View Full Version : madVR - high quality video renderer (GPU assisted)
sneaker_ger
27th April 2011, 16:24
How about glitches? Still there or gone for you with 0.59 / D3D11?
Never had any glitch problems with my HD 5850 to begin with, doesn't matter if with/without hack, D3D9 or D3D11. But I will keep my eyes open.
madshi
27th April 2011, 16:27
madVR v0.60 released
http://madshi.net/madVR.zip
* fixed: exclusive -> windowed with D3D11 made media player buttons disappear
* fixed: fullscreen exclusive mode was sometimes not activated properly
* Windows Key press now forces a 3 second delay before going back to exclusive
sneaker_ger
27th April 2011, 16:31
* fixed: exclusive -> windowed with D3D11 made media player buttons disappear
Not fixed for me.
Hprd
27th April 2011, 16:42
On my PC I need to have both new tweak options on or both off to get no glitches. If I have only the "run presentation in a separate thread" option on, then I get glitches. So let me ask you again: Please turn *BOTH* new tweak options off and test again. That's important for me because I actually want to delete the 2 new tweak options and treat them as always unchecked.
I tested several times b4 i made my first post, and several on my last one (and this one as well just to really make sure as per your suggestion), i can reproduce it nearly 100% of the time. With only the limit option on though, i did get 1 glitch on a short test (i guess would incidate the other option does do something, but not a whole lot), compared to literally zero with both on (playing quite a bit more). Both off, or run in another thread were pretty much the same (although, perhaps it is a little bit more resilient with the seperate thread option on, so i really shouldn't say it does nothing if it's the only thing checked. Could just be a coincidence on one test though...).
Also, the present que with only the "limit" option on goes up and down from 5-6/10 to 5-7/10, as compared to being pretty much locked at 5-6/10 with both on, and being 6-7/10 with only the "thread" option or both off. And, I also noticed that the render que (only when both options are checked) fluctuates from 7-8/8 to 6-8/8 then back again, seems to have no ill effects though. Not sure though if that's anything unusual at all anyways...
Hmm I'd hate if these options were removed, as they seem to work perfectly for me. Maybe my computer is just weird though from some reason, lol :o
EDIT: i see v60 just came out, i'll retest and see if anything's changed, lol.
iSunrise
27th April 2011, 16:46
Funny. On my NVidia 9400 HTPC it's exactly the opposite. With 24fps @ 60Hz, I get glitches without D3D11 and no glitches with D3D11 on.
Funny indeed. Just double-checked with restoring everything to the default settings (with the included batch file) just to be sure. I can reproduce it every time. Are you also using 270.61?
Trying 0.60 now.
someone`
27th April 2011, 16:49
With D3D11 enabled , I got crash when auto open next video in FSE mode , test with Potplayer and MPC-HC , madVR 0.60 .
Only crash when D3D11 is enabled and in FSE mode .
madshi
27th April 2011, 16:51
Not fixed for me.
It is for me... :(
I tested several times b4 i made my first post, and several on my last one (and this one as well just to really make sure as per your suggestion), i can reproduce it nearly 100% of the time. With only the limit option on though, i did get 1 glitch on a short test (i guess would incidate the other option does do something, but not a whole lot), compared to literally zero with both on (playing quite a bit more). Both off, or run in another thread were pretty much the same (although, perhaps it is a little bit more resilient with the seperate thread option on, so i really shouldn't say it does nothing if it's the only thing checked. Could just be a coincidence on one test though...).
Ok. Can you do a long test with (1) both options *on* compared to (2) both options *off*? Maybe the next time you watch a full movie you can try with both options on. Then the next movie you watch with both options off. Short time tests often have a bit of random nature. Long time tests are more reliable. Thx.
Funny indeed. Just double-checked with restoring everything to the default settings (with the included batch file) just to be sure. I can reproduce it every time. Are you also using 270.61?
Yep.
With D3D11 enabled , I got crash when auto open next video in FSE mode , test with Potplayer and MPC-HC , madVR 0.60 .
Only crash when D3D11 is enabled and in FSE mode .
Ok, will see if I can reproduce that...
