
[13:32:42] - Encode and Rebuild complete
[13:32:42] JOB: finished

DirectShow 68 min
[11/02/12] BD Rebuilder v0.42.03 (beta)
[14:33:37] Source:
- Input BD size: 28.31 GB
- Approximate total content: [01:43:00.799]
- Target BD size: 22.95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Auto Quality: Good (Very Fast), ABR
- Decoding/Frame serving: DirectShow [4-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[14:33:37] PHASE ONE, Encoding
- [14:33:37] Processing: VID_50201 (1 of 1)
- [14:33:37] Extracting A/V streams [VID_50201]
- [14:39:32] Reencoding video [VID_50201]
- Source Video: VC-1, 1920x1080
- Rate/Length: 23.976fps, 148,191 frames
- Bitrate: 27,957 Kbs
- [14:39:32] Reencoding: VID_50201, Pass 1 of 1
- [15:34:50] Video Encode complete
- [15:34:50] Processing audio tracks
- Track 4352 (eng): Keeping original audio
[15:34:50]PHASE ONE complete
[15:34:50]PHASE TWO - Rebuild Started
- [15:34:50] Rebuilding BD file Structure
[15:41:39] - Encode and Rebuild complete
[15:41:39] JOB: finished.

[jdobbs]

Interesting. DGDecNV is clearly the best -- but all the times look pretty good. Are your processors pegged at 100%? Have you tried it with 2-way or 6-way processing instead of 4, just as a test?

[soneca]

Here practically no differences between 2, 4 and 6 way.
Expecting a bigger difference...

DGDecNV [2-way]
~26min

- Input BD size: 26,89 GB
- Approximate total content: [01:42:22.302]
- Target BD size: 22,95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Quality: Good (Very Fast), ABR
- Decoding/Frame serving: DGDecNV [2-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[11:41:52] PHASE ONE, Encoding
- [11:41:52] Processing: VID_00000 (1 of 1)
- [11:41:52] Extracting A/V streams [VID_00000]
- [11:47:18] Reencoding video [VID_00000]
- Source Video: MPEG-4 (AVC), 1920x1080
- Rate/Length: 23,976fps, 147.268 frames
- Bitrate: 28.168 Kbs
- [11:47:18] Reencoding: VID_00000, Pass 1 of 1
- [12:07:20] Video Encode complete

DGDecNV [4-way]
~25min

- Input BD size: 26,89 GB
- Approximate total content: [01:42:22.302]
- Target BD size: 22,95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Quality: Good (Very Fast), ABR
- Decoding/Frame serving: DGDecNV [4-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[12:08:04] PHASE ONE, Encoding
- [12:08:04] Processing: VID_00000 (1 of 1)
- [12:08:04] Extracting A/V streams [VID_00000]
- [12:13:28] Reencoding video [VID_00000]
- Source Video: MPEG-4 (AVC), 1920x1080
- Rate/Length: 23,976fps, 147.268 frames
- Bitrate: 28.168 Kbs
- [12:13:28] Reencoding: VID_00000, Pass 1 of 1
- [12:33:27] Video Encode complete

DGDecNV [6-way]
~25min

- Input BD size: 26,89 GB
- Approximate total content: [01:42:22.302]
- Target BD size: 22,95 GB
- Windows Version: 6.1 [7601]
- MOVIE-ONLY mode enabled
- Quality: Good (Very Fast), ABR
- X264 Tweak(s) enabled
- Decoding/Frame serving: DGDecNV [6-way]
- Audio Settings: AC3=0 DTS=0 HD=0 Kbs=640
[10:46:55] PHASE ONE, Encoding
- [10:46:55] Processing: VID_00000 (1 of 1)
- [10:46:55] Extracting A/V streams [VID_00000]
- [10:52:20] Reencoding video [VID_00000]
- Source Video: MPEG-4 (AVC), 1920x1080
- Rate/Length: 23,976fps, 147.268 frames
- Bitrate: 28.168 Kbs
- [10:52:20] Reencoding: VID_00000, Pass 1 of 1
- [11:12:22] Video Encode complete

[jdobbs]

Thanks for the test. The only time you would expect to see a difference is when your processors are running at less than 100% at a lower setting. So your results probably just mean that your processors are being used at somewhere near 100% even when using 2-way.

[soneca]

So your results probably just mean that your processors are being used at somewhere near 100% even when using 2-way.

That's right, processors are used to 100% during almost the entire encode using 2 way.
In my case it's worth not more than 2 way because performance does not increase, but the memory consumption increases progressively.

[jdobbs]

That's right, processors are used to 100% during almost the entire encode using 2 way.
In my case it's worth not more than 2 way because performance does not increase, but the memory consumption increases progressively.Exactly.

