Page 2 of 3

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 1:46 am
by flyordie
neuron2 wrote:His original clip used for his performance results was 7:43. I can't believe CUDA is so close to VPU on the 460 for MPEG2. My testing doesn't agree with that. Maybe flyordie accidentally picked up the same DGDecodeNV.dll when doing the testing?
I did not make a mistake on the DGDecodeNV.dll. I've checked an rechecked. For me, using 1080i interlaced MPEG2 source, my GTX 560 Ti and GTX 460 did not use the Video engine during the testing unless deinterlace=1 or deinterlace=2 was specified. The GT 520, on the other-hand, utilized it's video engine on all of the VPU test avs scrips as shown in my earlier post.

Perhaps you intended for this to happen?
slomorbo wrote:I did the same test using x264.exe as a replacement for AvsMeter,

CODE: SELECT ALL
x264 --preset ultrafast --stats "x264.stats" --thread-input --output test.264 "VP.avs"
Using this method I ran the basic script with the 32-bit x264.exe on my GT 520 and GTX 560 Ti with my 7 minute source I used in my earlier post.
My command line was

Code: Select all

x264 --preset ultrafast --thread-input --output test.264 t1.avs
where t1.avs was substituted with the appropriate script for each version of DGDecodeNV.dll and appropriate .dgi for the given version of DGIndexNV and GPU.

My Intel i7 980x @3.95Ghz would spike early to around 70% utilization but most of each run stayed in the 30s-50s.
CUDA Runs
GT 520 - encoded 13896 frames, 99.75 fps, 11808.97 kb/s
GTX 560 Ti - encoded 13896 frames, 216.51 fps, 11808.97 kb/s
VPU Runs
GT 520 - encoded 13896 frames, 190.45 fps, 11805.91 kb/s
GTX 560 Ti - encoded 13896 frames, 216.86 fps, 11808.97 kb/s

I ran each of these runs multiple times. The VPU GT 520 run had the least amount of variation on FPS. None of the encodes varied on the kb/s. I think its odd all the runs came up with the exact same kb/s except for the VPU GT 520.

Is this because all the encodes used the same internal instructions except when VPU was enable and it saw the engine was VP5?

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 7:54 am
by laserfan
flyordie I'm probably being stoopid now, but I can't seem to sort-out what it means in your posts "CUDA runs" vs "VPU runs"? How exactly are you specifying to the cards and the DGDecNV process to use one or the other?

:?

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 10:50 am
by admin
There are two executables. The released one uses CUDA for MPEG2 and VPU for everything else.

The debug one here:

http://neuron2.net/board/viewtopic.php? ... t=10#p1455

...uses VPU for everything.

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 11:41 am
by slomorbo
I made a little "CUDA vs. VP TESTER" so everyone can test with the same base. It will further more give us consistent benchmark results and avoid involuntary error.
EDIT : I MADE CHANGES SO MAKE SURE TO USE VERSION 2
I REMOVED x264 TESTS
ADD SUPPORT FOR 2 CARDS

You can download it here -> http://www.megaupload.com/?d=V7WY7PEV or http://www.sendspace.com/file/7k6glx
if both link are down please PM me and I will fix it.
ARCHIVE PASSWORD : testingVPCUDA

Install instructions;
MUST BE decompressed directly in C:\ (will give you "C:\TEST_CUDA_VS_VP\")
YOU MUST copy your license.txt file in both DGdecNVxxx folders

There are 2 tests that your can run;

For the first card
First_Card_VP_AVSMETER.bat
Fisrt_Card_CUDA_AVSMETER.bat

if you have two card, for the second card run
Second_Card_VP_AVSMETER.bat
Second_Card_CUDA_AVSMETER.bat

Tested here with a "G210 VP4" and a "GT520 VP5", each CUDA test don't use the VP engine, and each VP test USE VP ENGINE (please confirm using GPU-Z)
As for my results they are exactly the same as in my previous table.

