DG NV tools

[Guest]

Looks like those earlier drivers don't support the cropping rectangle. So you'll have to specify no cropping and then crop in your script. There's nothing I can do to fix that up.

[Inspector.Gadget]

1920x1080, 29.970 fps, 1798 frames

Edit: OK, I'll test that out and see how it goes. Thanks.

[Varies]

Thanks for update! I will test with tv ts.
I hope that sometime nVidia will correct a decoding engine and the image will be normal ^ ^

[woah!]

so i tried a few things and so far my issues are pointing to these files: nvcuda.dll nvcuvid.dll

i get random crashes with drivers 190.62 so i installed drivers 186.38 which seem to be the last stable ones. ran lots of deint/resize stuff with no crashes.

i then put the 190.62 nvcuda.dll nvcuvid.dll in the system32 folder with 186.38 drivers and got my system to crash like it did with all the 190.62 drivers installed.

back to 186.38 nvcuda.dll nvcuvid.dll and everything is good again.

dont know if this helps at all really but at least i know my system is ok hardware wise....

[Varies]

woah!
uninstall video drivers completly and install 190.62 |reboot, unpack DG NV tools and replace dgdecode.dll in avisynth plugin derictory. make sure that you run latest cuvidserver :) beta 10 work as usual.

[woah!]

ahh i should have posted in this thread http://forum.doom9.org/showthread.php?t=149805

@Varies

i have done all that, many times actually, i even bought a new gcard thinking my one was borked... winxp 32bit sp3, you name the bench test and i have run it, prime95/occt /linpack etc... all stable until i use cuda encoding with deinterlace and resizing options turned on in dgsource() with nvidia drivers after 186.38

i now have 2 different make gcards, one is the older 192 shader 260gtx , and now i have a msi 216 shader 260gtx which does the same crash...

LoadPlugin("D:\NV10\DGDecodeNV.dll")
DGSource("H:\1.dgm" ,deinterlace=2 ,resize_w=1024 ,resize_h=576)



On Tue 9/22/2009 3:43:22 AM your computer crashed
This was likely caused by the following module: nv4_disp.dll
Bugcheck code: 0x100000EA (0x88B42890, 0x88DE39F0, 0xF78C6CBC, 0x1)
Error: THREAD_STUCK_IN_DEVICE_DRIVER_M
file path: H:\WINDOWS\system32\nv4_disp.dll
product: NVIDIA Compatible Windows 2000 Display driver, Version 190.62
company: NVIDIA Corporation
description: NVIDIA Compatible Windows 2000 Display driver, Version 190.62

[Varies]

I think it's attached to your source or pc :)
my pc ->
Processor: Intel(R) Core(TM)2 Quad CPU Q9300 @ 2.50GHz (2745 МГц)
Operation system: Microsoft Windows XP, 32-bit (Service Pack 3)
DirectX: 9.0c
Video processor: GeForce 9600 GT
Driver version: 190.62
x264: x264 r1271 by x264.nl
avs-script ->
DGsource("D:\video\103.dgm",deinterlace=0)
#
tdeintted = TDeint(mode=2,type=3,mtnmode=3,slow=2,AP=128)
tfm(order=-1,clip2=tdeintted, pp=7).tdecimate(mode=1,hybrid=0,nt=1,denoise=true)
#
BlackmanResize(960,720,181, 3, -182, -4,taps=8)
#
DGsource("D:\video\103.dgm",deinterlace=1)
#
tdecimate(mode=1,hybrid=0,nt=1,denoise=true)
#
BlackmanResize(960,720,181, 3, -182, -4,taps=8)
#
DGsource("D:\video\103.dgm",deinterlace=1, resize_w=960 ,resize_h=720) #crop in DGindex
#
tdecimate(mode=1,hybrid=0,nt=1,denoise=true)
#
DGsource("D:\video\103.dgm",deinterlace=2, resize_w=960 ,resize_h=720) #crop in DGindex
#
selecteven()
tdecimate(mode=1,hybrid=0,nt=1,denoise=true)
#

and it's work fine ^ ^

fixed

[deets]

would i really get an invalid license pop up by going from wifi connection to a LAN connection??? only thing thats changed :( can i have a new license id?

[Varies]

would i really get an invalid license pop up by going from wifi connection to a LAN connection???
yeah :D
I have same problem when i start virtual pc with adsl-lan connection.

[deets]

yeah :D
I have same problem when i start virtual pc with adsl-lan connection.

sigh good old "piracy protection" screwing over a legit paying customer again, nice :(

[GZZ]

all of you can just get the machine ID when machine have VPN open, wifi, lan etc and mail them to neuron2, then he can generate a new licens file that will work for you.