[Seajunk]

Out of interest i ran a high quality two pass on titanic last night which took 7 hours with x264/lavf.
Im running it with DGDecNV this morning and the eta is looking like the same. I guess when you increase the 'quality' the bottleneck shifts to the cpu making DGDecNV a bit of a wash.
Anyway, im happy with the 'Good ABR one pass' to a BD-25 target for most things.

[Mark_Venture]

Sorry to resurrect this old thread, but I'm looking at potential ways to speed up BD-Rebuilder times before I can do a full PC upgrade.

From this thread, it appears that using DGDecNV will provide a boost.

All things being equal (and my PC Specs are posted in here -> http://forum.doom9.org/showpost.php?p=1626004&postcount=74), if I upgrade my video card from the GTX275 to a GTX 660, and start using DGDecNV, will I see more of a speed increase than just using DGDecNV with my current rig?

I guess I'm asking will it be a worth while upgrade to a GTX660 & DGDecNV now over just adding DGDecNV to my current rig?

[gonca]

Your GTX275 is at the bottom end of DGDecNV support. I would recommend the 660 with DGDecNV. Check your power requirements to make sure your PSU is up to the job.

[omegaman7]

I more or less gave up on DGDecNV. I keep it installed however, for testing purposes. I have a GTX 570, paired with a 6 core AMD 1090t. My gains were quite minimal, dependent on settings. I would like a more state of the art GPU/VPU, before I begin playing with it again. Fact of the matter is, I was noticing a fairly substantial drop in quality, using DGDecNV. It seemed to oversimplify the color space. "Bad Teacher" comes to mind.

[RobertM]

I tried DGDecNV also, but I don't use it anymore. It didn't save much time for me and there was a conflict with my screen saver preferences. But, more importantly, it was just another variable that could potentially mess up the overall process. Same reason why I don't bother with the multiprocess setting either. If either of them saved a LOT of time then I might consider them -- but I'm not really in that much of a rush. I prefer a dependable result rather than the fastest speed.

With the power and speed of Blusky's system I really think he's in the realm of diminishing returns here, trying to wring the last handful of seconds out of his already awesome encode times. I'm not surprised that he wants to play around and see how fast it will go; I would most assuredly do the same thing -- for a while. ;)

[Mark_Venture]

SO then if it was your computer, you'd say its not worth just upgrading the video card and running with DGDecNV for now, until I'm ready to do the full system upgrade?

[omegaman7]

I'm afraid I've been out of the loop too long. My opinion would be nearly worthless. However, if it were me... I love playing with new hardware. So I'd probably do it :p Of course I would never go with a GPU for ONLY the video processing capability. I game ever now and again, so that is a more important factor. I could see building an encoding machine down the road though. Money is tight at the moment though :(

[Guest]

Fact of the matter is, I was noticing a fairly substantial drop in quality, using DGDecNV. It seemed to oversimplify the color space. "Bad Teacher" comes to mind. That is absurd, and anyone that knows anything about AVC video decoding understands why.

My opinion would be nearly worthless. I agree with that.

[omegaman7]

Don't you think that's a bit harsh?

This was only my testing. 'I' was noticing a drop in quality on the mentioned movie. And as another person said, it's another variable that could mess up a long process. I have been out of the loop for a while though. And it is possible I overlooked something. Decoding for playbook for instance is another factor. Perhaps that was the problem all along. But I find it interesting, that the Raw uncompressed played fine, and the resulting job did not. Highest/slowest quality setting by the way. Perhaps I should have posted my results :S

[Groucho2004]

Your GTX275 is at the bottom end of DGDecNV support. I would recommend the 660 with DGDecNV.
Where did you read this nonsense? A GTX275 has a VP4 chip which is 2 versions past the lowest supported VP2 (8300GS, for example). The latest is VP5.
Also, why would you recommend a 660 (~250$) for DGDecNV when a 610 (~40$) for example is just as fast for this purpose?

[omegaman7]

Where did you read this nonsense? A GTX275 has a VP4 chip which is 2 versions past the lowest supported VP2 (8300GS, for example). The latest is VP5.
Also, why would you recommend a 660 (~250$) for DGDecNV when a 610 (~40$) for example is just as fast for this purpose?

Perhaps he's a gamer like myself. Not sure of the 660's gaming capabilities though :p

[Groucho2004]

And as another person said, it's another variable that could mess up a long process.
You think the crappy DirectShowSource() is not a variable in the process?

[omegaman7]

You think the crappy DirectShowSource() is not a variable in the process?

My point, was that DGDecNV came last :p It's the newest variable.

I never use directshowsource...

[Groucho2004]

I never use directshowsource...What source filter do you use?

[gonca]

@Groucho2004
http://en.wikipedia.org/wiki/PureVideo#Table_of_PureVideo_.28HD.29_GPUs. You get there from neuron2's site. According to this the GTX275 has VP2.
I didn't recommend a 660. The poster asked about a 660 vs a 275.