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 12:10 pm
by admin
@slomorbo

Thanks for your test utility.
Groucho2004 wrote:I don't understand why you bring x264 into this. It adds another variable (of which there are already many) and the results will be different for pretty much every processor.
+1

Nvidia reports there is no way currently to query the VP engine version, but it may be implicated by the compute capability. Can a 520 user report the compute capability of the card, please? You can get it from Options/List GPU devices. In the absence of a reliable indicator, I will make CUDA/VPU configurable, per card and per video type.

@all

I'll be offline for several days so don't freak if I don't respond to a post. ;)

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 12:38 pm
by flyordie
At present I have the GTX 460 in another machine. Here is what DGIndexNV shows under Options > List GPUs for my two machines.

Image

Image
neuron2 wrote:@slomorbo

Thanks for your test utility.
Anon17 wrote:I don't understand why you bring x264 into this. It adds another variable (of which there are already many) and the results will be different for pretty much every processor.
+1
I'm happy to take x264 out of the equation. I'll try another MPEG2 clip and see if I can get VPU to activate on the GTX 460 or GTX 560 Ti.
slomorbo wrote:I made a little "CUDA vs. VP TESTER" so everyone can test with the same base. It will further more give us consistent benchmark results and avoid involuntary error.
I will give this a try and see if that changes any of my results.

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 1:24 pm
by slomorbo
@ Anon17 and @ admin
I remove the x264 tests and make a v.2 that add support for a second card.

@ flyordie, please make sure to use v.2 (and completly remove v.1 if you had downloaded it ;) )

@ admin
Image

Re: GT 520 VP5 performance

Posted: Sun Aug 14, 2011 3:51 pm
by flyordie
I just downloaded and used your V2_TEST_CUDA_VS_VP.rar. I added my license.txt file to each dgdecnv folder and double click on each batch file while I had gpu-z 0.5.4 open and with the appropriate gpu selected. Same results :( no Video Engine usage on either my GTX 460 nor GTX 560 Ti.
My GTX 460 is in a machine without avisynth.dll nor devil.dll. This got me thinking... what version of avisynth are you guys running? For these tests I'm using:
I've tried
avisynth.dll File version 2.6.0.0 Size 1.63 MB Date modified 5/3/2011 7:03 PM
avisynth.dll File version 2.6.0.2 Size 1.82 MB Date modified 7/19/2010 11:54 AM
avisynth.dll File version 2.5.8.5 Size 1.36 MB Date modified 7/11/2009 10:46 PM
with
devil.dll File version 0.1.6.5 Size 703 KB Date modified 5/26/2004
devil.dll File version 0.1.7.8 Size 746 KB Date modified 3/8/2009
and
ffavisynth.dll Size 52.5 KB Date modified 12/5/2009 6:42 PM

changing to any of the above .dll's did not fix my issue with the video engine not being used for VP on my GTX 560 Ti.

My results using V2_TEST_CUDA_VS_VP

Code: Select all

Video Card  CUDA     VP
GT 520      95.58   179.14
GTX 260    186.96   185.89  (30% GPU usage 3% Video Engine)
GTX 460    111.88   113.92
GTX 560 Ti 234.33   234.29
I have a space drive around. I'll do a fresh install of windows 7 x64 and see if that makes any difference. :|
Update: Fresh install of Windows 7 x64 and nVidia 280.26 made no difference with the GTX 560 Ti. :(

Here are my GPU-Z Graphics Card info pages. admin, perhaps your GTX 460 is different than mine .
ImageImageImage

Re: GT 520 VP5 performance

Posted: Mon Aug 15, 2011 6:38 am
by laserfan
neuron2 wrote:There are two executables. The released one uses CUDA for MPEG2 and VPU for everything else.

The debug one here:

http://neuron2.net/board/viewtopic.php? ... t=10#p1455

...uses VPU for everything.
Thank you DG for clarifying--the earlier posts somehow whistled right over (or through) my head! :oops:

Re: GT 520 VP5 performance

Posted: Wed Aug 17, 2011 9:14 pm
by Bertie
Not sure how useful this is . . . . .

Using DGIndexNV for a few weeks and stumbled across this thread which interests me as the card I'm using is a Gigabyte GT520OC at stock speeds, as both the display and encoding card.

