GT 520 VP5 performance

Support forum for DGDecNV
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

GT 520 VP5 performance

Post by flyordie »

I have started this thread in an attempt to continue the off topic conversation from this thread viewtopic.php?f=8&t=162

My ECS nVidia GeForce GT 520 came in today. It took some trial and error to get the speeds the doom9 forum mentioned in this thread http://forum.doom9.org/showthread.php?p ... ost1513742. The quick and dirty of it is... there is a line in every .dgi file "DEVICE 0" This will use your first video card shown from the DGIndexNV menu : Options > List GPU Devices.

For me this is my main GTX 560 Ti. To use the alternative GT 520 I changed the line to read "DEVICE 1"
The software does this automatically if you modify the DGIndexNV.ini line "CUDA_Device=255" to whichever device you want to force. For me this was "CUDA_Device=1" and then rerun the project. (It was quicker for me to just edit the .dgi file)

I get the following on the rat.264 test file with the GT 520:

Code: Select all

AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]

Number of frames:             839
Length (h:m:s.ms):    0:00:34.993
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (24000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12

Hit ESC to exit...
Frame 839/839, fps (min/max/avg): 67.73 | 146.18 | 131.69
I get the following on the rat.264 test file with the GTX 560 Ti:

Code: Select all

AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]

Number of frames:             839
Length (h:m:s.ms):    0:00:34.993
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (24000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12

Hit ESC to exit...
Frame 839/839, fps (min/max/avg): 35.95 | 75.51 | 63.00
At present my capture content is 1080i. My avs files look like this:

Code: Select all

DGSource("Q:\DVR\e1.dgi",resize_w=768, resize_h=432, use_pf=true, deinterlace=2)
TDecimate(mode=2, rate=23.976,m2PA=true)
My x264_64.exe command line looks like this:

Code: Select all

x264_x64.exe --thread-input --direct auto --crf 17.5 e1.avs -o e1.17.5.mkv --ref 9 --me umh --subme 8 --merange 24 --trellis 1 --bframes 9 --b-adapt 2 --rc-lookahead 40 --fullrange on
My GT 520 outputs the following:

Code: Select all

Q:\DVR>x264_x64.exe --thread-input --direct auto --crf 17.5 e1.avs -o e1.17.5.mkv --ref 9 --me umh --subme 8 --merange 24 --trellis 1 --bframes 9 --b-adapt 2 --rc-lookahead 40 --fullrange on
avs [info]: 768x432p 0:0 @ 24000/1001 fps (cfr)
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 FastShuffle SSE4.2
x264 [info]: profile High, level 3.1
x264 [info]: frame I:150   Avg QP:14.02  size: 46081
x264 [info]: frame P:2414  Avg QP:17.27  size: 12413
x264 [info]: frame B:8044  Avg QP:20.53  size:  2376
x264 [info]: consecutive B-frames:  3.3%  1.6%  7.6% 27.6% 31.3% 17.4%  5.9%  4.6%  0.8%  0.0%
x264 [info]: mb I  I16..4: 10.1% 65.6% 24.3%
x264 [info]: mb P  I16..4:  2.8%  5.9%  1.5%  P16..4: 38.7% 25.5% 14.2%  0.0%  0.0%    skip:11.3%
x264 [info]: mb B  I16..4:  0.1%  0.2%  0.1%  B16..8: 36.6%  5.7%  1.2%  direct: 3.0%  skip:52.9%  L0:39.3% L1:48.2% BI:12.5%
x264 [info]: 8x8 transform intra:60.0% inter:76.5%
x264 [info]: direct mvs  spatial:99.9% temporal:0.1%
x264 [info]: coded y,uvDC,uvAC intra: 71.6% 85.2% 67.2% inter: 17.1% 23.1% 3.5%
x264 [info]: i16 v,h,dc,p: 47% 11%  4% 38%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 18% 12%  8%  7% 10% 12% 11% 11% 12%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 18% 10%  6%  8% 13% 13% 11% 10% 11%
x264 [info]: i8c dc,h,v,p: 42% 22% 23% 14%
x264 [info]: Weighted P-Frames: Y:1.9% UV:0.7%
x264 [info]: ref P L0: 51.1% 12.9% 17.3%  5.3%  4.5%  2.9%  2.7%  1.5%  1.6%  0.2%  0.0%
x264 [info]: ref B L0: 83.7%  9.2%  3.4%  1.5%  0.9%  0.8%  0.4%  0.2%
x264 [info]: ref B L1: 93.9%  6.1%
x264 [info]: kb/s:1012.43

encoded 10608 frames, 25.90 fps, 1012.44 kb/s
Here is a gpu-z snapshot of how the GT 520 is being utilized just before it finishes.
Image