Hprd
27th April 2011, 16:56
Hmm, v60 is the same as 59b, glitches no matter what i do, even though the present and render ques behave exactly the same as 59 with the exact same settings. Also, the exit, minimze etc, buttons still aren't showing up on mph-hc, and it sill goes to "fullscreen" on fse close most of the time :( Also sometimes when i goto fullscreen (did this on 59b a couple times too) the image is just frozen, but i can double click to goto windowed mode and the video is playing there, then if i try to switch back, it'll just have the image there paused in fullscreen... I think i'll go back to 59 for now then.
Also, for the record I'm using the latest 270.61 nvidia drivers (quadro version, not sure if that has anything to do with it specifically).
Oh and on v59, the glitches would start happening right away (like any settings on v59b/60) and would continue at a semi-random interval (or basically it seemed the more dense the video was the more you'd see them) as compared to with the tweaks on, they wouldn't show up at all (after the first couple of seconds anyways when it's switching to FSE), period. I'll put v59 back in and play a 10+ minute video and see for sure (i'll take screenshots to compare) ;)
iSunrise
27th April 2011, 17:07
D3D11 (MPC-HC and Potplayer)
- When switching back from fullscreen to windowed the media player doesnīt return to the native (windowed) resolution of the movie, but instead it tries to display the movie in the fullscreen resolution in windowed mode. This works perfectly fine with the D3D9 path and the old FSE path.
This bug is still there with 0.60. Happens only when the D3D11 path is enabled, all other settings donīt matter.
Budtz
27th April 2011, 17:29
I get an accessviolation error whenever i switch to the next video file in a folder from within mpc-hc. this bug started with 0.59 and is also present in 0.60
Hprd
27th April 2011, 17:41
Well it seems that the glitches don't carry over when the video ends, and it switches to windowed mode (which is where i can capture screenshots currently). So I couldn't take any.
Ok first off i went to videohelp to get v59 again, as v59b/60 of course broke whatever was working with that version. Then, what i did was play a near 13 minute video, put it to fullscreen, let it go for a couple of seconds, hit control+r then let it play through. I found that with both tweaks enabled i got a grand total of, 0 glitches... With both off, i had exactly 190. Now i might retest with both on again to see if it was a fluke that there wasn't any, but i think it's pretty conclusive the differences between them being on and off on this computer, lol.
EDIT: Ok i tried it again with both on, same result as b4 (0).
madshi
27th April 2011, 18:09
This bug is still there with 0.60. Happens only when the D3D11 path is enabled, all other settings donīt matter.
Argh, I can't reproduce that here... :(
I get an accessviolation error whenever i switch to the next video file in a folder from within mpc-hc. this bug started with 0.59 and is also present in 0.60
I'll see if I can reproduce this.
Well it seems that the glitches don't carry over when the video ends, and it switches to windowed mode (which is where i can capture screenshots currently). So I couldn't take any.
Ok first off i went to videohelp to get v59 again, as v59b/60 of course broke whatever was working with that version. Then, what i did was play a near 13 minute video, put it to fullscreen, let it go for a couple of seconds, hit control+r then let it play through. I found that with both tweaks enabled i got a grand total of, 0 glitches... With both off, i had exactly 190. Now i might retest with both on again to see if it was a fluke that there wasn't any, but i think it's pretty conclusive the differences between them being on and off on this computer, lol.
EDIT: Ok i tried it again with both on, same result as b4 (0).
Ok, 190 vs 0 is pretty clear. And with v59b/60 do you get that many glitches, too?
I've checked and there was a bug in v59 which is not in v59b: It eventually used a wrong number of backbuffers. Weird that that actually seemed to have helped in your case. What is your setting for "how many video frames shall be presented in advance"? And in your "windowed mode settings" tab what have you set "how many backbuffers shall be used" to?
dansrfe
27th April 2011, 18:11
Ok so I'm loving the D3D11 option. But I still have the glitch where when I lock the screen and unlock it (via password) then the player starts skipping frames (windowed mode only).