Unable to download slomorbo test files, therefore in real world encoding I have used both the standard dll and the VPU dll :-

Input file: MPEG TS, MPEG-2 720x576i 25fps 16:9 76500 frames 51min duration
Using [F5] preview with display disabled, the CUDA gives 739.35fps, with 99% GPU/0% Video Engine Load and 20C temperature increase
Using [F5] preview with display disabled, the VPU gives 799.35fps, with 3% GPU/99% Video Engine Load and no temperature increase

Ouput file via MeGui to an mkv - x.264, trim each side by 96 to have 528x576p 25fps forced 4:3 resolution:

CUDA - 1st pass - 226.47fps - ave/max GPU = 64.48/69.00 - ave/max VEL = 0
CUDA - 2nd pass - 71.3fps - ave/max GPU = 27.73/38.00 - ave/max VEL = 0
Total time 23m30s

VPU - 1st pass - 243.91fps - ave/max GPU = 39.00/55.00 - ave/max VEL = 33.12/53.00
VPU - 2nd pass - 73.26fps - ave/max GPU = 11.15/27.00 - ave/max VEL = 9.64/33.00
Total Time 22m37s

NOTE: fps taken from MeGUI queue, video card details from GPU-Z logging file via spreadsheet for the calculations.

For a cheap card the GT520 suits my purpose and with VPU enabled is slightly faster with no obvious visual differences in the resultant file although there is a 3KB difference in file size VPU>CUDA.

Re: GT 520 VP5 performance

Posted: Fri Aug 19, 2011 9:02 pm
by slomorbo
@ Bertie
I fix the download link ;)

Re: GT 520 VP5 performance

Posted: Sat Aug 20, 2011 3:18 am
by Bertie
So using slomorbo's test setup the results are:

CUDA - min/max/avg 74.58/103.46/88.11
VP - min/max/avg 154.83/190.77/178.91

Using the respective engines as confirmed in GPU-Z.

ETA Borrowed a Gigabyte GT440 1GB DDR3 and ran the test:

CUDA - min/max/avg 106.73/142.63/124.86
VP - min/max/avg 85.95/106.40/97.85

Re: GT 520 VP5 performance

Posted: Wed Oct 05, 2011 5:17 pm
by laserfan
neuron2 wrote:Due to a bug in the VP3+ engines for MPEG2, I have to use CUDA instead of the VP engine for MPEG2. I need to check with Nvidia about the status of that for VP5. If it was fixed, I can revert to the video engine if a VP5 is detected.
Got my 520 today, though I've not installed it yet, and am trying to get my head-around your use of the terms "CUDA" and "CUVID". When I search on these terms, it seems they have been used interchangeably, so I'm confused. :oops:

CUVID means decoding using the graphics engine (VPU) exclusively? And CUDA means some Software and some Graphics Engine (cuz you can't use CUDA w/o an Nvidia chipset)?

Right now my 9600GT uses exclusively "cuvid" according to List GPU Devices, but when I install my 520 I'll want to change MPG to CUDA (still)?

Re: GT 520 VP5 performance

Posted: Wed Oct 05, 2011 8:57 pm
by admin
Almost correct.

CUVID = decoding is performed on the dedicated video processor (VP). Note that the VP is not the graphics engine.

CUDA = decoding is performed on a combination of the CPU and the CUDA processors. I don't know the breakdown of tasks between the two.

You can leave everything set for CUVID if you are running latest drivers.

Re: GT 520 VP5 performance

Posted: Thu Oct 06, 2011 5:18 am
by laserfan
neuron2 wrote:Almost correct.

CUVID = decoding is performed on the dedicated video processor (VP). Note that the VP is not the graphics engine.

CUDA = decoding is performed on a combination of the CPU and the CUDA processors. I don't know the breakdown of tasks between the two.

You can leave everything set for CUVID if you are running latest drivers.
>Note that the VP is not the graphics engine.

Thanks alot for correcting me. I will study-up some more on VPU vs Graphics Engine vs CUDA processors--it can be very puzzling to a non-programmer. :?