My GTX 560 Ti outputs the following:

Code: Select all

Q:\DVR>x264_x64.exe --thread-input --direct auto --crf 15 e1.avs -o e1.15.mkv --ref 9 --me umh --subme 8 --merange 24 --trellis 1 --bframes 9 --b-adapt 2 --rc-lookahead 60 --fullrange on
avs [info]: 768x432p 0:0 @ 24000/1001 fps (cfr)
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 FastShuffle SSE4.2
x264 [info]: profile High, level 3.1
x264 [info]: frame I:150   Avg QP:11.48  size: 60158
x264 [info]: frame P:2414  Avg QP:14.69  size: 18424
x264 [info]: frame B:8044  Avg QP:17.97  size:  3643
x264 [info]: consecutive B-frames:  3.3%  1.6%  7.4% 27.9% 31.1% 17.1%  6.5%  4.4%  0.8%  0.0%
x264 [info]: mb I  I16..4:  7.6% 65.6% 26.8%
x264 [info]: mb P  I16..4:  2.7%  8.0%  2.1%  P16..4: 32.1% 29.7% 18.0%  0.0%  0.0%    skip: 7.5%
x264 [info]: mb B  I16..4:  0.1%  0.4%  0.1%  B16..8: 35.6%  9.6%  2.2%  direct: 5.1%  skip:46.8%  L0:38.2% L1:43.9% BI:17.8%
x264 [info]: 8x8 transform intra:63.1% inter:71.3%
x264 [info]: direct mvs  spatial:99.9% temporal:0.1%
x264 [info]: coded y,uvDC,uvAC intra: 79.7% 90.2% 79.4% inter: 24.2% 28.9% 7.6%
x264 [info]: i16 v,h,dc,p: 48% 10%  4% 38%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 19% 12% 11%  7%  9% 11%  9% 10% 11%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 18% 11%  6%  8% 12% 12% 11% 10% 11%
x264 [info]: i8c dc,h,v,p: 47% 20% 21% 12%
x264 [info]: Weighted P-Frames: Y:1.9% UV:0.7%
x264 [info]: ref P L0: 50.2% 12.3% 17.8%  5.6%  4.8%  3.0%  2.8%  1.6%  1.6%  0.2%  0.0%
x264 [info]: ref B L0: 82.1% 10.0%  3.5%  1.7%  1.1%  0.9%  0.5%  0.2%
x264 [info]: ref B L1: 94.1%  5.9%
x264 [info]: kb/s:1497.26

encoded 10608 frames, 44.19 fps, 1497.27 kb/s

Q:\DVR>F:\Video\Sirens\avsmeter\AVSMeter.exe F:\Video\Sirens\avsmeter\e2.avs
Here is a gpu-z snapshot of how the GTX 560 Ti is being utilized just before it finishes.
Image

Neither of these are utilizing the Video Engine. I have encoded in the past and seen the Video engine utilized at 80%-99%, but at the moment I cannot remember what the content type was specifically. For a $50, before a $20 mail in rebate, it appears to perform at 50% of the speed of my GTX 560 Ti which I feel is still rather impressive. :D

Could you recommend different settings that would give a better comparison of the hardware with respect to your awesome software?

Before you ask, UseD3D=0 is the setting in my DGIndexNV.ini :D
As an after thought I ran my .avs through AVSMeter to see what it thought. It speeds up if I turn off deinterlace=2 and usepf=1, but I guess for Mpeg2 Main@High the video engine isn't used. :(
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

flyordie wrote:The quick and dirty of it is... there is a line in every .dgi file "DEVICE 0" This will use your first video card shown from the DGIndexNV menu : Options > List GPU Devices.
Don't edit the DGI file. If you do that DGIndexNV will not use your setting, but only DGDecodeNV. To get both to use your chosen device, you have to edit the INI file.

When testing for decoding speed, you should use a bare script, i.e., no resizing, deinterlacing, TDecimate, etc.