Hprd
27th April 2011, 18:16
It's 10 for the FSE (on 59b/60 raising it didn't help at all, plus the glitches seemed to behave exactly the same on those versions with any setting, since you asked. Although to be honest i only turned it up to 12 briefly, and it didn't make any difference so i put it back to 10. I've done so much though i probably should retest to make sure, lol.), and 8 for windowed, although i did change it to 8 as i had it to 1 b4 (for the rare occasion i play something seriously in windowed), although this was after i first saw the differences on 59 with the tweaks on and off.
EDIT: okay, i put v60 back in, reset to default this time, and reguardless of the backbuffer/tweaks and whatnot, i seemed to be getting glitches at the exact same rate. Put v59 back in again, and perfect.
Ya know there's also something i rememberd i found out on v58 and some earlier versions too. It was if i put (didn't matter what else i changed there, nothing else seemed to do much of anything aside from this) "copy to backbuffer" to flush&wait (sleep) (or loop i think did the same thing too), i wouldn't get any glitches at all, but the render and present que would eventually go down to zero. At 60hz the stats would show no dropped frames or anything. But at lower refresh rates it would drop frames like crazy... I figured it was just the OSD reporting wrong on 60hz, but maybe it had something to do with 60hz that made the playback actually work. I guess i wouldn't really know the effects of what i was actually doing though, lol.
pouyoux
27th April 2011, 18:24
juste tried 0.60 and obviously the D3D11 presentation feature needs some more love :
- I launch a video, go to fullscreen : everyhting works well, and it looks smoother to me than D3D9
- I go back to windowed mode, there are lots of dropped frames with render queue 1-2/8 et backbuffer queue 0-1/3
if I uncheck D3D11 I can go back to windowed mode from fullscreen without problem
6233638
27th April 2011, 18:32
The display mode changer is what it is for now. I said right from the start that it's not the final version, but a simple and stupid first version. You guys can use it in its current form if you find it useful, otherwise simply don't use it.Absolutely: I'm just basically putting out what I hope the display mode changer turns into.
The problem for me is that anything in HD at 50Hz under 70-80 minutes is PAL television content that should be played back at 25fps and anything else should be played back at 24p.
I can't think of an "on/off" checkbox that would cover all scenarios. It could be done so you just add one other option that takes care of my needs, but I'm trying to think of the best solution that would cover what most people would ever need, based on what kind of content they watch.
The minimum options you could get away with to cover most scenarios would be:
1. HD < 80 minutes
2. SD < 80 minutes
3. HD > 80 minutes
4. SD > 80 minutes
12bit output is not possible, there's no Direct3D support for that. You can use either 10bit or you can directly go 16bit, which will also give you extended gamut support. But with 16bit you have to render in linear light and trust in that the GPU/drivers convert that to gamma light properly. Which is a design catastrophe, IMHO. Someone at Microsoft should get fired for such a stupid decision. GPU manufacturers don't exactly have a track record at getting things right. IMHO the only useable alternatives right now are 8bit and 10bit output.10-bit out would be nice, 16-bit seems pointless anyway. Virtually no device will accept a 16-bit signal. The only thing I can think of that does is a Pioneer Kuro, and they are not even 8-bit native displays so it doesn't matter what you're sending them really.
But these are not so important, I have good news!
With 0.60 (skipped 0.59) and the new "limit rendering times to avoid glitches" option, I seem to have smooth playback with 24p@24Hz, 25p@50Hz, and 30p@60Hz again!
This is with all other options reset to the defaults, and the buffers turned up to the maximum.
I tried out the D3D11 renderer (with and without this option) but it doesn't seem to work well here unfortunately.
With 30p @ 60Hz, the present queue is at 5-7/16 compared to 10-12/16 in D3D9.
If I hit Ctrl + Enter to quit fullscreen mode, MPC-HC (3052) stays full size rather than shrinking to 100% as it typically does, and there are no window controls. (min/max/close)
With 24p@24Hz, I just get a black screen (or stats that stay at 0/8, 0/16 etc.) and have to force quit the player.