I think yr post would make a nice addition to the DGIndexNV manual discussion (at the end)! ;)

Oh thanks also for the new "List GPU devices"/INI/selection features which are very nice indeed. :geek:

Re: GT 520 VP5 performance

Posted: Thu Oct 06, 2011 11:36 am
by laserfan
Hey guys, given the decoding fps of this 520GT is 3x that of my previous 9600GT, I am so underwhelmed by the real-world Indexing results that I wonder if I might be doing something wrong. I'd just indexed a VC-1 program with my 9600, so after installing the 520GT I did it several times again, using the simple CLI command:

"C:\Program Files (x86)\DGIndexNV\DGIndexNV" -i D:\video\videoIN.vc1 -o D:\video\videoIN.dgi -e

Here's the log using the 9600GT, noting it took 6:01 to complete:

Code: Select all

Stream Type: Elementary
Video Type: VC1
Profile: Advanced
Level: 3
Coded Size: 1920x1088
DAR: 16x9
Display Size: 1920x1080
Frame Rate: 23.976024 fps
Colorimetry: BT.709 [1]*
Frame Structure: 
Frame Type: 
Coded Number: 142811
Playback Number: 142811
Frame Repeats: 0
Field Repeats: 0
Bitrate: 
Bitrate (Avg): 
Bitrate (Max): 
Elapsed: 0:06:01
Remain: 0:00:00
FPS: 
Info: Finished!
Then after installing the 520 and running with the default Decode_Modes=0,0,0
Elapsed 0:05:52

Next using the 520GT and CUDA (Decode_Modes=1,1,1):
Elapsed 0:05:40

Using 520 and CUDA and High Priority (vs. Normal above):
Elapsed 0:05:47

So CUDA mode was slightly faster than CUVID (I expected it would be slower) but neither is notably better than my previous 9600GT. I've not compared x264 encoding times, but it seems given the bulk of the time is in Encoding anyway that I'm not likely to see an improvement in wall time there either. Was I wrong to expect the 520 would speed-up these processes? Maybe times vary widely between codecs and AVC/MPEG2 will be greatly different? I will try of course.

BTW here's my INI file:

Code: Select all

Version=DGIndexNV 2041
Window_Position=100,100
Info_Window_Position=1066,100
Process_Priority=1
Playback_Speed=3
AVS_Template_Folder=C:\Program Files (x86)\DGIndexNV
AVS_Template_File=template.avs
AVS_Enable_Template=1
AVS_Overwrite=0
Full_Path_In_Files=1
MRUList[0]=D:\video\videoIN.vc1
MRUList[1]=R:\Clash of the Titans - Trailer.track_1.264
MRUList[2]=D:\Clash of the Titans - Trailer.track_1.264
MRUList[3]=E:\TiVo\The War Part 7 - A World Without War.ts
Enable_Info_Log=1
Loop_Playback=0
AVC_Extension=264
MPG_Extension=m2v
VC1_Extension=vc1
Deinterlace=0
UsePF=0
AlwaysCrop=1
UseD3D=0
Snapped=1
AlwaysCancelOnAudioMismatch=0
Enable_Audio_Demux=1
CUDA_Device=0
Decode_Modes=1,1,1
Full_Info=1
Oh, and the List GPU Devices dialog looks like this:
>0:"GeForce GT 520",Compute 2.1, 993 MB, 1SMs

AVC: cuda
MPG: cuda
VC1: cuda

Re: GT 520 VP5 performance

Posted: Thu Oct 06, 2011 6:20 pm
by admin
Indexing does not use the GPU! All it does is scan through the source stream and write an index file. There is no *decoding* done at all when indexing.

You need to measure the decoding FPS rate when serving via your script.

The gain in CPU can be substantial in some scenarios, such as when resizing and deinterlacing on the GPU, or when you have a low end CPU. (But primarily, DGDecNV is about frame accuracy and integrated AVC/MPEG2/VC1 handling. Other source filters have some serious difficulties with M2TS streams, especially field-based ones.)