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.
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

Using your your comments about not changing the .dgi file, I ran the sample through DGIndexNV twice. Once with CUDA_Device=255 (My GTX 560 Ti) and once with CUDA_Device=1 (My GT 520)
I updated my script so it only has this for the GTX 560 Ti:

Code: Select all

DGSource("Q:\DVR\wc_560ti.dgi")

Code: Select all

Q:\DVR>AVSMeter.exe wc_560ti.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13260
Length (h:m:s.ms):    0:07:22.442
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12

Hit ESC to exit...
Frame 13260/13260, fps (min/max/avg): 176.09 | 409.37 | 254.13
and for the GT 520:

Code: Select all

DGSource("Q:\DVR\wc_520.dgi")
Here is the throughput for each:

Code: Select all

Q:\DVR>AVSMeter.exe wc_520.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13260
Length (h:m:s.ms):    0:07:22.442
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12

Hit ESC to exit...
Frame 13260/13260, fps (min/max/avg): 66.13 | 165.62 | 99.63
I initially misread your statement and thought if I used a VP3/VP2 GPU I would get the video engine instead of the CUDA. Before realizing my mistake I ran it on a 8600 GT I have in another computer.
I realize now that you changed it so all GPUs use CUDA for MPEG2, but nevertheless this is the throughput on the 8600GT:

Code: Select all

AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]

Number of frames:           13260
Length (h:m:s.ms):    0:07:22.442
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12

Hit ESC to exit...
Frame 13260/13260, fps (min/max/avg): 51.43 | 120.07 | 72.72
The GTX 560 Ti has 384 CUDA Cores and the GT 520 has 48 CUDA Cores.
When running AVSMeter, my GTX 560 Ti maxes GPU load around 46% and the GT 520 gets around 79% GPU load.
Would having more CUDA Cores improve my throughput or does it look like a diminishing return?
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

>Would having more CUDA Cores improve my throughput or does it look like a diminishing return?

I don't know because I have only one card and I know nothing about the internal implementation.
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

How are you calling the CUDA API? Is it possible to have a look at that section of the source code?
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

My source is closed. However, the DGAVCDecNV development dialog at my website will show you the basic idea, but without all the extra stuff needed to implement DGDecNV functionality.

http://rationalqm.us/dgdecnv/cuda/cuda.html

What are you after in asking that question?
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

More of a curiosity than anything. My coding abilities haven't progressed beyond the C++ classes I took in college. If anything, I was most likely going to compare to LAV CUVID Decoder and see if I could see any similar approaches and see if there was anyway to improve your stuff. Its all probably way beyond my level of understanding at this point. I think that link you posted in your last post will probably satiate my interest for now. It just seemed odd it was only utilizing 46% of my GPU.
DAE avatar
slomorbo
Posts: 10
Joined: Tue Jan 25, 2011 9:28 pm

Re: GT 520 VP5 performance

Post by slomorbo »

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.
I was wondering if this bug is hardware related, I mean, will it be fix on a new driver release or it's a permanant bug on vp3-4 engine?
In the other hand, on my G210 with my test clip i got ~20fps more using CUDA instead of the VP4 engine (tested with v.2039 vs v.2037).
The result flyordie got with his 560ti are really amazing considering that if he use an older dgdecnv version that use the VP4 engine on his card he'll get far less fps.
If that bug does not affect the VP5 engine, can you add an option to either use CUDA or the VP engine?

Thank a lot
User avatar
zys4416
Posts: 9
Joined: Fri Jul 29, 2011 9:04 pm

Re: GT 520 VP5 performance

Post by zys4416 »

Wow, nice post!
I am also prepared to buy an Nvidia card for dgdecnv.
But just from the post itself, it doesn't seem to specify whether I should choose GT520 or other high-end cards (considering the quality and stability only). Can you please advise?
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

The currently available high-end cards do not have the VP5 engine. This will change in the fall.

