VDPAU testing tool - Page 5

crisalide · 06-01-09, 07:18 AM

Quote:

Originally Posted by cehoyos

To add some alternative numbers:

Code:

SURFACE GET BITS: 47.0154 M/s
SURFACE PUT BITS: 46.6615 M/s

MPEG DECODING (1920x1080): 34 pic/s
H264 DECODING (1920x1080): 44 pic/s
VC1 DECODING (1440x1080): 26 pic/s

MIXER WEAVE (1920x1080): 255 frames/s
MIXER BOB (1920x1080): 466 fields/s
MIXER TEMPORAL (1920x1080): 116 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 63 fields/s

Geforce 9400 PCI on a Pentium III 500

Real-life MPEG and VC1 HD samples do not decode in real-time with this setup (H264 works fine even for bitrates higher than level 4.1 and using de-interlacing).

Carl Eugen

Looks like a cpu bottleneck since MPEG and VC1 are partially accelerated on this GPU. For such a cpu, i think a G98 (with complete acceleration of MPEG, H264 and VC1) would be a better choice.

jyavenard · 06-01-09, 08:44 AM

Quote:

Originally Posted by crisalide

I would say it's a powermizer issue. Look at previous posts, some users had the same kind of problem and i think some have found solutions.

It's running 185.18.14 which is supposed to fix powermizer issues,
plus I have in xorg.conf Option "RegistryDwords" "PowerMizerLevelAC=0x1"

supposed to force performance mode... to no available

plus it's surprising to get good results on the deinterlacers but not on the decoding

Quantumstate · 06-01-09, 09:18 AM

Quote:

Originally Posted by jyavenard

you must install libqt4-dev if you want to compile qt4 programs....

Ah, that was the needed magick. I'll bet that should be in the ReadMe.

Quadro 1600M

Code:

SURFACE GET BITS: 640.084 M/s
SURFACE PUT BITS: 567.362 M/s

MPEG DECODING (1920x1080): 62 pic/s
H264 DECODING (1920x1080): 43 pic/s
VC1 DECODING (1440x1080): 82 pic/s

MIXER WEAVE (1920x1080): 878 frames/s
MIXER BOB (1920x1080): 1286 fields/s
MIXER TEMPORAL (1920x1080): 282 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 88 fields/s

So am I understanding that this is fast enough to do Advanced2X deinterlacing for 1080p?

And am I understanding that it is not fast enough to decode H.264?

jyavenard · 06-01-09, 09:24 AM

Quote:

Originally Posted by Quantumstate

Ah, that was the needed magick. I'll bet that should be in the ReadMe.

Quadro 1600M

Code:

SURFACE GET BITS: 640.084 M/s
SURFACE PUT BITS: 567.362 M/s

MPEG DECODING (1920x1080): 62 pic/s
H264 DECODING (1920x1080): 43 pic/s
VC1 DECODING (1440x1080): 82 pic/s

MIXER WEAVE (1920x1080): 878 frames/s
MIXER BOB (1920x1080): 1286 fields/s
MIXER TEMPORAL (1920x1080): 282 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 88 fields/s

So am I understanding that this is fast enough to do Advanced2X deinterlacing for 1080p?

And am I understanding that it is not fast enough to decode H.264?

looks fast enough to me for everything.

your typical HD h264 content is from a bluray, which is 24 pic/s ; your card can do twice that.

Quantumstate · 06-01-09, 10:58 AM

Quote:

Originally Posted by jyavenard

looks fast enough to me for everything.

your typical HD h264 content is from a bluray, which is 24 pic/s ; your card can do twice that.

Here is 100% of the HD content I watch:

DishNetwork satellite H.264/AVC

Code:

Video
ID                               : 4642 (0x1222)
Menu ID                          : 1 (0x1)
Format                           : AVC
Format/Info                      : Advanced Video Codec
Format profile                   : Main@L4.0
Format settings, CABAC           : Yes
Format settings, ReFrames        : 4 frames
Duration                         : 59mn 51s
Bit rate                         : 5 279 Kbps
Width                            : 1 440 pixels
Height                           : 1 080 pixels
Display aspect ratio             : 16/9
Frame rate                       : 29.970 fps
Standard                         : Component
Resolution                       : 24 bits
Colorimetry                      : 4:2:0
Scan type                        : Interlaced
Scan order                       : Top Field First
Bits/(Pixel*Frame)               : 0.113
Stream size                      : 2.21 GiB (87%)

