vdpau slowness with 260.19.06 - Page 8

lexa2 · 10-17-11, 06:17 PM

285.05.09

Code:

qvdpautest 0.5.1
AMD Phenom(tm) II X4 955 Processor
NVIDIA GPU GeForce GTX 550 Ti (GF116) at PCI:2:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 19:17:37 PDT 2011

SURFACE GET BITS: 1252.72 M/s
SURFACE PUT BITS: 1180.51 M/s

MPEG DECODING (1920x1080): 107 frames/s
MPEG DECODING (1280x720): 239 frames/s
H264 DECODING (1920x1080): 67 frames/s
H264 DECODING (1280x720): 141 frames/s
VC1 DECODING (1440x1080): 83 frames/s
MPEG4 DECODING (1920x1080): 72 frames/s

MIXER WEAVE (1920x1080): 377 frames/s
MIXER BOB (1920x1080): 522 fields/s
MIXER TEMPORAL (1920x1080): 251 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 216 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 310 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 130 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 117 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 144 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 371 fields/s
MIXER TEMPORAL_SPATIAL + HQSCALING (720x576 video to 1920x1080 display): 240 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 107 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 237 fields/s

RobertMassaioli · 10-18-11, 01:45 AM

Reading through this thread it seems like the exact same problem that I am getting. Except that I keep calling the problem "High X11 CPU Usage". The issue that I notice the most is that X11 just goes to pot and the entire system becomes laggy. However, I also have significantly slower VDPAU playback too.

My system, when it starts always starts just fine: no performance issues. But, after some random amount of time (it is never consistent) the system will start performing horribly. And just take a look at the difference.

I do not really know what to make of this but I am attempting to solve the problem as best as I can and I am keeping my thoughts and musings in this Google Doc which I just made publicly viewable.

I suspect that all of the issues that say that "X11 has high CPU usage" will also be linked to the "VDPAU performing poorly" issues. Please let me know what I can do to help. I have already provided nvidia bug reports but I can do so again. In-fact next time I might attach them to my Google Doc instead of these forums. (We should have a public bug tracker to keep track of this; did somebody say Jira?)

weltall · 10-18-11, 02:17 AM

In my case at least vdpau misbehaves always. just after a reboot, just after an x restart, after hours or days. it's years i don't see it behaving well. But I've noticed the X cpu use happening from time to time like 5 times/year or something like that

Blackkitten · 10-19-11, 06:17 AM

Hello again !
I started from advice about "un-redirect full screen" that I was given on ubuntuforums.
I didn't have it ticked. When I selected it the vdpau perfomance became smooth ( tested on video from Nikon D7000 that is 1080p@24fps ). Then I stared to measure perfomance with qvdpau testing tool... In the end I dicovered next things:

1. It looks like that "un-redirect full screen" option doesn't affect perfomance.

2. Vdpau perfomance is switching from smooth to bad without any noticable causes. Usually it is smooth after restart or changing the session, but after some time it becomes too slow.
Usually Unity2d session has smoother results, than Unity 3d, but in general they are on par. I have next qvdpautest results:

Unity2d session. Smooth VDPAU.

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7500  @ 2.20GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 821.865 M/s
SURFACE PUT BITS: 737.28 M/s

MPEG DECODING (1920x1080): 57 frames/s
MPEG DECODING (1280x720): 124 frames/s
H264 DECODING (1920x1080): 25 frames/s
H264 DECODING (1280x720): 57 frames/s
VC1 DECODING (1440x1080): 72 frames/s

MIXER WEAVE (1920x1080): 531 frames/s
MIXER BOB (1920x1080): 894 fields/s
MIXER TEMPORAL (1920x1080): 204 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 135 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 277 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 70 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 58 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 78 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 270 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 60 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 172 fields/s

Unity3d session. Smooth vdpau.

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7500  @ 2.20GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 812.373 M/s
SURFACE PUT BITS: 684.472 M/s

MPEG DECODING (1920x1080): 57 frames/s
MPEG DECODING (1280x720): 123 frames/s
H264 DECODING (1920x1080): 24 frames/s
H264 DECODING (1280x720): 53 frames/s
VC1 DECODING (1440x1080): 72 frames/s

MIXER WEAVE (1920x1080): 529 frames/s
MIXER BOB (1920x1080): 672 fields/s
MIXER TEMPORAL (1920x1080): 152 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 99 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 208 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 54 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 44 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 63 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 202 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 61 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 171 fields/s

Unity3d session. Bad vdpau perfomance.