[Guest]

If you don't like my policies you don't have to use my tools. I provide additional license IDs immediately when I receive the request by email.

[Guest]

I have refreshed NV beta 10 with a fix for a serious CUVIDServer bug. It could cause these horrible things:

1. Opening a new stream could return as the first frame a frame from the previous stream opened.

2. Random access could produce errored frames.

I think some people learned to avoid this by killing and restarting CUVIDServer for each new stream/pass. This should no longer be necessary.

http://neuron2.net/dgavcdecnv/NVbeta10.zip

[laserfan]

I have refreshed NV beta 10 with a fix for a serious CUVIDServer bug.OMG neuron2, I came here to report a problem, and you already have the fix!

I had found that with the newest vers of CUVIDServer, when I was in Vdub jumping to specific frames (for new chapters) that the previously-viewed frame was being retained i.e. Vdub was not refreshing to the GoTo'ed frame. On a hunch that CUVIDServer was to blame, I reverted-back to the one you published with the original beta9 (file modified Aug-29) before it would work again.

But I am happy to report that the new CUVIDServer you published w/this post has fixed the problem! I wonder how you fixed this before I was able to report it--I haven't seen anyone mention this! No matter, very happy here! :)

[Guest]

I wonder how you fixed this before I was able to report it--I haven't seen anyone mention this! No matter, very happy here! :) Once in a while I pretend to be Joe Six Pack and go through an encode using my tools just to see what I might run into. I ran into it!

The funny thing is that the bug has been there all along. Fortuitous timing may have saved us before. Now with D3D-less operation timings have changed.

[laserfan]

The funny thing is that the bug has been there all along. Fortuitous timing may have saved us before. Now with D3D-less operation timings have changed.I can say that I've had problems in the past seeking (as above) but could never quite put my finger on a repeatable incidence i.e. was always able to fiddle back/forth and get to the right frame(s) eventually. But it seems since the 1st beta9 CUVIDServer got completely broke. I barely know how to spel D3D but am happy if the swarm (of bugs) moves along with it. ;)

Thanks for fixing it so quickly, and saving me having to try to explain what the heck I was seeing! :scared: :)

[woah!]

well so far it seems this new release doesnt crash my system with 190.62 .dll's :) i have tried many of the prior source files that did crash my setup, but no issues and encodes run completely to end.

it looks like i have a 260gtx to SLI which i thought was borked...

well done :)

[Guest]

@woah!

Are you running with D3D=1?

[woah!]

no all indexers are set to D3D=0 as is the server.

[Jeff Flowerday]

191.07 drivers have been released. Anyone running them?

[Guest]

All I get when going to nvidia.com is a gray screen!

But nvidia.com/page/home.html gets me through. I'll test the new driver.

[Jeff Flowerday]

All I get when going to nvidia.com is a gray screen!

But nvidia.com/page/home.html gets me through. I'll test the new driver.

Excellent!

[Guest]

Seems to be working fine for me.

[Jeff Flowerday]

It might be my imagination, but the new drivers seem to give a little boost in performance...

[Fotis_Greece]

Just downloaded v10 and tried to demux some vob files. Unfortunatelly it does not give me choice to select audio, as if it does not exist, not even in the preview it shows no audio. I used dgmpgdec155 and properly demuxed video and audio.
Is there something am I doing wrong?

[Guest]

Is there something am I doing wrong? Yes, you are failing to give me a stream sample that I can use to fix the problem. You didn't even tell me the type of audio so I can tell you if it is supported yet!

[Fotis_Greece]

Oh yes, sorry. It was a DD 5.1 audio as came from the 4 vob files ripped from a DVD. Unfortunatelly I deleted the vob files after successfully demuxed with dgmpgdec155.
These vob files are almost 1gb, how can I send? But i tried many times with DGMPGIndexNV and there was no audio shown in the preview and ofcourse no audio stream available to set for demuxing (it didn't come with the message that there is no audio stream, it was just blank).
Anyway next time.

[Guest]

Oh yes, sorry. It was a DD 5.1 audio as came from the 4 vob files ripped from a DVD. Unfortunatelly I deleted the vob files after successfully demuxed with dgmpgdec155. Re-rip the DVD. Then use DGSplit to split 50MB from the front of the first one.

I'm sure you'd be willing to do that to help improve the software and contribute to the community.

[Fotis_Greece]

OK, just allow me a day or two to re-rent the DVD disc.

[Guest]

OK, just allow me a day or two to re-rent the DVD disc. Don't bother. We can't help you rip rented material. Struck for rule 6.

[Fotis_Greece]

OK, I will try with a non retail, home-made disc.