US ATSC mpeg2:

Code:

Video
ID                               : 224 (0xE0)
Format                           : MPEG Video
Format version                   : Version 2
Format profile                   : Main@High
Format settings, Matrix          : Default
Duration                         : 1h 52mn
Bit rate mode                    : Variable
Bit rate                         : 10.1 Mbps
Nominal bit rate                 : 19.4 Mbps
Width                            : 1 280 pixels
Height                           : 720 pixels
Display aspect ratio             : 16/9
Frame rate                       : 59.940 fps
Colorimetry                      : 4:2:0
Scan type                        : Progressive
Bits/(Pixel*Frame)               : 0.183
Stream size                      : 7.93 GiB (89%)

OK, so satellite is 29.970 interlaced, fair enough my decoder can handle that. But ATSC is 59.970 progressive, and the decoder is rated at 62 pic/s. However this is at 1080 and my sample is 720. I guess this is OK.

I've just updated my nVidia driver from 185.18.10 to 185.18.14 and here are the results:

Quadro 1600M (HP 8710w laptop):

Code:

SURFACE GET BITS: 528.471 M/s
SURFACE PUT BITS: 474.799 M/s

MPEG DECODING (1920x1080): 60 pic/s
H264 DECODING (1920x1080): 43 pic/s
VC1 DECODING (1440x1080): 83 pic/s

MIXER WEAVE (1920x1080): 891 frames/s
MIXER BOB (1920x1080): 1304 fields/s
MIXER TEMPORAL (1920x1080): 286 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 90 fields/s

9300 IGP (Asus P5N7A-VM mobo)

Code:

SURFACE GET BITS: 984.095 M/s
SURFACE PUT BITS: 1112.54 M/s

MPEG DECODING (1920x1080): 60 pic/s
H264 DECODING (1920x1080): 53 pic/s
VC1 DECODING (1440x1080): 67 pic/s

MIXER WEAVE (1920x1080): 297 frames/s
MIXER BOB (1920x1080): 451 fields/s
MIXER TEMPORAL (1920x1080): 109 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 33 fields/s

I am surprised to find that the 9300 performs worse than the 1600M. It is supposed to be so much better.

And I am not clear on how to apply these findings to whether these deinterlacers are fast enough for Advanced2X?

And disturbingly, when I close the final window for qvdpautest on my 9300 system, it locks up the machine entirely. I have to power down and up, to get it back.

crisalide · 06-01-09, 11:15 AM

@Quantumstate

If you can provide a 720p MPEG2 sample, i would add 720p mpeg decoding test. (never seen any until now)

Quantumstate · 06-01-09, 12:48 PM

Uploading now, but my ISP has upload throttled to 40Kbs, so this 194MB file will take 6 hours. I would fire Queerwire, except they are so cheap and download is 1.5Mbps.

garyl · 06-01-09, 01:56 PM

Hilary Duff - With love 720p (mpeg2, 280MB)
http://www.hotshare.net/en/file/92795-2898639442.html

- or -
GTA4 trailer 720p (mpeg4, 19MB)
http://www.megaupload.com/?d=5EH4GH4Z
You can convert it to mpeg2 by mencoder.
$ mencoder 1280x720.mp4 -o new.mpg \
-oac copy \
-ovc lavc -lavcopts vcodec=mpeg2video:vbitrate=3200

I played "Hilary Duff - With love 720p".
$ mplayer -vo vdpau -vc ffmpeg12vdpau Hilary_Duff-With_Love-HD_720p\ -279MB.mpg
The CPU usage is higher than 1080p MGS4 trailer playback (532MB).
http://www.4gamer.net/games/048/G004886/20080815040/

Quantumstate · 06-01-09, 02:24 PM

I suggest that it's never a good idea to do any transcoding for a purpose such as this. The testing ought to be done with the TS just exactly as it comes from the provider, as I am (slowly) uploading.