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7500  @ 2.20GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 816.466 M/s
SURFACE PUT BITS: 630.061 M/s

MPEG DECODING (1920x1080): 34 frames/s
MPEG DECODING (1280x720): 74 frames/s
H264 DECODING (1920x1080): 15 frames/s
H264 DECODING (1280x720): 33 frames/s
VC1 DECODING (1440x1080): 42 frames/s

MIXER WEAVE (1920x1080): 523 frames/s
MIXER BOB (1920x1080): 671 fields/s
MIXER TEMPORAL (1920x1080): 151 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 98 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 207 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 54 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 43 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 63 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 202 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 38 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 174 fields/s

3. The funny thing.
The results above was take using notebook's display: 1280x800.
Switching to external display(1920x1080) gives better vdpau perfomance!

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7500  @ 2.20GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 825.82 M/s
SURFACE PUT BITS: 742.96 M/s

MPEG DECODING (1920x1080): 71 frames/s
MPEG DECODING (1280x720): 143 frames/s
H264 DECODING (1920x1080): 41 frames/s
H264 DECODING (1280x720): 86 frames/s
VC1 DECODING (1440x1080): 105 frames/s

MIXER WEAVE (1920x1080): 492 frames/s
MIXER BOB (1920x1080): 836 fields/s
MIXER TEMPORAL (1920x1080): 193 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 129 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 262 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 68 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 56 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 76 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 263 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 67 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 161 fields/s

P.S. I'm using mplayer from this PPA.
And nvidia-drivers from this PPA.

weltall · 10-26-11, 04:04 AM

290.03... still the same soup

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7300  @ 2.00GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  290.03  Mon Oct 17 10:57:02 PDT 2011

SURFACE GET BITS: 720.83 M/s
SURFACE PUT BITS: 688.561 M/s

MPEG DECODING (1920x1080): 35 frames/s
MPEG DECODING (1280x720): 76 frames/s
H264 DECODING (1920x1080): 15 frames/s
H264 DECODING (1280x720): 33 frames/s
VC1 DECODING (1440x1080): 44 frames/s

MIXER WEAVE (1920x1080): 513 frames/s
MIXER BOB (1920x1080): 865 fields/s
MIXER TEMPORAL (1920x1080): 199 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 131 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 270 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 69 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 56 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 77 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 264 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 38 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 171 fields/s

MadCatX · 10-26-11, 12:51 PM

Here is something you might try.

1st run, PowerMizer set to "Adaptive"

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T9550  @ 2.66GHz
NVIDIA GPU GeForce 9800M GTS (G94) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 769.197 M/s
SURFACE PUT BITS: 1011.51 M/s

MPEG DECODING (1920x1080): 51 frames/s
MPEG DECODING (1280x720): 102 frames/s
H264 DECODING (1920x1080): 29 frames/s
H264 DECODING (1280x720): 64 frames/s
VC1 DECODING (1440x1080): 83 frames/s

MIXER WEAVE (1920x1080): 1800 frames/s
MIXER BOB (1920x1080): 3158 fields/s
MIXER TEMPORAL (1920x1080): 730 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 424 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 983 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 225 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 182 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 245 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 833 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 37 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 337 fields/s

I've seen worse results, but my card can do better

2nd run, PowerMizer set to "Max performance" few seconds _after_ the test started

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T9550  @ 2.66GHz
NVIDIA GPU GeForce 9800M GTS (G94) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 970.785 M/s
SURFACE PUT BITS: 1011.1 M/s

MPEG DECODING (1920x1080): 69 frames/s
MPEG DECODING (1280x720): 137 frames/s
H264 DECODING (1920x1080): 40 frames/s
H264 DECODING (1280x720): 87 frames/s
VC1 DECODING (1440x1080): 109 frames/s

MIXER WEAVE (1920x1080): 1909 frames/s
MIXER BOB (1920x1080): 2983 fields/s
MIXER TEMPORAL (1920x1080): 724 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 424 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 983 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 225 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 182 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 245 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 833 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 38 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 346 fields/s

Card ran at performance level 3 all the time during the test.

3rd run, PowerMizer set back to "Adaptive" few seconds _after_ the test started

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T9550  @ 2.66GHz
NVIDIA GPU GeForce 9800M GTS (G94) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  285.05.09  Fri Sep 23 17:55:14 PDT 2011

SURFACE GET BITS: 957.537 M/s
SURFACE PUT BITS: 1008.73 M/s

MPEG DECODING (1920x1080): 69 frames/s
MPEG DECODING (1280x720): 132 frames/s
H264 DECODING (1920x1080): 39 frames/s
H264 DECODING (1280x720): 84 frames/s
VC1 DECODING (1440x1080): 108 frames/s

MIXER WEAVE (1920x1080): 1709 frames/s
MIXER BOB (1920x1080): 3160 fields/s
MIXER TEMPORAL (1920x1080): 731 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 423 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 983 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 225 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 181 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 244 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 833 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 45 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 413 fields/s

As you can see, there's a noticeable performance boost, it even beats the "Max performace" mode. The card was switching between performance levels 2 and 3 over the course of the test.
It's not only the qvdpautest where I can observe this odd behavior, Full HD movies with high bitrate are quite unwatchable in mplayer unless I do this adaptive->max->adaptive toggle.