For now, if you have high-end needs for gaming, etc., you'll have to settle for the VP4 engine on a high-end card. If high-end graphics is not a requirement and you are focused on video decoding then the 520 will give you better bang for your buck.
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

slomorbo wrote:[I was wondering if this bug is hardware related, I mean, will it be fix on a new driver release or it's a permanant bug on vp3-4 engine?
At the time I made the change, Nvidia said a driver fix would be done, but they weren't sure which version it would get into. I assume it is in there by now. However, they also said that for MPEG2, performance would better with CUDA than the VP engine, and my testing confirmed it. So I have been uninterested in reverting to the VP engine for MPEG2. I do not know whether the situation has changed for VP5 as I do not have a card yet. When I do I will reconsider the situation.
If that bug does not affect the VP5 engine, can you add an option to either use CUDA or the VP engine?
As I said, I will re-assess things when I have a VP5 to test. I could ask Nvidia but I don't like to bother them over things I can work out for myself. If someone with a 520 wants to test this I'd be happy to provide a test build that is reverted to the VP engine for MPEG2.
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

I would be ELATED to test VP5 for you! :D

I would like to note the obvious, progressive AVC, VC-1 HD video is where the GT 240 really shines (at present). Interlaced and MPEG2 HD content is the only thing that performs better on the GTX 560 Ti probably because of the CUDA cores or however nVidia CUVID deals with deinterlacing.
I have a an original source that is 31.1Mbps, AVC, High@4.1, 23.976 fps, 1080p.
I'm using DGIndexNV to crop the top and bottom black bars. The Aspect is 2.37.

Code: Select all

DGSource("Q:\Go.dgi",resize_w=1280, resize_h=544, use_pf=true)
RemoveGrain(mode=2).Sharpen(0.2)
GT 520

Code: Select all

AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:          149665
Length (h:m:s.ms):    1:44:02.278
Frame width:                 1280
Frame height:                 544
Framerate:                 23.976 (24000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 24384/149665, fps (min/max/avg): 64.86 | 150.25 | 126.95
GTX 560 Ti

Code: Select all

AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:          149665
Length (h:m:s.ms):    1:44:02.278
Frame width:                 1280
Frame height:                 544
Framerate:                 23.976 (24000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13325/149665, fps (min/max/avg): 45.49 | 82.17 | 54.55
In real world performance with this source:
Using x264_x64 --quality medium, the GT 520 out performs my GTX 560 Ti.
On the encode, the best I can get on GTX 560 Ti is around 54.05 fps (100% on the Video Engine and like 1-8% on the GPU).
The GT 520 is able to achieve 103 fps! (93% on the Video Engine and like 12-21% on the GPU)

If the --quality preset is any higher than medium (slow, slower, veryslow, placebo) then the two cards perform within 2-8 fps of each other since the bottleneck becomes the CPU.
x264_x64 --quality slow gets around 45.25 fps on the GT 520. (20-38% Video Engine and 4-8% GPU usage)
x264_x64 --preset slow gets around 44.77 fps on the GTX 560 Ti. (60-100% Video Engine and 1% GPU usage)
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

flyordie wrote:I would be ELATED to test VP5 for you!
Hey, that's great. A lot of people will be curious about this, including the white cat. :)

Here is a 32-bit build of 2040 reverted to the VPU for MPEG2:

http://rationalqm.us/misc/DGDecNV2040mpeg2vpu.zip
I would like to note the obvious, progressive AVC, VC-1 HD video is where the GT 240 really shines (at present). Interlaced and MPEG2 HD content is the only thing that performs better on the GTX 560 Ti probably because of the CUDA cores or however nVidia CUVID deals with deinterlacing.
It's not so obvious. CUVID does nothing with interlacing/deinterlacing. There's no reason for any big difference between progressive and interlaced content.
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

neuron2 wrote:It's not so obvious. CUVID does nothing with interlacing/deinterlacing. There's no reason for any big difference between progressive and interlaced content.
To answer this I've put in some data points where Deinterlacing is used.

OK, I've broken this post into 4 sections. Each section has 5 encodings I've specified the .avs files before the results.
For a source file I have an opening section from a TV Show.