[Groucho2004]

According to this the GTX275 has VP2.
Oops, my bad.

[omegaman7]

What source filter do you use?

Generally X264/LAVF. But given today's circumstances, I may be reemploying DGDecNV :p On some jobs, it did cut an hour off the process time. So it's worth looking into.

[gonca]

@Grouch2004
No harm, no foul.
VP4 > VP5 is still a big jump in speed.

@Omegaman7
That quote actually belongs to Groucho2004

[omegaman7]

@Grouch2004
No harm, no foul.
VP4 > VP5 is still a big jump in speed.

@Omegaman7
That quote actually belongs to Groucho2004

Oops.

Ha ha, I say that all the time LOL!

[Guest]

Definitely you want a VP5 or better. It is twice as fast as a VP4.

But it's not all about speed. It's about a reliable frame accurate framework for all the video types with proper pulldown and interlace handling treated in a consistent way. If you only do linear encodes, then maybe other source filters would be useful too. I try to make DGDecNV a sort of compliance standard for proper decoding and display when serving video via Avisynth.

[gonca]

I am not an expert at this. I just know that with DGDecNV I get good results and I speed up the process.

[RobertM]

Actually, now that I think about it, perhaps I discounted DGDecNV too quickly after my initial trials.

My initial trouble was that the encoding would stop when the screensaver turned off the monitor. So I had to remember to disable the screensaver for each job. Is this a normal thing, or something wrong with my setup related to DGDecNV?

[Guest]

The screensaver unloads the video driver. I can't control that. Find a screensaver that doesn't do that or don't use one.

[jdobbs]

Sorry to resurrect this old thread, but I'm looking at potential ways to speed up BD-Rebuilder times before I can do a full PC upgrade.

From this thread, it appears that using DGDecNV will provide a boost.

All things being equal (and my PC Specs are posted in here -> http://forum.doom9.org/showpost.php?p=1626004&postcount=74), if I upgrade my video card from the GTX275 to a GTX 660, and start using DGDecNV, will I see more of a speed increase than just using DGDecNV with my current rig?

I guess I'm asking will it be a worth while upgrade to a GTX660 & DGDecNV now over just adding DGDecNV to my current rig?It depends on the system. In most cases it gives a slight speed increase, others not so much. It also makes MULTIPROCESS work better because it is frame accurate and eliminates the splitting function.

[Mark_Venture]

Also, why would you recommend a 660 (~250$) for DGDecNV when a 610 (~40$) for example is just as fast for this purpose?
Yeah I had asked specifically about the 660... while I don't game often, its nice to have a slightly better card just in case. :D

[Mark_Venture]

I am not an expert at this. I just know that with DGDecNV I get good results and I speed up the process.
can you expand a little on what "speed up" you've seen?

ANy idea what percentage it sped things up, or how much time saved? Are your PC specs listed anywhere?

[omegaman7]

I sincerely apologize Neuron2. I shall be very thorough in the following attempt to use DGDecNV, on "The Smurfs" to AVCHD backup. Not only will I compare Players(Decoding) on my main PC, I will also burn the result to DVD-9 and play it back on the living room player.

I didn't mean to insult you or the software. It was merely my personal experience with it. This particular title, also allows me to view a Cinavia infected audio track more closely as well.

[Seajunk]

can you expand a little on what "speed up" you've seen?

On my system (http://forum.doom9.org/showthread.php?p=1598875#post1598875)i was going from ~75 minute bd-25 (very fast) directshow to 45 minutes with DGDecNV.

The higher the quality the less of a difference DGDecNV seemed to make as the workflow became cpu bound.

I no longer use DGDecNV as I switched the processing to a headless server machine and access it via remote desktop.

[Guest]

I shall be very thorough in the following attempt to use DGDecNV, on "The Smurfs" to AVCHD backup. Not only will I compare Players(Decoding) on my main PC, I will also burn the result to DVD-9 and play it back on the living room player. Remember, however, that DGDecNV is only a decoder. If you see some issues at the end of a transcode chain, i.e., when playing back, it could be and likely is due to other tools and processing choices, the encoding, muxing, etc.

I didn't think agreeing with you was rude, but anyway I accept and return your sentiments. Good luck with your processing, and if there is anything I can do to help, please ask here.

[blusky]

DGDecNV uses the PureVideo engine, not the GPU cores, for decoding.

Unless you are transcoding or doing something else, CPU use should not be so high.

Make sure you have decode modes in the INI as 0,0,0.

Can anyone be kind enough to explain what this decode mode does??

Thanks in advance.

[Guest]

You could try reading the documentation.

[blusky]

You could try reading the documentation.

Sorry, Missed it. Will do. Thanks.

