Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 08-18-11, 10:48 AM   #37
danix
NVIDIA Corporation
 
danix's Avatar
 
Join Date: Feb 2010
Location: Santa Clara, CA
Posts: 237
Default Re: GF100 VDPAU performance is really not good

Let's try to clear up some confusion here:

The original bug that was reported in this thread ("GF100 VDPAU performance is really not good") was reproduced at NVIDIA using the same hardware (HP EliteBook 8740w) as the original reporter (camelreef). After some investigation, it was determined that the reason PowerMizer was unable to enter the highest performance level was because of an interaction between the DSDT in HP's BIOS and the Linux kernel's ACPI subsystem; in fact, the exact same issue which is responsible for a blank screen on EliteBook 8x40w systems that have the optional DreamColor eDP panel installed [1].

camelreef, sorry for not communicating this back to you sooner. The thread about the blank screen issue contains details on how to patch your DSDT to enable correct functionality. HP sampled us a BIOS with which we were able to verify that both issues (the blank screen and the low performance) were resolved, but it appears that they have not yet made a public release containing the needed changes.

For those of you who are experiencing similar issues on other systems than HP 8x40w series notebooks: if you are using a notebook system or other system with the GPU on an MXM, please check your DSDT for an issue similar to what is described in the thread about the blank screens. If you discover a similar issue, please check to see if patching your DSDT resolves it. (Note: on the HP 8x40w notebooks it is sufficient to change the order of the two device definitions; on others, this may not be possible, and the workaround may not be as trivial.)

If you are on a non-MXM system, or an MXM system without aliased device definitions or on which a patched DSDT does not resolve your issue, please start a separate thread and include details sufficient to reproduce your reported problem. (See the stickies on this forum.)

[1] http://www.nvnews.net/vbulletin/showthread.php?t=159544
danix is offline   Reply With Quote
Old 08-26-11, 11:10 AM   #38
camelreef
Registered User
 
Join Date: Nov 2006
Posts: 27
Default Re: GF100 VDPAU performance is really not good

Hello Danix,

Back from vacation, and I will test this tonight!

Now I need to chase HP, I guess...

Thanks!

Nico
camelreef is offline   Reply With Quote
Old 08-26-11, 03:00 PM   #39
camelreef
Registered User
 
Join Date: Nov 2006
Posts: 27
Default Re: GF100 VDPAU performance is really not good

oh, well, a DSDT tweak, a kernel recompile and a reboot later:

Code:
qvdpautest 0.5.1
Intel(R) Core(TM) i7 CPU       Q 840  @ 1.87GHz
NVIDIA GPU Quadro 5000M (GF100GL) at PCI:1:0:0 (GPU-0)

VDPAU API version : 1
VDPAU implementation : NVIDIA VDPAU Driver Shared Library  280.13  Wed Jul 27 17:15:20 PDT 2011

SURFACE GET BITS: 1099.51 M/s
SURFACE PUT BITS: 1139.4 M/s

MPEG DECODING (1920x1080): 104 frames/s
MPEG DECODING (1280x720): 232 frames/s
H264 DECODING (1920x1080): 63 frames/s
H264 DECODING (1280x720): 131 frames/s
VC1 DECODING (1440x1080): 82 frames/s
MPEG4 DECODING (1920x1080): 71 frames/s

MIXER WEAVE (1920x1080): 123 frames/s
MIXER BOB (1920x1080): 171 fields/s
MIXER TEMPORAL (1920x1080): 76 fields/s
MIXER TEMPORAL + IVTC (1920x1080): 64 fields/s
MIXER TEMPORAL + SKIP_CHROMA (1920x1080): 93 fields/s
MIXER TEMPORAL_SPATIAL (1920x1080): 39 fields/s
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080): 33 fields/s
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080): 43 fields/s
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display): 107 fields/s
MIXER TEMPORAL_SPATIAL + HQSCALING (720x576 video to 1920x1080 display): 71 fields/s

MULTITHREADED MPEG DECODING (1920x1080): 95 frames/s
MULTITHREADED MIXER TEMPORAL (1920x1080): 72 fields/s

Woot! More than enough for my needs!

And Powermizer goes to full perfs too.

Thanks for pointing this out.

I a really going to bug HP on this BIOS issue now!

Nico

Last edited by camelreef; 08-26-11 at 03:03 PM. Reason: code tags
camelreef is offline   Reply With Quote
Old 08-29-11, 09:12 AM   #40
camelreef
Registered User
 
Join Date: Nov 2006
Posts: 27
Default Re: GF100 VDPAU performance is really not good

hmmm, in fact this one is on the low side for my use (1080i50), i was expecting better results from such a card!

MIXER TEMPORAL_SPATIAL (1920x1080): 39 fields/s

Is this normal?




just for fun, here are the improvements introduced by the DSDT fix:

Code:
Function(s)	Before	After	Unit	Improvement
SURFACE GET BITS	673.407	1099.51	M/s	63%
SURFACE PUT BITS	748.643	1139.4	M/s	52%
				
MPEG DECODING (1920x1080)	78	104	frames/s	33%
MPEG DECODING (1280x720)	175	232	frames/s	33%
H264 DECODING (1920x1080)	44	63	frames/s	43%
H264 DECODING (1280x720)	88	131	frames/s	49%
VC1 DECODING (1440x1080)	62	82	frames/s	32%
MPEG4 DECODING (1920x1080)	53	71	frames/s	34%
				
MIXER WEAVE (1920x1080)	24	123	frames/s	413%
MIXER BOB (1920x1080)	40	171	fields/s	328%
MIXER TEMPORAL (1920x1080)	10	76	fields/s	660%
MIXER TEMPORAL + IVTC (1920x1080)	11	64	fields/s	482%
MIXER TEMPORAL + SKIP_CHROMA (1920x1080)	19	93	fields/s	389%
MIXER TEMPORAL_SPATIAL (1920x1080)	7	39	fields/s	457%
MIXER TEMPORAL_SPATIAL + IVTC (1920x1080)	5	33	fields/s	560%
MIXER TEMPORAL_SPATIAL + SKIP_CHROMA (1920x1080)	8	43	fields/s	438%
MIXER TEMPORAL_SPATIAL (720x576 video to 1920x1080 display)	19	107	fields/s	463%
MIXER TEMPORAL_SPATIAL + HQSCALING (720x576 video to 1920x1080 display)	12	71	fields/s	492%
				
MULTITHREADED MPEG DECODING (1920x1080)	22	95	frames/s	332%
MULTITHREADED MIXER TEMPORAL (1920x1080)	13	72	fields/s	454%
camelreef is offline   Reply With Quote
Old 09-10-11, 08:03 AM   #41
TheSwissKnife
Registered User
 
Join Date: Sep 2010
Posts: 10
Default Re: GF100 VDPAU performance is really not good

Well I assume a Shuttle XS35GT with ION NG is not MXM. It seems that each and every system must then be handled as a separate thread? Can it at least be confirmed there is no way in general to force a performance level when the interaction with the system is preventing powermizer from working? If this is the case then an explanation of why would be very helpful in this thread and given its title - it is one thing for powermizer to not work but another to prevent the card from operating at max performance level.
TheSwissKnife is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 06:20 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.