Code: Select all

Size                        : 724MB (759,565,722 bytes)
Format                      : MPEG Video
Format version              : Version 2
Format profile              : Main@High
Format settings, BVOP       : Yes
Format settings, Matrix     : Default
Bit rate mode               : Variable
Length                      : 7 Minutes 43 seconds
Maximum bit rate            : 18.516 Mbps
Average bit rate            : 12.763 Mbps
Width                       : 1 920 pixels
Height                      : 1 080 pixels
Display aspect ratio        : 16:9
Frame rate                  : 29.970 fps
Color space                 : YUV
Chroma subsampling          : 4:2:0
Bit depth                   : 8 bits
Scan type                   : Interlaced
Scan order                  : Top Field First
Compression mode            : Lossy
C:\video\DGDecNV contains DVDecNV Release 2040
C:\video\DGDecNV_2 contains DVDecNV Version 2040 with VPU enabled (from the previous post)

First up is the GT 520 with the VPU enabled version for MPEG2 (32-Bit):

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\s.dgi")

Code: Select all

Q:\Iso>avsmeter s1.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 120.32 | 213.21 | 190.76
GPU Load: 22-24%
Video Engine Load: 99%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\s.dgi", use_pf=true)

Code: Select all

Q:\Iso>avsmeter s1_usepf.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 120.62 | 212.41 | 190.77
GPU Load: 22-24%
Video Engine Load: 99%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\s.dgi", use_pf=true, deinterlace=1)

Code: Select all

Q:\Iso>avsmeter s1_deint1_usepf.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 61.11 | 103.32 | 102.05
GPU Load: 81-82%
Video Engine Load: 54-55%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\s.dgi", use_pf=true, deinterlace=2)

Code: Select all

Q:\Iso>avsmeter s1_deint2_usepf.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           27792
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 59.940 (60000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 27792/27792, fps (min/max/avg): 66.85 | 115.82 | 114.00
GPU Load: 80-81%
Video Engine Load: 29-31%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\s.dgi", use_pf=true, deinterlace=2)
TDecimate(mode=2, rate=23.976,m2PA=true)

Code: Select all

Q:\Iso>avsmeter s1_deint2_usepf_Decimate.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           11116
Length (h:m:s.ms):    0:07:43.630
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (3528096 / 147151)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 11116/11116, fps (min/max/avg): 3.96 | 42.66 | 29.91
GPU Load: 69-72%
Video Engine Load: 27%


Second up is the GT 520 using the CUDA enabled version for MPEG2 (32-Bit):

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\t.dgi")

Code: Select all

Q:\Iso>avsmeter T1.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 63.84 | 126.42 | 101.71
GPU Load: 74-77%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\t.dgi", use_pf=true)

Code: Select all

Q:\Iso>avsmeter T1_usePF.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 63.86 | 126.35 | 101.67
GPU Load: 74-76%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\t.dgi", use_pf=true, deinterlace=1)

Code: Select all

Q:\Iso>avsmeter T1_usePF_deint1.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 34.14 | 68.27 | 60.42
GPU Load: 84-86%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\t.dgi", use_pf=true, deinterlace=2)

Code: Select all

Q:\Iso>avsmeter T1_usePF_deint2.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           27792
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 59.940 (60000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 27792/27792, fps (min/max/avg): 44.10 | 92.15 | 82.01
GPU Load: 81-83%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\t.dgi", use_pf=true, deinterlace=2)
TDecimate(mode=2, rate=23.976,m2PA=true)

Code: Select all

Q:\Iso>avsmeter T1_usePF_deint2_decimate.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           11116
Length (h:m:s.ms):    0:07:43.630
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (3528096 / 147151)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 11116/11116, fps (min/max/avg): 2.64 | 34.98 | 22.26
GPU Load: 79-87%
Video Engine Load: 0%

Third up is the GTX 560 Ti with the VPU enabled version for MPEG2 (32-Bit):

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\w.dgi")

Code: Select all

Q:\Iso>avsmeter w.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 167.18 | 332.73 | 260.58
GPU Load: 35-39%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\w.dgi", use_pf=true)

Code: Select all

Q:\Iso>avsmeter w_usepf.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 167.23 | 333.86 | 261.05
GPU Load: 36-38%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\w.dgi", use_pf=true, deinterlace=1)

Code: Select all

Q:\Iso>avsmeter w_usepf_deint1.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 50.63 | 125.32 | 106.54
GPU Load: 32-36%
Video Engine Load: 99%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\w.dgi", use_pf=true, deinterlace=2)

Code: Select all

Q:\Iso>avsmeter w_usepf_deint2.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           27792
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 59.940 (60000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 27792/27792, fps (min/max/avg): 67.24 | 212.36 | 206.27
GPU Load: 63-65%
Video Engine Load: 94-99%