Can you replicate this on your machine?

The test ran on Arch Linux 64bit, kernel 3.0.7 with CK patches, KDE 4.7.2 with compositing on, Xorg 11.1.901.

weltall · 10-27-11, 02:25 AM

unfortunately in my case nothing changes by switching that option my card also stays on 2 whathever i do (0-2 are the available levels on this card)
now I'm also using 3.1 kernel.

Blackkitten · 10-29-11, 05:23 PM

Quote:

Originally Posted by weltall

unfortunately in my case nothing changes by switching that option my card also stays on 2 whathever i do (0-2 are the available levels on this card)
now I'm also using 3.1 kernel.

Same thing for me. But if set powermizer to max.perf. right after restart/boot/waking up from sleep mode then vdpau perfomance is always good and it doesn't degrade as I've reported earlier.

It looks like powermizer is the cause: if it goes to "0" mode perfomance then vdpau perfomance also goes down no matter what you do (even setting poewrmizer to max.perf doesn't help) until restart.

weltall · 11-09-11, 10:15 AM

new driver and old problems
linux 3.2-rc1 + nvidia 290.06 no changes at all still slow and unusable.
As an additional thing if i switch monitor config while vdpau is running the vdpau context is lost. (but in this case it's the first time ever I've tried to do such a thing)

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7300  @ 2.00GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  290.06  Mon Oct 31 08:48:41 PDT 2011

SURFACE GET BITS: 593.024 M/s
SURFACE PUT BITS: 574.048 M/s

MPEG DECODING (1920x1080): 39 frames/s
MPEG DECODING (1280x720): 85 frames/s
H264 DECODING (1920x1080): 16 frames/s
H264 DECODING (1280x720): 35 frames/s
VC1 DECODING (1440x1080): 48 frames/s

MIXER WEAVE (1920x1080): 445 frames/s
MIXER BOB (1920x1080): 759 fields/s
MIXER TEMPORAL (1920x1080): 180 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 124 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 238 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 64 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 54 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 72 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 250 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 38 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 151 fields/s

weltall · 03-26-12, 12:58 PM

seems it finally got better:

Code:

qvdpautest 0.5.1
Intel(R) Core(TM)2 Duo CPU     T7300  @ 2.00GHz
NVIDIA GPU GeForce 8600M GT (G84) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  295.33  Sat Mar 17 15:20:42 PDT 2012

SURFACE GET BITS: 685.673 M/s
SURFACE PUT BITS: 690.281 M/s

MPEG DECODING (1920x1080): 72 frames/s
MPEG DECODING (1280x720): 144 frames/s
H264 DECODING (1920x1080): 40 frames/s
H264 DECODING (1280x720): 82 frames/s
VC1 DECODING (1440x1080): 108 frames/s
Profile unsupported.

MIXER WEAVE (1920x1080): 518 frames/s
MIXER BOB (1920x1080): 871 fields/s
MIXER TEMPORAL (1920x1080): 200 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 134 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 272 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 69 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 57 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 76 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 266 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 67 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 165 fields/s

10-18-11, 01:45 AM	#86
RobertMassaioli Developer Join Date: Sep 2011 Location: Australia Posts: 5	Re: vdpau slowness with 260.19.06 Reading through this thread it seems like the exact same problem that I am getting. Except that I keep calling the problem "High X11 CPU Usage". The issue that I notice the most is that X11 just goes to pot and the entire system becomes laggy. However, I also have significantly slower VDPAU playback too. My system, when it starts always starts just fine: no performance issues. But, after some random amount of time (it is never consistent) the system will start performing horribly. And just take a look at the difference. qvdpautest results when the system is misbehaving qvdpautest results when the system is performing well I do not really know what to make of this but I am attempting to solve the problem as best as I can and I am keeping my thoughts and musings in this Google Doc which I just made publicly viewable. I suspect that all of the issues that say that "X11 has high CPU usage" will also be linked to the "VDPAU performing poorly" issues. Please let me know what I can do to help. I have already provided nvidia bug reports but I can do so again. In-fact next time I might attach them to my Google Doc instead of these forums. (We should have a public bug tracker to keep track of this; did somebody say Jira?)

10-18-11, 02:17 AM	#87
weltall Registered User Join Date: Feb 2007 Posts: 115	Re: vdpau slowness with 260.19.06 In my case at least vdpau misbehaves always. just after a reboot, just after an x restart, after hours or days. it's years i don't see it behaving well. But I've noticed the X cpu use happening from time to time like 5 times/year or something like that

10-27-11, 02:25 AM	#91
weltall Registered User Join Date: Feb 2007 Posts: 115	Re: vdpau slowness with 260.19.06 unfortunately in my case nothing changes by switching that option my card also stays on 2 whathever i do (0-2 are the available levels on this card) now I'm also using 3.1 kernel.