Re: GT 520 VP5 performance

Posted: Fri Oct 07, 2011 6:48 am
by laserfan
neuron2 wrote:Indexing does not use the GPU! All it does is scan through the source stream and write an index file. There is no *decoding* done at all when indexing.
I didn't know that! :oops:
The gain in CPU can be substantial in some scenarios, such as when resizing and deinterlacing on the GPU, or when you have a low end CPU.
I haven't done any of this yet, so will keep an eye out for same, thanks.
DGDecNV is about frame accuracy and integrated AVC/MPEG2/VC1 handling. Other source filters have some serious difficulties
Man I surely did know that! My encodings have been completely glitch-free ONLY since your release of DGDecNV! :D

Anon17 you guessed pretty much exactly right--my source file was about 18GB. I didn't reencode it from scratch to see if there's a noticeable difference there and maybe I will try that today, but I assume now that a straight re-encode will not see much benefit.

Thanks y'all--at least getting that monster N9600GT out of my PC means better airflow and lower power consumption. It is amazing how tiny the N520GT is by comparison.

Re: GT 520 VP5 performance

Posted: Mon Oct 10, 2011 7:51 am
by laserfan
Well, so much for benchmarking & framerates. Reencoding with x264 there is virtually no improvement with the 520, at least with the VC-1 program I'd done right before swapping-out my 9600 for the 520:

With the 9600GT:

Code: Select all

Wed 10/05/2011
06:01 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 1 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me hex --sub
me 7 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output NUL  --frames 142811 --demuxer y4m --stdin y4m - :
4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile Main, level 4.1
x264 [info]: frame I:3870  Avg QP:17.08  size:133148
x264 [info]: frame P:73607 Avg QP:20.13  size: 35278
x264 [info]: frame B:65334 Avg QP:22.05  size: 11165
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 48.6%  0.0% 51.4%
x264 [info]: mb P  I16..4: 25.1%  0.0%  0.0%  P16..4: 52.3%  0.0%  0.0%  0.0%  0.0%    skip:22.6%
x264 [info]: mb B  I16..4:  3.3%  0.0%  0.0%  B16..8: 21.5%  0.0%  0.0%  direct:16.2%  skip:58.9%  L0:33.3% L1:48.2% BI:18.5%
x264 [info]: final ratefactor: 19.67
x264 [info]: direct mvs  spatial:97.8% temporal:2.2%
x264 [info]: coded y,uvDC,uvAC intra: 46.8% 49.0% 7.0% inter: 16.1% 16.5% 0.1%
x264 [info]: i16 v,h,dc,p: 35% 24% 30% 10%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 22% 21% 20%  3%  9%  6%  6%  6%  7%
x264 [info]: i8c dc,h,v,p: 54% 22% 21%  3%
x264 [info]: Weighted P-Frames: Y:1.1% UV:0.5%
x264 [info]: kb/s:5159.46

encoded 142811 frames, 21.04 fps, 5159.46 kb/s

-----------------
Beginning Pass 2
-----------------

Wed 10/05/2011
07:54 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 2 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me umh --sub
me 9 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output D:\video\videoOUT.264 --frames 142811 --demuxer y4m
 --stdin y4m - : 4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile High, level 4.1
x264 [info]: frame I:3870  Avg QP:20.47  size:100827
x264 [info]: frame P:73607 Avg QP:21.61  size: 36448
x264 [info]: frame B:65334 Avg QP:23.81  size: 12583
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 24.1% 67.6%  8.3%
x264 [info]: mb P  I16..4:  1.6%  4.2%  0.3%  P16..4: 53.3% 12.7%  7.5%  0.0%  0.0%    skip:20.5%
x264 [info]: mb B  I16..4:  0.1%  0.3%  0.0%  B16..8: 42.2%  2.8%  0.5%  direct: 2.2%  skip:52.0%  L0:39.7% L1:53.1% BI: 7.2%
x264 [info]: 8x8 transform intra:68.4% inter:81.5%
x264 [info]: direct mvs  spatial:88.8% temporal:11.2%
x264 [info]: coded y,uvDC,uvAC intra: 70.3% 64.3% 16.7% inter: 22.8% 27.8% 0.2%
x264 [info]: i16 v,h,dc,p: 34% 22% 17% 27%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu:  8%  7% 10% 10% 14% 13% 13% 12% 13%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 11%  9%  4%  9% 14% 14% 14% 11% 15%
x264 [info]: i8c dc,h,v,p: 36% 31% 21% 12%
x264 [info]: Weighted P-Frames: Y:1.2% UV:0.5%
x264 [info]: ref P L0: 70.1% 17.7% 12.1%
x264 [info]: kb/s:5231.52