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\w.dgi", use_pf=true, deinterlace=2)
TDecimate(mode=2, rate=23.976,m2PA=true)

Code: Select all

Q:\Iso>avsmeter w_usepf_deint2_decimate.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           11116
Length (h:m:s.ms):    0:07:43.630
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (3528096 / 147151)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 11116/11116, fps (min/max/avg): 6.04 | 72.44 | 49.79
GPU Load: 44-71%
Video Engine Load: 77-99%

Fourth up is the GTX 560 Ti with the CUDA enabled version for MPEG2 (32-Bit):

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\x.dgi")

Code: Select all

Q:\Iso>avsmeter x.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 167.08 | 331.74 | 260.67
GPU Load: 36-39%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\x.dgi", use_pf=true)

Code: Select all

Q:\Iso>avsmeter x_usepf.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 169.64 | 332.29 | 260.50
GPU Load: 35-39%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\x.dgi", use_pf=true, deinterlace=1)

Code: Select all

Q:\Iso>avsmeter x_usepf_deint1.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           13896
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 13896/13896, fps (min/max/avg): 88.99 | 178.25 | 155.30
GPU Load: 59-65%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\x.dgi", use_pf=true, deinterlace=2)

Code: Select all

Q:\Iso>avsmeter x_usepf_deint2.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           27792
Length (h:m:s.ms):    0:07:43.663
Frame width:                 1920
Frame height:                1080
Framerate:                 59.940 (60000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 27792/27792, fps (min/max/avg): 104.52 | 200.96 | 179.54
GPU Load: 59-66%
Video Engine Load: 0%

Code: Select all

loadplugin("C:\video\DGDecNV\DGDecodeNV.dll")
DGSource("Q:\Iso\x.dgi", use_pf=true, deinterlace=2)
TDecimate(mode=2, rate=23.976,m2PA=true)

Code: Select all

Q:\Iso>avsmeter x_usepf_deint2_decimate.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:           11116
Length (h:m:s.ms):    0:07:43.630
Frame width:                 1920
Frame height:                1080
Framerate:                 23.976 (3528096 / 147151)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 11116/11116, fps (min/max/avg): 6.16 | 70.41 | 47.71
GPU Load: 55-72%
Video Engine Load: 0%


My GTX 460 arrived back today. If you would like, I can also run these tests on it. The GT 520 has 48 CUDA cores, GTX 460 has 336 CUDA cores, and the GTX 560 Ti has 384 CUDA cores. This may help show the scale of the CUDA cores... I'm thinking I could adjust the speeds down so that the frequencies match if you think that would help with this comparison. (Assuming you would like this test data on GTX 460) Let me know.
These are my card specs:
GeForce GTX 460 (Fermi)
Core Clock 715MHz
Shader Clock 1430MHz
Stream Processors 336 Processor Cores
Effective Memory Clock 3600MHz
Actual Memory Clock 900MHz
Memory Size 1GB
Memory Interface 256-bit
Memory Type GDDR5

GeForce GTX 560 Ti (Fermi)
Core Clock 850MHz
Shader Clock 1700MHz
Stream Processors 384 Processor Cores
Effective Memory Clock 4104MHz
Actual Memory Clock 1050MHz
Memory Size 1GB
Memory Interface 256-bit
Memory Type GDDR5
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

Thanks for the extensive testing. All we really care about is CUDA versus VP5 for plain source script on the 520. You report this:

CUDA: 101
VP5: 190

So for VP5 we should revert to the VPU, as long as the decoding bug is fixed. I will post a link to a test stream for that tomorrow morning.

VP5 is looking better and better!
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

Flagship VP5 won't get here till 2012 :(

I ran the GTX 460 through the same test.

If you re-enable VPU would you allow the end user to switch between the two? Perhaps a DecodeEngine=1 or something in the AVS call?