EDIT: I should point out that the MPC-HC build I use has been hex-edited to support the Arcsoft DTS-HDMA decoders, if that makes a difference.
renq
27th April 2011, 18:39
I have a glitch in D3D11 mode- after going FSE and then into window mode, the 3 buttons (Close, minimise, resize window or whatnot..) on the top right dissappear completely.
W7 SP1
6950 + Cat11.4b
MadVR 0.6
MPC-HC 3030
EDIT: Tried 3050, still happens
Log: http://www.upload.ee/files/1302102/madVR_-_log.txt.html
iSunrise
27th April 2011, 18:45
If I hit Ctrl + Enter to quit fullscreen mode, MPC-HC (3052) stays full size rather than shrinking to 100% as it typically does, and there are no window controls. (min/max/close
Same problem here. Hopefully madshi can find the cause for this, as he seemed unable to reproduce it, yet.
madshi
27th April 2011, 18:45
I still have the glitch where when I lock the screen and unlock it (via password) then the player starts skipping frames (windowed mode only).
<sigh> I hate fixing bugs.
It's 10 for the FSE and 8 for windowed, although i did change it to 8 as i had it to 1 b4
Try setting it to 8 for FSE. Does that bring v0.60 back to v0.59 smoothness for you? I think v0.59 accidently used the windowed setting for exclusive mode.
juste tried 0.60 and obviously the D3D11 presentation feature needs some more love :
- I launch a video, go to fullscreen : everyhting works well, and it looks smoother to me than D3D9
- I go back to windowed mode, there are lots of dropped frames with render queue 1-2/8 et backbuffer queue 0-1/3
if I uncheck D3D11 I can go back to windowed mode from fullscreen without problem
Ok.
With 0.60 (skipped 0.59) and the new "limit rendering times to avoid glitches" option, I seem to have smooth playback with 24p@24Hz, 25p@50Hz, and 30p@60Hz again!
Good to hear.
I tried out the D3D11 renderer (with and without this option) but it doesn't seem to work well here unfortunately.
With 30p @ 60Hz, the present queue is at 5-7/16 compared to 10-12/16 in D3D9.
With D3D9 I have to present one frame per vsync. Which means that when playing back 30p @ 60Hz, I have to present 60x per second. D3D11 has a feature where I can tell it to keep using a frame for up to 4 vsyncs. Which means I only need to present 30x per second, when playing back 30p @ 60Hz. So a queue of 5-6/16 with D3D11 is practically the same as a queue of 10-12/16 with D3D9, when talking about 30p @ 60Hz.
If I hit Ctrl + Enter to quit fullscreen mode, MPC-HC (3052) stays full size rather than shrinking to 100% as it typically does, and there are no window controls. (min/max/close)
With 24p@24Hz, I just get a black screen (or stats that stay at 0/8, 0/16 etc.) and have to force quit the player.
Ok. But if you disregard these problems, just for testing. How does the D3D11 option behave in terms of glitches? Identical to D3D9 or better/worse? Do you even need the "limit rendering times to avoid glitches" option with D3D11? Because I don't, when using D3D11.
pirlouy
27th April 2011, 18:46
I confirm the bug with D3D11 option mentionned earlier (in .60). Window size and position is not what it was, after exiting Exclusive mode.
But the .60 allows me to go into fullscreen whereas 59 did not.
zeroo
27th April 2011, 18:55
In 0.58 I had approximately one glitch every second and no dropped and delayed frames.
With .59/60:
Video 23.976, display 24Hz: 10-15 glitches per minute.
Video 23.976, display 60Hz: 25-30 glitches per minute.
D3D11, video 23.976, display 24Hz: no glitches.
D3D11, video 23.976, display 60Hz: up to 10 glitches per minute.
But now there is one dropped frame every 2-3 minute with and without D3D11.
Hprd
27th April 2011, 18:55
Nope, no difference. Also setting it to 8 on v59 produced the same as anything on 60, then back to 10 was ok again...
dansrfe
27th April 2011, 19:07
Ok so I go into exclusive mode with D3D11 and when I come out of it I lose the minimize, maximize and close (X) buttons in the upper right hand corner :O
6233638
27th April 2011, 19:15
Good to hear.Unfortunately, it seems that it's not completely eliminated. After watching something longer, I do seem to still get the odd presentation glitch here or there. After 20 minutes in one video I had 5 glitches for example.