[squid_80]

191.07 drivers have been released. Anyone running them?
Just an FYI, these drivers fix a bug that was causing corrupt decoding of streams with high amounts of reference frames (when not using D3D).

[Guest]

That's good to know. There was another bug fix also related to improper honoring of quant matrices in some streams.

[Fotis_Greece]

There you are my good friend neuron. This comes from a magazine dvd trailer so no worries.
http://rapidshare.com/files/292158770/VTS_01_1.zip

NV beta10 does not recognise any audio stream.
But the good old dgmpgdec155 works just fine!!
I would really appreciate if you could probably have a look.

[Fotis_Greece]

And something else. This comes from a sat receiver's recording (DGS Cuberevo)
Nvbeta10 does not recognise any video or audio (!!) while dgmpgdec155 works again just fine!
http://rapidshare.com/files/292166458/ant1121009.zip

By the way, why in nvbeta10 in the preview there is no info about field order while in dgmpgdec155 the info preview is more detailed including the field order?

[Guest]

Re-download NV beta 10 and get the new DGMPGIndexNV. (That is the only component I refreshed.) Thank you for pointing out these issues, Fotis_Greece.

By the way, why in nvbeta10 in the preview there is no info about field order while in dgmpgdec155 the info preview is more detailed including the field order? Because it's not finished yet.

[Klipper]

Hi,

I'm was using DGMPGIndexNV lately. My Problem here is that it is slower than with just the CPU.

CPU only: http://nopaste.info/78caa049a4.html
CPU + GPU: http://nopaste.info/5468d41d31.html
Logfile: http://nopaste.info/764a2f8412.html
x264/megui: http://nopaste.info/9442ed9672.html

The system I'm using:
XP 32bit SP3
Nvidia GTX 295
Intel i7 950
megui/x264
avisynth 2.5.8 / MT-Plugin

I get around 12 fps with just the CPU. When I'm using the GPU/NV tools it goes down to 4 fps.

Does someone else has the same problem as me or solutions for it?

[squid_80]

Hardware MPEG2 decoding typically isn't quicker than today's CPUs.

[Guest]

That's a giant slowdown, though, and it's hard to suppose that decoding is a large fraction of the total load. I think something else must be going on here.

[Revgen]

@Clipper

Stop using the MT Plugin when using DGDecodeNV. See what happens.

[Klipper]

This time I tested the DGAVC-Versions on an AVC-FullHD-Source.
Now I get 36 fps with CPU+GPU and just 18 fps with CPU only.

Looks like the DGMPG GPU is slower than the CPU like squid_80 said. I'm glad that the AVC-Version isn't as the MP-Version ;)

[Guest]

I'm still not satisfied with that. Can yo do an experiment for me please?

Load the source into DGMPGDecNV. Set Video/Disable display and then hit F5. Wait for a while and report the FPS from the info display.

[Guest]

Here is beta 11. It adds some fixes for CUDA context creation that may explain some failures. It also contains experimental support for multinstance support. Please read the Readme.txt file carefully if you want to experiment with multiple instance support.

http://neuron2.net/dgavcdecnv/NVbeta11.zip

[Klipper]

I'm still not satisfied with that. Can yo do an experiment for me please?

Load the source into DGMPGDecNV. Set Video/Disable display and then hit F5. Wait for a while and report the FPS from the info display.

Done it, I got ~60 fps with that

[Guest]

So there is no way to attribute the slow-down to the decoding.

[squid_80]

So there is no way to attribute the slow-down to the decoding.
You don't think 60fps is a bit slow for simple MPEG2 decoding?

[GZZ]

You don't think 60fps is a bit slow for simple MPEG2 decoding?

You have to remeber that the Cuda driver can only use part of your video card to decode video. I got a GTX 260 and get simular result on decoding, so it is as fast at it can go with the current driver. But getting 60 fps on 2 pass encoding will also require some serious CPU power.

[kemuri-_9]

But getting 60 fps on 2 pass encoding will also require some serious CPU power.

you're being assumptive.
sure I've got a quad-core processor,
but i can get >300 fps with x264 using dgmpgdec as my source plugin for some 640x480 material.

[Guest]

You don't think 60fps is a bit slow for simple MPEG2 decoding? For HD it's not terrible. Anyway that's not the point. I cannot see how that rate can bottle neck the encode to 4 fps.

[PatlaborForce]

but i can get >300 fps with x264 using dgmpgdec as my source plugin for some 640x480 material.

Which is hardly comparable since they are getting 60fps on HD content. I'd say that going up almost 7 times in resolution and the speed only dropping to 1/5th of what you get is quite good. That seems to scale quite well. If you can get more than 60fps on HD content using the software decoder then you would have a point.