encoded 142811 frames, 7.24 fps, 5231.52 kb/s

------------------
Encoding Complete
------------------

Thu 10/06/2011
01:23 AM
With the 520GT:

Code: Select all

Sun 10/09/2011
03:55 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 1 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me hex --sub
me 7 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output NUL  --frames 142811 --demuxer y4m --stdin y4m - :
4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile Main, level 4.1
x264 [info]: frame I:3875  Avg QP:17.08  size:133209
x264 [info]: frame P:73610 Avg QP:20.13  size: 35267
x264 [info]: frame B:65326 Avg QP:22.05  size: 11164
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 48.6%  0.0% 51.4%
x264 [info]: mb P  I16..4: 25.1%  0.0%  0.0%  P16..4: 52.3%  0.0%  0.0%  0.0%  0.0%    skip:22.6%
x264 [info]: mb B  I16..4:  3.3%  0.0%  0.0%  B16..8: 21.5%  0.0%  0.0%  direct:16.2%  skip:58.9%  L0:33.3% L1:48.2% BI:18.5%
x264 [info]: final ratefactor: 19.67
x264 [info]: direct mvs  spatial:97.8% temporal:2.2%
x264 [info]: coded y,uvDC,uvAC intra: 46.8% 49.0% 7.0% inter: 16.1% 16.5% 0.1%
x264 [info]: i16 v,h,dc,p: 35% 24% 30% 10%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 22% 21% 20%  3%  9%  6%  6%  6%  7%
x264 [info]: i8c dc,h,v,p: 54% 22% 21%  3%
x264 [info]: Weighted P-Frames: Y:1.1% UV:0.5%
x264 [info]: kb/s:5159.46

encoded 142811 frames, 21.48 fps, 5159.46 kb/s

-----------------
Beginning Pass 2
-----------------

Sun 10/09/2011
05:45 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 2 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me umh --sub
me 9 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output D:\video\videoOUT.264 --frames 142811 --demuxer y4m
 --stdin y4m - : 4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile High, level 4.1
x264 [info]: frame I:3875  Avg QP:20.47  size:100847
x264 [info]: frame P:73610 Avg QP:21.61  size: 36441
x264 [info]: frame B:65326 Avg QP:23.81  size: 12583
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 24.1% 67.6%  8.3%
x264 [info]: mb P  I16..4:  1.6%  4.2%  0.3%  P16..4: 53.3% 12.6%  7.5%  0.0%  0.0%    skip:20.5%
x264 [info]: mb B  I16..4:  0.1%  0.3%  0.0%  B16..8: 42.2%  2.8%  0.5%  direct: 2.2%  skip:52.0%  L0:39.7% L1:53.1% BI: 7.2%
x264 [info]: 8x8 transform intra:68.4% inter:81.5%
x264 [info]: direct mvs  spatial:88.8% temporal:11.2%
x264 [info]: coded y,uvDC,uvAC intra: 70.4% 64.3% 16.7% inter: 22.8% 27.8% 0.2%
x264 [info]: i16 v,h,dc,p: 34% 22% 17% 27%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu:  8%  7% 10% 10% 14% 13% 13% 12% 13%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 11%  9%  4%  9% 14% 14% 14% 11% 15%
x264 [info]: i8c dc,h,v,p: 36% 31% 21% 12%
x264 [info]: Weighted P-Frames: Y:1.2% UV:0.5%
x264 [info]: ref P L0: 70.1% 17.7% 12.1%
x264 [info]: kb/s:5231.59

encoded 142811 frames, 7.19 fps, 5231.59 kb/s

------------------
Encoding Complete
------------------

Sun 10/09/2011
11:17 PM
9600 wall time 7h 22m
520 wall time 7h 22m

My avs was this simple:"DGSource("videoIN.dgi")". Maybe it would do better with AVC or MPEG2. Or maybe as Anon17 suggests, my Q6600 and/or hard drives are the limiting factor, but this is disappointing nonetheless. At least I only paid $20 bucks for the new card!