I notice in my tests above that 185.18.10 decodes faster than 185.18.14, but it deinterlaces slower. I'd rather have faster decoding. Is powermizer fixed in .10?

crisalide · 06-01-09, 02:50 PM

A few seconds is all i need (the test will use 25 pictures)

**Stephen Warren** · 06-01-09, 03:27 PM

No VDPAU PowerMizer issues were fixed in 185.18.10

sl1pkn07 · 06-01-09, 04:27 PM

Quote:

Originally Posted by sl1pkn07

Code:

SURFACE GET BITS: 825.522 M/s
SURFACE PUT BITS: 870.204 M/s

MPEG DECODING (1920x1080): 76 pic/s
VC1 DECODING (1440x1080): 84 pic/s

MIXER WEAVE (1920x1080): 237 frames/s
MIXER BOB (1920x1080): 373 fields/s
MIXER TEMPORAL (1920x1080): 74 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 23 fields/s

8400GS (G98) + Kubuntu 64bits KDE 4.2.3 Without Effects + 185.18.10

Code:

SURFACE GET BITS: 738.163 M/s
SURFACE PUT BITS: 868.294 M/s

MPEG DECODING (1920x1080): 75 pic/s
H264 DECODING (1920x1080): 66 pic/s
VC1 DECODING (1440x1080): 84 pic/s

MIXER WEAVE (1920x1080): 244 frames/s
MIXER BOB (1920x1080): 392 fields/s
MIXER TEMPORAL (1920x1080): 74 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 24 fields/s

8400GS (G98) + Kubuntu 64bits KDE 4.2.3 Without Effects + 185.18.14

the difference is not relevant...

06-01-09, 11:15 AM	#54
crisalide Registered User Join Date: Dec 2008 Posts: 173	Re: VDPAU testing tool @Quantumstate If you can provide a 720p MPEG2 sample, i would add 720p mpeg decoding test. (never seen any until now)

06-01-09, 12:48 PM	#55
Quantumstate Registered User Join Date: May 2008 Posts: 166	Re: VDPAU testing tool Uploading now, but my ISP has upload throttled to 40Kbs, so this 194MB file will take 6 hours. I would fire Queerwire, except they are so cheap and download is 1.5Mbps.

06-01-09, 01:56 PM	#56
garyl Registered User Join Date: May 2009 Posts: 24	Re: VDPAU testing tool Hilary Duff - With love 720p (mpeg2, 280MB) http://www.hotshare.net/en/file/92795-2898639442.html - or - GTA4 trailer 720p (mpeg4, 19MB) http://www.megaupload.com/?d=5EH4GH4Z You can convert it to mpeg2 by mencoder. $ mencoder 1280x720.mp4 -o new.mpg \ -oac copy \ -ovc lavc -lavcopts vcodec=mpeg2video:vbitrate=3200 I played "Hilary Duff - With love 720p". $ mplayer -vo vdpau -vc ffmpeg12vdpau Hilary_Duff-With_Love-HD_720p\ -279MB.mpg The CPU usage is higher than 1080p MGS4 trailer playback (532MB). http://www.4gamer.net/games/048/G004886/20080815040/

06-01-09, 02:24 PM	#57
Quantumstate Registered User Join Date: May 2008 Posts: 166	Re: VDPAU testing tool I suggest that it's never a good idea to do any transcoding for a purpose such as this. The testing ought to be done with the TS just exactly as it comes from the provider, as I am (slowly) uploading. I notice in my tests above that 185.18.10 decodes faster than 185.18.14, but it deinterlaces slower. I'd rather have faster decoding. Is powermizer fixed in .10?

06-01-09, 02:50 PM	#58
crisalide Registered User Join Date: Dec 2008 Posts: 173	Re: VDPAU testing tool A few seconds is all i need (the test will use 25 pictures)

06-01-09, 03:27 PM	#59
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: VDPAU testing tool No VDPAU PowerMizer issues were fixed in 185.18.10

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Valve to release its trailer-making tool: Source Filmmaker	dikinher	Gaming Central	2	06-28-12 12:28 PM
xorg locks-up with newest nvidia drivers w/ vdpau.	theroot	NVIDIA Linux	1	06-24-12 11:04 AM
Load Testing with the Cloud ' This Week on inside* Publications	News	Latest Tech And Game Headlines	0	05-20-12 05:30 PM
Portal 2 Perpetual Testing Initiative Starts Now	News	Latest Tech And Game Headlines	0	05-08-12 05:00 PM
tv tool 6.8	thcdru2k	NVIDIA Windows Graphics Drivers	4	10-18-02 01:10 PM