Ok. But if you disregard these problems, just for testing. How does the D3D11 option behave in terms of glitches? Identical to D3D9 or better/worse? Do you even need the "limit rendering times to avoid glitches" option with D3D11? Because I don't, when using D3D11.After watching a few minutes of the same clip in D3D11 and D3D9 I'd say it's perhaps slightly better, but still unwatchable. It's maybe a 20% drop in presentation glitches, which still meant 16 glitches after a couple of minutes vs 20. Due to the fact it seems somewhat random (there are spots where they always show up, but not in exactly the same place) they may just be the same over a longer video, but that would take a few hours to test.
As with D3D9, it tends to be groups of glitches, so 3-5 at once rather than just one every 20 seconds or whatever.
As previously mentioned, I don't get any video with 24p @ 24Hz in D3D11 mode, just a black screen.
I'm sure there must be something different going on with the 500-series cards from Nvidia to be causing this. If I knew that the 520 or 550 series cards were similarly affected, I'd happily buy you one if it meant getting it sorted, but I have no idea if the same issues I'm having would show up there, or if it would have to be a 560/570/580.
6233638
27th April 2011, 19:30
OK, more testing:
If I enable both "run presentation in a separate thread" and "limit rendering times to avoid glitches" I can use D3D11 mode with 24p @ 24Hz without getting a black screen.
Using just one or the other on their own still results in a black screen, it must be both that are enabled.
Using the same 25p clip at 50Hz that normally results in around 15-20 presentation glitches after a couple of minutes playback, this combination is giving me 0 dropped/delayed, and 0 presentation glitches.
Telling madVR & reclock to treat that clip as 24p (it is actually 25p native but I wanted to keep things the same) I am still getting 0,0,0 at 24Hz.
I need to do longer testing to be confident of this (running in D3D9 appeared to be glitch-free for a short period of time) but it's another combination that should, at the very least, give me a low number of glitches.
Still having the issues leaving D3D11 Fullscreen Exclusive, but after 20 minutes of 24p @ 24Hz and 20 minutes of 30p at 60Hz, I still have 0,0,0 with smooth playback.
fps
27th April 2011, 20:20
juste tried 0.60 and obviously the D3D11 presentation feature needs some more love :
- I launch a video, go to fullscreen : everyhting works well, and it looks smoother to me than D3D9
- I go back to windowed mode, there are lots of dropped frames with render queue 1-2/8 et backbuffer queue 0-1/3
if I uncheck D3D11 I can go back to windowed mode from fullscreen without problemI've also got this problem most of the time (DX11 mode, didn't test with DX9 mode yet). In addition the the massive frame drops in windowed mode the maximize, minimize, etc. buttons vanish, too.
The best way for me to reproduce this behaviour in MPCHC is the follwing:
1) Start movie in windowed mode
2) Right Click -> Fullscreen
3) Let the video play for some seconds in FSE mode
4) Right Click -> Fullscreen
5) Let the video play for some seconds in windowed mode
6) Right Click -> Fullscreen
7) Let the video play for some seconds in FSE mode
8) Double left click to enter windowed mode => massive frame drops most of the time and buttons are gone
The frame drops are more reproducible with a video "that doesn't match the display resolution very good". For example with a 720p video at my 1280x760@60Hz laptop display I only lose the buttons, while there are no frame drops in windowed mode.
However with a 640x360 video I can reproduce both the drops and the missing buttons (forgot to mention the wrong window size as mentioned by others before, too).
These test were all made with v0.60, Win7/32, Nvidia Geforce Go 7400, 1280x760@60Hz.
Prior to v0.60 I let a 40 min video run with v0.59 and the DX11 mode enabled (new tweak options disabled, flush to don't/don't/flush & wait(sleep)/don't, 4 pre rendered frames). I got about 1 glitch every 8-9 minutes. Not too important but I also noticed I can force glitches with every seek or if I close my laptop and hereby turn of the display and then reopen it.