I thought it was strange the GTX 560 Ti had virtually the same results for CUDA and VPU. Did that package only utilize VP5? I did notice the video engine engage when I was deinterlacing. I've been having to deal with a lot of 1080i source lately. I prefer to use your program for all my deinterlacing and resizing. It keeps my avs scripts short and my encodes fast. :D

To summarize my previous post and group them by type of encode to better compare the different cards:
Plain Script (Average FPS)

Code: Select all

                   CUDA        VPU
GT 520           101.71 fps  190.76 fps
GTX 460          240.49 fps  241.98 fps
GTX 560 Ti       260.67 fps  260.58 fps
DeInterlace=1 Script (Average FPS)

Code: Select all

                   CUDA        VPU
GT 520            60.42 fps  102.05 fps
GTX 460          129.61 fps  106.51 fps
GTX 560 Ti       155.30 fps  106.54 fps
DeInterlace=2 Script (Average FPS)

Code: Select all

                   CUDA        VPU
GT 520            82.01 fps  114.00 fps
GTX 460          149.61 fps  172.17 fps
GTX 560 Ti       179.54 fps  206.27 fps
DeInterlace=2,Decimate Script (Average FPS)

Code: Select all

                   CUDA        VPU
GT 520            22.26 fps   29.91 fps
GTX 460           39.51 fps   42.77 fps
GTX 560 Ti        47.71 fps   49.79 fps
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

flyordie wrote:If you re-enable VPU would you allow the end user to switch between the two? Perhaps a DecodeEngine=1 or something in the AVS call?
I plan to detect the VPU version if possible and choose based on that.
I thought it was strange the GTX 560 Ti had virtually the same results for CUDA and VPU. Did that package only utilize VP5?
560 does not have VP5.

Here is the stream to test the VP5 with:

http://rationalqm.us/guest/Olivier/4.demuxed.m2v

Please test it and advise if you see any artifacts on the 520 with VPU.
DAE avatar
slomorbo
Posts: 10
Joined: Tue Jan 25, 2011 9:28 pm

Re: GT 520 VP5 performance

Post by slomorbo »

I download your test clip and I test it with your modified version.
Using Geforce G210 that have a VP4 Engine to see if I will see actifact/corruption
Using basic avisynth script

Code: Select all

LoadPlugin("C:\video\tools\dgdecnv_2\DGDecodeNV.dll")
DGSource("C:\Users\slomorbo\Desktop\test\4demuxed.dgi")
Using driver version 275.33 (dated 2011-05-20) -> no actifact/corruption
Using driver version 280.26 (dated 2011-08-03) -> no actifact/corruption

I unfortunatly don't have drivers in between to test.
User avatar
flyordie
Posts: 34
Joined: Thu Nov 18, 2010 10:07 am

Re: GT 520 VP5 performance

Post by flyordie »

I'm using 280.26 nVidia drivers on the GT 520.
Since this is so short I think the avsmeter results are inflated...nevertheless
The CUDA dll resulted in this:

Code: Select all

Q:\Iso>avsmeter 4.demuxed.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:             173
Length (h:m:s.ms):    0:00:05.772
Frame width:                  720
Frame height:                 480
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 173/173, fps (min/max/avg): 308.96 | 497.37 | 449.15
For VPU, running this avs script:

Code: Select all

loadplugin("C:\video\DGDecNV_2\DGDecodeNV.dll")
DGSource("Q:\Iso\4.demuxed.dgi")
Resulted in this:

Code: Select all

Q:\Iso>avsmeter 4.demuxed.avs
AVSMeter v1.12 (Aug  4 2011) by Groucho2004
AviSynth 2.60, build:Aug 13 2009 [15:07:40]
Number of frames:             173
Length (h:m:s.ms):    0:00:05.772
Frame width:                  720
Frame height:                 480
Framerate:                 29.970 (30000 / 1001)
Progressive:                  Yes
Colorspace:                  YV12
Hit ESC to exit...
Frame 173/173, fps (min/max/avg): 610.15 | 1028.69 | 929.26
The output looked good. I used the avs script above and encoded it with x264 (x86) with crf 0 (lossless)

