Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
3rd June 2013, 16:41 | #18964 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
You don't have to use it if you don't need it. I'm just wondering why your composition rate seemingly changes, depending on which video you're playing. You do mean the composition rate reported by the madVR OSD (Ctrl+J), right? This composition rate usually only changes if the display mode is changed. So how can it change in your case, if you don't use any refresh rate / display mode changer?
|
3rd June 2013, 17:02 | #18965 | Link | ||
black stain
Join Date: Apr 2013
Posts: 46
|
Quote:
Quote:
whats more interesting, after starting f.lux again the same movies composition rate now stays 60.000 (after restarting the player of course) |
||
3rd June 2013, 19:38 | #18967 | Link |
Registered User
Join Date: Sep 2012
Location: Turin
Posts: 104
|
Here is a test with Haswell + MadVR + LAV Filters
http://www.anandtech.com/show/7007/i...pc-perspective
__________________
Raven RVZ01 * i7-4790k * 16GB RAM * Zotac GTX 970 4G * SSD 850Evo 500GB * Blu-Ray Burner Slot-In * PSU SFX 80+ Gold 450Watt * Windows 10 64bit * MPCHC+MadVR+SVP * Panasonic 50" VT30 ^^ |
3rd June 2013, 20:00 | #18968 | Link |
Registered User
Join Date: May 2013
Posts: 10
|
More configurable CPU queue size
Hi madshi. Really appreciate your work on this unique piece of software. I'm using madVR for everything I watch.
Recently I've hit the CPU limit on some 1080p video. I've figured out the exact reason and I'd like to request one small change in madVR config. On like 5 highly detailed fragments (with duration ~15 seconds each) of the entire video, I have stuttering and AV desync (100% CPU usage, decoder queue slowly becomes empty -> hiccups), then everything gets back to normal (60-70% CPU usage, full decoder queue). The bottleneck side-effects appeared even if using EVR renderer, so its the decoder/cpu speed issue. However I've noticed that madVR can pre-buffer decoded frames in advance, which can help my PC to distribute CPU load between heavy and light parts of the video, but the 32 queue size is not enough to achieve this (at least for this video), so my question/feature request is: Can you increase the configurable CPU queue size to smth like 128 , so I could use all my RAM for buffering ? I tried to hack the madVR registry config values to set 64 queue size, so I could see how that would work out for me, but that didnt work - when I ran MPC-HC the queue size became 12 instead. In case you don't want to mess up the UI with overTICKed trackbar, then perhaps can you allow manual registry config value changes outside of the UI range to be properly processed by madVR ? Or maybe some numeric box under the trackbar for "above 32" values (could be made as a trackbar value multiplier) ? My specs are: Athlon II X2 215 ~2.7GHz; NVIDIA GeForce 8800 GTS (320 MB); 3 GB RAM Win XP SP3; 1920x1080@72Hz MPC-HC(v1.6.7.7091); LAV(v0.57)-Splitter/Video/Audio; madVR(v0.86.1); ReClock; Video material: 1860x1048 @ 23.976 fps (1080p x264 Hi10p) madVR rendering settings: CPU queue: 32 GPU queue: 8 Setting CPU queue size to 4 doesn't help - it plays the fragment slower/longer with frequent but small frame drops and I think the video looks less stuttering because of that, but the AV desync is still there - even a bit worse than with 32 queue size. |
3rd June 2013, 20:46 | #18969 | Link | |
Registered User
Join Date: Nov 2012
Posts: 138
|
Quote:
if you have 100% cpu usage its going to drop frames no matter what settings you use, sounds like you need a faster cpu to play that video |
|
3rd June 2013, 21:02 | #18970 | Link | |
Registered User
Join Date: May 2013
Posts: 10
|
Quote:
I love my PC because its properly configured and works like a clock without glitches. This is the only video so far that I got unsolvable problems with (I've been watching many 1080p videos just fine) and I'd rather try out some options such as these before thinking about selling the case . Last edited by Danat; 3rd June 2013 at 21:07. |
|
3rd June 2013, 21:08 | #18971 | Link |
Registered User
Join Date: Mar 2010
Posts: 14
|
Anyone can help me with ringing artifact issue? I have downgraded Intel HD4000 driver to 29XX in order to use Madvr but I'm experiencing some ringing. Tried a lot of 1080p mkv's played on a 1080p LCD therefore I can exclude resize related issues.
|
3rd June 2013, 21:35 | #18972 | Link | ||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Argh...
Quote:
Nice - good to see Haswell finally gets the refresh rates right! Quote:
I'm not sure I understand what you mean. Do you think that your playback chain somehow adds ringing artifacts to the movie which aren't in the original movie source? Can you show me a screenshot of the exact artifact you mean? |
||
3rd June 2013, 22:49 | #18973 | Link |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Not that that 23.97605 Hz really helps much when the clock deviation is so high at 0.00472% with a dropped frame every 14.84 minutes like their screenshot. The refresh rate of 23.97605 is actually very far from optimal on Ananatech's ASRock Z87E-ITX Haswell testbed in this instance, and still means using something like Reclock is required. I wish review sites like Anandtech would do more in-depth testing on the clock deviation with various hardware setups to find the best matches at standard refresh rates.