Will try with v0.60 and DX9 mode now.
Qaq
27th April 2011, 20:25
Tried v.060 with D3D11 option enabled (didn't touch anything else).
Win7, ATI 5450, 1080p/23,976 (movie and display), madVR settings: flash/flash&wait/don't/don't.
Everything seems fine, no drops, no glitches while playback in FSE mode. Thanks!
fairchild
27th April 2011, 20:43
Thanks for continuing to work on MadVR madshi. I have a bug which is reproducable everytime when I enable DX11 rendering path. Basically if I am in FSE mode and I close the file, the screen stays black and/or sometimes the media player crashes.
If I disable the new DX11 rendering path, this doesn't happen and I'm able to close the media file while in FSE as I have been able to do with previous versions. All settings at default aside from chaging Luma scalers.
Plutotype
27th April 2011, 20:45
Ok so I go into exclusive mode with D3D11 and when I come out of it I lose the minimize, maximize and close (X) buttons in the upper right hand corner :O
confirmed here too, this is exit from MPC-HC when in FSE:
http://img155.imageshack.us/f/exitmpchcfromfse.png/
Thunderbolt8
27th April 2011, 20:53
Im getting a few glitches now, just some, but not too many (winXP). let a movie play for ~5 mins and got ~7 glitches. seems they appear when the present queue goes down to from 2-3 / 3 to 1-3 / 3 in between sometimes. so far, this hasnt happened with the versions before (always used the same movie).
dansrfe
27th April 2011, 20:55
Basically D3D11 needs a lot of fixes...reverting back to D3D9 for now I guess...
e-t172
27th April 2011, 21:08
madVR 0.60 TEST REPORT
Win7 x64 - NVidia GTX 280 270.61 - MPC-HC 1.5.1.2903 - 1920x1080p24@48.034Hz - Reclock fully enabled
All test parameters are stated as a differential with default settings. Indications of dropped/duplicated frames are not derived from counters but from actual observation using Reclock's tearing bar test. madVR was completely shut down (MPC-HC file close) between each test.
- Default settings: presentation glitches counter increasing at 3-5 per second, dropped/duplicated frames, all queues stay full
- "use D3D11 for presentation" checked: presentation glitches counter increasing at 1-2 per second, dropped/duplicated frames, all queues stay full
- "run presentation in a separate thread" checked: presentation glitches counter increasing at 3-5 per second, dropped/duplicated frames, all queues stay full
- "limit rendering times to avoid glitches" checked: presentation glitches counter increasing at 2-3 per second, dropped/duplicated frames, all queues stay full
- "run presentation in a separate thread" checked, "limit rendering times to avoid glitches" checked: presentation glitches counter increasing at 1-3 per second, dropped/duplicated frames, all queues stay full
- "run presentation in a separate thread" checked, "limit rendering times to avoid glitches" checked, "use D3D11 for presentation" checked: presentation glitches counter increasing at 0-1 per second, dropped/duplicated frames, all queues stay full
- "run presentation in a separate thread" checked, "limit rendering times to avoid glitches" checked, "use a separate device for presentation" unchecked: presentation glitches counter increasing at 3-5 per second, dropped/duplicated frames, all queues stay full
- "present several frames in advance" unchecked: no counters increasing, dropped/duplicated frame every 5 seconds or so, all queues stay full
- "present several frames in advance" unchecked, "use a separate device for presentation" unchecked: no counters increasing, rare dropped/duplicated frames (once per 30 seconds or so), all queues stay full
Tried to play with other options like flush parameters. Didn't change anything.
A word of caution, though: I thought I had perfect playback with 0.58, but when I tried 0.58 again, I realized that I was unable to get perfect playback with it, no matter how hard I tried (the best I could achieve is a dropped/duplicated frame every 30 seconds or so). In fact, I realize that I currently have no solution for really smooth playback on my PC :( . So I think that these test results should be interpreted as absolute, not relatively to previous versions. This is quite disappointing. In fact, I think I lost perfect playback when updating my graphics drivers from a quite old version (260.89, 2010-10-20) to 270.61.