If there's something wrong with my command-line (or expectation/thought process) by all means straighten me out! :?

BTW I'd also done a CUDA run and it only slightly slower at 7h 28m:

Code: Select all

Sat 10/08/2011
04:24 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 1 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me hex --sub
me 7 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output NUL  --frames 142811 --demuxer y4m --stdin y4m - :
4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile Main, level 4.1
x264 [info]: frame I:3875  Avg QP:17.08  size:133219
x264 [info]: frame P:73610 Avg QP:20.13  size: 35267
x264 [info]: frame B:65326 Avg QP:22.05  size: 11164
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 48.6%  0.0% 51.4%
x264 [info]: mb P  I16..4: 25.1%  0.0%  0.0%  P16..4: 52.3%  0.0%  0.0%  0.0%  0.0%    skip:22.6%
x264 [info]: mb B  I16..4:  3.3%  0.0%  0.0%  B16..8: 21.5%  0.0%  0.0%  direct:16.2%  skip:58.9%  L0:33.3% L1:48.2% BI:18.5%
x264 [info]: final ratefactor: 19.67
x264 [info]: direct mvs  spatial:97.8% temporal:2.2%
x264 [info]: coded y,uvDC,uvAC intra: 46.8% 49.0% 7.1% inter: 16.1% 16.5% 0.1%
x264 [info]: i16 v,h,dc,p: 35% 24% 30% 10%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 22% 21% 20%  3%  9%  6%  6%  6%  7%
x264 [info]: i8c dc,h,v,p: 54% 22% 21%  3%
x264 [info]: Weighted P-Frames: Y:1.1% UV:0.5%
x264 [info]: kb/s:5159.46

encoded 142811 frames, 20.44 fps, 5159.46 kb/s

-----------------
Beginning Pass 2
-----------------

Sat 10/08/2011
06:21 PM

D:\video>"C:\Program Files (x86)\x264\pipebuf.exe" "C:\Program Files (x86)\x264\avs2yuv.exe" D:\video\videoIN.avs - : "C:\Program Files (x86)\
x264\x264_x64.exe" --pass 2 --tune film  --keyint 48 --rc-lookahead 48 --bitrate 5235 --vbv-maxrate 15000 --vbv-bufsize 15000 --min-keyint 1 -
-open-gop --bluray-compat --b-pyramid none --slices 4 --stats D:\video\videoIN.stats --qpfile D:\video\qpfile.txt --direct auto --me umh --sub
me 9 --colorprim bt709 --transfer bt709 --colormatrix bt709 --level 4.1 --sar 1:1 --output D:\video\videoOUT.264 --frames 142811 --demuxer y4m
 --stdin y4m - : 4
D:\video\videoIN.avs: 1920x1080, 24000/1001 fps, 142811 frames
y4m [info]: 1920x1080p 1:1 @ 24000/1001 fps (cfr)
x264 [info]: using SAR=1/1
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
x264 [info]: profile High, level 4.1
x264 [info]: frame I:3875  Avg QP:20.47  size:100861
x264 [info]: frame P:73610 Avg QP:21.61  size: 36439
x264 [info]: frame B:65326 Avg QP:23.80  size: 12583
x264 [info]: consecutive B-frames: 20.4% 53.0%  8.2% 18.4%
x264 [info]: mb I  I16..4: 24.1% 67.6%  8.3%
x264 [info]: mb P  I16..4:  1.6%  4.2%  0.3%  P16..4: 53.3% 12.6%  7.5%  0.0%  0.0%    skip:20.5%
x264 [info]: mb B  I16..4:  0.1%  0.3%  0.0%  B16..8: 42.2%  2.8%  0.5%  direct: 2.2%  skip:52.0%  L0:39.7% L1:53.1% BI: 7.2%
x264 [info]: 8x8 transform intra:68.4% inter:81.5%
x264 [info]: direct mvs  spatial:88.8% temporal:11.2%
x264 [info]: coded y,uvDC,uvAC intra: 70.4% 64.3% 16.7% inter: 22.8% 27.8% 0.2%
x264 [info]: i16 v,h,dc,p: 34% 22% 17% 27%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu:  8%  7% 10% 10% 14% 13% 13% 12% 13%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 11%  9%  4%  9% 14% 14% 14% 11% 15%
x264 [info]: i8c dc,h,v,p: 36% 31% 21% 12%
x264 [info]: Weighted P-Frames: Y:1.2% UV:0.5%
x264 [info]: ref P L0: 70.1% 17.7% 12.1%
x264 [info]: kb/s:5231.52