IMHO, the optimal setup is one which madVR reports drop/repeat every 3-7+ days when not using Reclock (or similar) with default GPU refresh rate settings. Matching the refresh rate to clock deviance is pretty much the only way to achieve perfect sync without degrading quality in some way, but unfortunately many GPU drivers are too inflexible to lock in fine-tuned refresh rates. This is the main thing going for using madVR with a high refresh rate multiple instead of a matching refresh rate if you seek to fine-tune refresh rate around stubborn drivers. Assuming the driver allow refresh rate timing tweaks of some kind, it becomes exponentially easier to lock in a clock deviance matched refresh rate each time you increase the refresh rate multiple. If you do implement a high CPU queue like 128, you may want to consider adding an additional slider for the Subtitle queue and/or have it match the GPU queue size instead. Setting the Subtitle queue smaller than than CPU queue I would actually expect would improve reliability in cases of CPU bottlenecks, similar to how setting the CPU queue slightly higher than the GPU queue does. Last edited by cyberbeing; 3rd June 2013 at 23:02. |
3rd June 2013, 23:02 | #18975 | Link |
Registered Developer
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,346
|
Usually precision issues with the clock generators, or inflexibility issues. It may be designed for 23.9760 specifically, but that doesn't mean it can do 23.977 to compensate for audio deviation.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders |
3rd June 2013, 23:19 | #18976 | Link |
Registered User
Join Date: Jan 2008
Posts: 589
|
Wait. If I understand this correctly, if you use the same HDMI port for audio *and* video, they should be governed by the same clock (since, AFAIK, in HDMI audio is transferred during the blanking interval between two frames). If that's the case, why is clock deviation still an issue? Shouldn't they be perfectly in sync since it's the *same* clock? Or are we only discussing the case where audio is transported separately (e.g. analog)?
|
3rd June 2013, 23:20 | #18977 | Link |
Registered User
Join Date: Dec 2008
Posts: 18
|
I did some more viewing with Smooth Motion rendering on tonight - watching snippets of films I've watched over and over again (you know those AV demo worthy parts that are worth watching over and over again )
what has really shocked me is how with Smooth Motion on - I'm noticing details I've never noticed before - when I say details I don't mean as in it looks sharper, or there is more detail on screen so to speak, but just that there seems to be so much more content in any shot with motion genuinely leagues better - ok there is the odd very minor additional artefact - but perfectly acceptable I'm not normally one for "post-processing" - but whatever Smooth Motion is doing it works for my setup - so thanks ! I'm comparing 24p from MadVR as 24p input into TV (my Panasonic plays 24p at 23.976hz which is quite unusual) - you get noticeable flicker in Windows Desktop - but no obvious flicker whilst watching films to 60p from MadVR as a 60p input into TV in theory the Panasonic playing 24p content with near to no processing without Smooth Vision should be the ideal setup - but with Smooth Vision on its leagues better |
4th June 2013, 00:23 | #18978 | Link | |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Quote:
I have no idea what you consider as "video clock", but the "audio clock" is what DirectShow uses as a reference clock as interpreted by the "system clock". The clock deviance shown by madVR, multiplied by and added to the video frame rate is the "real" playback rate in DirectShow before anything ever gets passed over HDMI or other output. For example, in Anandtech's screenshot with 0.00472% clock deviance, the "real" video & audio playback rate is equivalent to ~23.97715 fps, which is higher than his display refresh rate of ~23.97605 Hz. I've found madVR's "clock deviation" + "drop/repeat every" measurements to be highly accurate on my systems, but ultimately you should trust your eyes in combination with the number of actual frame drops/repeats/glitches madVR records in the OSD stats. At a certain point, if you are unable to perceive or detect any anomalies, its good enough. Last edited by cyberbeing; 4th June 2013 at 00:44. |
|
4th June 2013, 16:23 | #18979 | Link |
Registered User
Join Date: Apr 2013
Posts: 6
|
Hi, I seem to be having a slight problem. Usually on playback I get no dropped/delayed frames at all, but sometimes I can watch the exact same video again (that I've previously had no dropped/delayed frames) on a different day and receive a significant amount of delayed frames (usually ranging from around 70-130 I've noticed). Does anybody know why this is?
If it helps, here are my specs: Samsung NPC700G7C, 16.0GB RAM, Intel Core 17 3610QM @ 2.30GHz, GTX 675M I'm also using Niyawa's Guide and use it exclusively for anime playback. I'm also using his 'highest settings' for the scaling algorithm's. If anyone could help that would be great! |
4th June 2013, 18:46 | #18980 | Link | |||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Quote:
Quote:
(1) Make sure you disable tools/applications which might eat GPU performance. E.g. newer Firefox/Chrome/IE versions might use the GPU for rendering tasks. Or recently it has been reported that the "f.lux" and "GpuZ" tools can also result in video playback problems. So close them before watching a movie. (2) Make sure your GPU doesn't clock down. This has happened to some users. I think some tweak tools allow to fix the clock to some specific value. If these hints don't help, make a screenshot of the madVR OSD (Ctrl+J) in the moment when those frame drops occur. If you can't make a screenshot (FSE mode?), please at least write down the state of the queues and report them here. Then we can go from there... |
|||||
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
Thread Tools | Search this Thread |
Display Modes | |
|
|