Code: Select all

Q:\Iso>x264 --preset placebo --crf 0 4.demuxed.avs -o 4.losless.mkv
avs [info]: 720x480p 0:0 @ 30000/1001 fps (cfr)
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 FastShuffle SSE4.2
x264 [info]: profile High 4:4:4 Predictive, level 4.0, 4:2:0 8-bit
x264 [info]: frame I:6     Avg QP: 0.00  size:155896
x264 [info]: frame P:167   Avg QP: 0.00  size: 88490
x264 [info]: mb I  I16..4: 40.7% 21.4% 37.9%
x264 [info]: mb P  I16..4:  4.4%  7.1%  4.4%  P16..4: 39.1% 11.2%  4.7%  0.9%  0.4%    skip:27.9%
x264 [info]: 8x8 transform intra:40.4% inter:56.2%
x264 [info]: coded y,uvDC,uvAC intra: 93.4% 92.7% 92.6% inter: 57.2% 47.3% 47.0%
x264 [info]: i16 v,h,dc,p: 40% 56%  3%  0%
x264 [info]: i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 35% 60%  4%  0%  0%  0%  0%  0%  0%
x264 [info]: i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 32% 56%  3%  2%  2%  2%  1%  1%  1%
x264 [info]: i8c dc,h,v,p:  8% 56% 35%  1%
x264 [info]: Weighted P-Frames: Y:3.0% UV:2.4%
x264 [info]: ref P L0: 64.6%  7.5% 15.2%  5.7%  2.9%  1.7%  0.5%  0.1%  0.2%  0.6%  0.3%  0.1%  0.1%  0.1%  0.2%  0.2%
x264 [info]: kb/s:21776.91
encoded 173 frames, 15.99 fps, 21777.66 kb/s
Here is the output file http://www.mediafire.com/?458ssqu248c184b
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

flyordie wrote:To summarize my previous post and group them by type of encode to better compare the different cards:
Plain Script (Average FPS)

Code: Select all

                   CUDA        VPU
GT 520           101.71 fps  190.76 fps
GTX 460          240.49 fps  241.98 fps
GTX 560 Ti       260.67 fps  260.58 fps
I think something must have gone wrong with your testing. On my 460, CUDA runs 4 times faster than VPU for an HD MPEG2 stream (tested in DGIndexNV with display disabled). Also, Nvidia states that on the 460 CUDA will perform way better for MPEG2.

@slomorbo

Thank you for your test results. That is good news.
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

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?
DAE avatar
slomorbo
Posts: 10
Joined: Tue Jan 25, 2011 9:28 pm

Re: GT 520 VP5 performance

Post by slomorbo »

I when out and I bought a gt520 and I done some testing.
My Test clip ;
1080i MPEG2 @~17.5mbps

I've done the same test using a G210 (VP4) and a GT520(vp5), both using driver version 280.26
In each test I used the benchmark function in DGIndexNV (f5 with display disable) and via AVSMETER (just Loadplugin and DGSource)
I used DGindexNV v.2040 stock(with stock 2040 DGDecodeNV.dll) for the CUDA testing and DGindexNV v.2040 with mod(with mod DGDecodeNV.dll) for VP testing


Image

One thing surprised me, it's seem that DGDecodeNV.dll (used in a avisynth script in CUDA mode) is not able to deliver the same level of fps than DGIndexNV. It only give ~64% on my G210 and ~55.75% on my GT520 :shock:
I used Avisynth 2.6 alpha 3 for testing, and after those results I got back to 2.5.8 just to be sure and I got the exact same results.

hope it's help :)
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

slomorbo wrote:hope it's help
It's very helpful, thank you! What are the percentages in the table? Also, your table should show VP5 for the 520.

I have never used AvsMeter but will look into it to see if there is a reason for the anomaly you noted.
User avatar
admin
Site Admin
Posts: 4449
Joined: Thu Sep 09, 2010 3:08 pm

Re: GT 520 VP5 performance

Post by admin »

Groucho2004 wrote:I meant the short MPEG-2 clip with 173 frames you uploaded.
I know what you meant. I'm just pointing out that the small clip was not intended for performance testing, but just to test that the decoding bug was fixed in current driver versions.
DAE avatar
slomorbo
Posts: 10
Joined: Tue Jan 25, 2011 9:28 pm

Re: GT 520 VP5 performance

Post by slomorbo »

neuron2 wrote:
slomorbo wrote:hope it's help
It's very helpful, thank you! What are the percentages in the table? Also, your table should show VP5 for the 520.

I have never used AvsMeter but will look into it to see if there is a reason for the anomaly you noted.
I fix my typo error on the table ;)
percentages in the table for CUDA are the GPU utilization from GPU-Z
percentages in the table for VP are the VP engine utilization from GPU-Z

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"
CUDA -> ~77.5fps using ~55% of my cpu (so not cpu limited) versus 77.5fps on AvsMeter
VP5 -> 160fps using 100% CPU (so CPU LIMITED) versus 176.50fps on AvsMeter
Post Reply