Oh, and like everyone else, I lose MPC-HC close/min/max buttons when using D3D11, and dimensions of the MPC-HC window get altered.
SamuriHL
27th April 2011, 21:11
I got a chance to quickly test this out. I only enabled D3D11 and changed nothing else. It works fine in MC16. No issues at all.
LoRd_MuldeR
27th April 2011, 21:26
Ok so I go into exclusive mode with D3D11 and when I come out of it I lose the minimize, maximize and close (X) buttons in the upper right hand corner :O
I noticed that as well :confused:
BTW: Does the D3D11 path actually work with GPU's without D3D11 hardware support? (like GTX 260)
Reith
27th April 2011, 21:27
Personally, I'm able to get zero presentation glitches with the "use a separate device for presentation" option checked. But if I use D3D11, I start getting glitches again (and there's also the problem with the media player buttons disappearing still) so I'll stick with D3D9 for now.
The options to limit rendering times and run presentation in a separate thread don't seem to make a difference as I can get zero glitches without them, but they don't hurt either.
Hypernova
27th April 2011, 21:58
OK, more testing:
If I enable both "run presentation in a separate thread" and "limit rendering times to avoid glitches" I can use D3D11 mode with 24p @ 24Hz without getting a black screen.
I have the "windows size doesn't change back problem" as well. Enable "run presentation in a separate thread" make the windows size return, but still no windows control buttons. Disable "present several frames in advance altogether" seems to fix both problems. Note that while this result is consistent on my machine now, it didn't seem that way when I was playing with the options the first couple times i.e. the windows control buttons doesn't return on my first two tries of disable "present several..." maybe I didn't restart the player though.
Gleb Egorych
27th April 2011, 22:01
Here are my test results of 0.60, compared to 0.58. Win7 x64, 8800GT, 260.71, Zoom Player.
1) D3D11 path: it works, haven't discovered any benifits over D3D9.
2) D3D11 path: I can confirm lost focus bug after entering fullscreen exclusive. D3D9 is OK.
3) D3D11 path: after exiting fullscreen Zoom Player's video is resized to fullscreen resolution. D3D9 is OK.
4) "limit rendering times" reduces number of glitches for 25fps @ 60Hz (comparing to 0.58).
5) "run presentation in a separate thread" increases number of glitches for 25fps @ 60Hz.
6) 0.60, 8 pre-presented frames, separate device for presenting, limit rendering times: 60fps @ 60Hz are perfect smooth with ReClock (like 0.58 on same settings except "limit render times")
moophone
27th April 2011, 22:43
Unfortunately logs rarely help with crashes. Which splitter/decoder are you using? Have you tried different splitters and different decoders?
I tried with both CoreAVC 2.0 and latest ffdshow trials for decoders. Both crash mpchc.
I am using the internal mpchc splitter. I also tried haali.
EDIT: The problem is in the latest mpchc builds. I reverted to the latest official and the problem isnt there.
Andy o
27th April 2011, 22:44
When coming out of D3D11 exclusive, I'm also missing the min/max/close window buttons and also, the original window size in MPC-HC is not respected. I have auto zoom on, to auto size, but 50% or 100% don't work either. The window becomes the size of the screen regardless.
Using .60, btw, Win 7-64, ATI 5770.
e-t172
27th April 2011, 23:28
Ok, I've done some experiments and I think I have some VERY interesting information to report.
I reported earlier (http://forum.doom9.org/showpost.php?p=1496176&postcount=7440) that with NVidia drivers 270.61, I was unable to get perfect playback, no matter what options I tried. The best I could achieve is one frame drop/duplicate every 30 seconds or so.
So I've downloaded a bunch of earlier NVidia driver versions, and I did some tests. Tests were made using madVR 0.60, 24p@48.034Hz, with everything to defaults except "present several frames in advance", which was disabled. This is the configuration which, historically, gave me the best results.