encoded 142811 frames, 7.18 fps, 5231.52 kb/s

------------------
Encoding Complete
------------------

Sat 10/08/2011
11:52 PM

Re: GT 520 VP5 performance

Posted: Tue Oct 11, 2011 1:30 am
by clumpco
laserfan wrote:Or maybe as Anon17 suggests, my Q6600 and/or hard drives are the limiting factor, but this is disappointing nonetheless. At least I only paid $20 bucks for the new card!
Seven hours for 142811 frames is about 11 fps per pass (unless I dropped a 0). That's about the maximum I could get out of my Q6600 on the higher quality x264 settings. Using a "fast" profile I could get about 25fps on the first pass and 12 fps on the second. DGIndexNV is helping you, but the Q6600 is the limiting factor I'm afraid.

Re: GT 520 VP5 performance

Posted: Tue Oct 11, 2011 8:57 am
by laserfan
clumpco wrote:... the Q6600 is the limiting factor I'm afraid.
I just realized I bought the Q6600 for its "overclock-ability" yet I have never attempted that--maybe I will do so next.

Re: GT 520 VP5 performance

Posted: Wed Nov 02, 2011 8:28 pm
by nibus
How fast of a CPU do you need in order to notice a substantial speed improvement with DGDecNV? Sandy Bridge or the new Bulldozer? I only noticed about .25fps max improvement on my encodes since I bought my 520. I have an i7 930.

Re: GT 520 VP5 performance

Posted: Fri Nov 04, 2011 2:08 am
by clumpco
nibus wrote:How fast of a CPU do you need in order to notice a substantial speed improvement with DGDecNV? Sandy Bridge or the new Bulldozer? I only noticed about .25fps max improvement on my encodes since I bought my 520. I have an i7 930.
Since DGDecNV effectively removes any decoding bottleneck until the encoding frame-rate reaches that of the decoding frame-rate any speed improvement is more or less proportional to the raw processing power of your CPU. Take look at your i7 930's score here http://www.cpubenchmark.net/high_end_cpus.html

Re: GT 520 VP5 performance

Posted: Fri Nov 04, 2011 7:40 am
by laserfan
I wonder can you clumpco or anyone here confirm e.g. that an i5-2500K or i7-2600 or 2700K will encode >2x the Q6600?

Re: GT 520 VP5 performance

Posted: Fri Nov 04, 2011 12:48 pm
by clumpco
laserfan wrote:I wonder can you clumpco or anyone here confirm e.g. that an i5-2500K or i7-2600 or 2700K will encode >2x the Q6600?
You're in luck, I have both a Q6600 and a i7 2600K.
Here are the 1st/2nd pass results for encoding a 1 hour show to 720p @ 3000 kb/s with exactly the same settings for both:

Q6600 - slightly overclocked to 103 MHz FSB
Image
Total encode/mux time: 78m17s

i7 2600 K unlocked to 43x
Image
Total encode/mux time: 32m20s

Will that do?