[gonca]

@Mark_Venture
See Hardware considerations --- posts 105 & 107. Note the cpu usage, the directshow was 100% on pass 2 and pretty close to that on pass 1. I could have pushed the cpu harder, higher MP number, on the DGDecNV encode

Correction In pass 1 of the DGDecNV encode I was getting 135 FPS which is about the limit for VP5. I might not have been able to shave much more time with a higher MP number.

[omegaman7]

The higher the quality the less of a difference DGDecNV seemed to make as the workflow became cpu bound.

My tests seem to conclude the same. In the past, and today.

When I run High quality, it SEEMS as though the GPU may have to wait for the CPU to finish, before delivering the next frame/s. Because X264/LAVF frame server, beat DGDecNV by over 6 min. However, I've used the FASTER setting in the past, and got encouraging results, when comparing the two frame servers. In which DGDecNV shines. So the real question a person has to ask themselves, is: Does HIGH QUALITY really make much difference? I seem to recall a scene in "The Other Guys", that I had to run slowest encode, to correct an artifacting issue. But I suppose that could have been a Decoding/player issue. I don't like to burn discs, until I'm reasonably sure things are squared away ;) But Jdobbs did elect to choose "High Quality" as default :S

I really need a VP5 chip :( Wish I could justify the cost right now.
--------------------------
The Smurfs (movie only - High Quality setting - Default)

DGDecNV AVCHD (DVD-9)
11:03:05
14:17:39
3hrs14min34sec

X264/LAVF AVCHD (DVD-9)
16:16:18
19:24:29
3hrs8min11sec

[omegaman7]

Yeah, I'm feeling like this testing is pointless(til I have a VP5 chip anyway). I have outdated hardware, and I'm just not seeing a benefit on my machine. Good news is, the video quality of both outputs, are highly similar. So similar, I can't really tell the difference. As well I shouldn't ;)

The Smurfs (movie only - Faster setting)

DGDecNV AVCHD (DVD-9)

00:27:24
02:45:58
2hrs18min34sec

The Smurfs (movie only - Faster setting)

X264/LAVF AVCHD (DVD-9)

16:09:17
18:27:32
2hrs18min15sec

[soneca]

omegaman7

Another test using slower settings and VP5.
The gain DGDecNV is mainly on the first pass. Here's what makes the big difference in speed.

FLYBOYS - DVD-5 - LAVF - [2-way]

- Approximate total content: [02:18:42.147]
- RESIZE 1080p to 720p enabled
- Quality: Ultra High Quality (Extremely Slow), Two Pass
- [10:17:23] Reencoding: VID_00000, Pass 1 of 2
- [10:56:20] Reencoding: VID_00000, Pass 2 of 2
- [12:56:58] Video Encode complete

Pass 1 - 39:03
Pass 2 - 02:00:38
Total - 02:39:41

FLYBOYS - DVD-5 - DGDecNV - [2-way]

- Approximate total content: [02:18:42.147]
- RESIZE 1080p to 720p enabled
- Quality: Ultra High Quality (Extremely Slow), Two Pass
- [13:41:13] Reencoding: VID_00000, Pass 1 of 2
- [14:05:03] Reencoding: VID_00000, Pass 2 of 2
- [15:56:10] Video Encode complete

Pass 1 - 22:50
Pass 2 - 1:51:07
Total - 02:13:57

[omegaman7]

omegaman7

Another test using slower settings and VP5.
The gain DGDecNV is mainly on the first pass.

Mmm hmm. I've noticed that. I don't think I'll do any more testing, til I have a VP5 engine ;) I may sell my GTX 570, and revert to my GTX 260. As I discovered recently, the meltdown didn't harm my 260 :D I think the Power supply simply shutdown, to protect the components. (when my liquid cooling failed)

[soneca]

I never liked water cooling...:p
the difference between a good air cooler and water is not so big.

[omegaman7]

I never liked water cooling...:p
the difference between a good air cooler and water is not so big.

It was quite good while it lasted. Though it didn't help my Northbridge any :( The cooler in my signature is almost as good. I went for a VERY large heat sink this time! Plus it's triple push pull :) I have a controller on them too. It'll do.

[jdobbs]

I never liked water cooling...:p
the difference between a good air cooler and water is not so big.It's huge on my system. I see a good 10C degree difference between my best air cooler (which cost a lot) and my Corsair water cooling.

[soneca]

Jdobbs, 10 degrees really is a big difference!
Corsair would be the H100?
This temperature can vary greatly depending on the level of ventilation cabinet.
Already measured the temperature of the cpu(environment too, delta) running prime?

[Mark_Venture]

... I may sell my GTX 570, and revert to my GTX 260. As I discovered recently, the meltdown didn't harm my 260 :D

OT, but why would you sell the 570 and go back to the 260?