Here are the results:
- NVidia 260.89: perfect playback
- NVidia 260.99: perfect playback
- NVidia 266.35: one frame drop/duplicate every 30 seconds or so
- NVidia 266.58: one frame drop/duplicate every 30 seconds or so
- NVidia 270.61: one frame drop/duplicate every 30 seconds or so (see details (http://forum.doom9.org/showpost.php?p=1496176&postcount=7440))
I'm positive that the NVidia drivers version 266.35 introduce a regression, at least with my configuration. I reinstalled 260.99 and 266.35 multiple times to make sure, there is no doubt in my mind about this. I would not be surprised if it had something to do with the fact that I use an exotic refresh rate (48.034Hz).
Now the good news is, NVidia 260.99 is like the madVR dream for me. No matter what options I choose, playback is ALWAYS perfect! Old path? Check. New path? Check. D3D9? Check. D3D11? Check. Without hacks? Check. With hacks? Check. Presenting 16 frames in advance? Check. All flush settings to "Don't flush"? Check. It doesn't matter what options I choose, I will always get silky-smooth playback, all the time. No presentation glitches, no dropped frames, nothing.
The irony is, when I was on 270.61 it was impossible to get smooth playback, now with 260.99 it is impossible NOT to get smooth playback!
To everyone experiencing issues with madVR on an NVidia card, I highly recommend to try downgrading to NVidia 260.99. I would be surprised to be the only one experiencing such a huge difference between two driver versions.
leeperry
27th April 2011, 23:30
* Windows Key press now forces a 3 second delay before going back to exclusive
I don't get the start menu when PotP's working fine(as it enforces "always on top" behavior) but when PotP's frozen due to an error at decoding, it works wonders...great stuff, thanks a bunch!
I've also played around w/ the new path and its 2 options, whatever I do presentation glitches still increase...but very very slowly(I also get 2/3 or 3/3 buffered frames, and that changes very slowly too). The old path is still butter smooth AFAICS...the only times I see hiccups, it's always in the original telecine when I rewind, duh :p
deptel
27th April 2011, 23:44
When I turn on the D3D11 option, entering full screen exclusive mode simply crashes the player, gives a typical windows close dialog and instantly shuts off the player.
But still playing fine without it enabled, just thought i'd report it anyway :).
6233638
27th April 2011, 23:59
I'm positive that the NVidia drivers version 266.35 introduce a regression, at least with my configuration.Please list what your configuration actually is in your signature. It seems that different generations of Nvidia card are having different issues.
As I do more than just run MadVR on my PC, the latest Nvidia drivers are essential. (considerable performance improvements in many games)
For what it's worth though, I tried a number of driver versions and saw essentially no difference as far as MadVR was concerned.
mrcorbo
28th April 2011, 00:00
I'm having some issues with the new DX11 path and I'll detail those when I have more time. For now though, I'd like to report one big success. I have a 720p60 ATSC recording that no matter what I did at best dropped one frame every several seconds. With the DX11 path enabled, in FSE mode playback on this file is now PERFECT.
pankov
28th April 2011, 01:03
madshi,
I didn't have a lot of time to test v0.59/59b/60 but here are my results (from v0.60)
1. I've tried D3D11 path and at my PC it doesn't make any difference in glitches/dropped/delayed frames - everything stays at 0/0/0. I've noticed a couple of problems though:
a) if I enable the "run presentation in a separate thread" option and I start the playback in FSE mode I get stuck frame on screen with audio playing. If I go out of exclusive and back to it everything plays fine. Here is a log http://www.mediafire.com/?tbnakeac9bqxrlr
b) exiting the player (ZoomPlayer) while in FSE mode results in a crash ... also jumping to the next file. And this crash is not the access violation I was getting with previous versions but a lock which Windows 7 shows as "Zoom Player has stopped working. A problem caused the program to stop working correctly. Windows will close the program and notify you if a solution is available.". If I exit the player / jump to next file in windowed mode everything is fine. btw I've noticed that the access violation with D3D9 mode happens with every jump to next file on my secondary monitor and very seldom (close to never) if I do the same on the primary monitor. (the above log file also shows this new "lock on exit" problem)
c) I too get the problem as Gleb Egorych and others about the size of the player window. For me I have to "exit fullscreen" twice to go back to the "normal" window. "exiting fullscreen" only once makes the player window "maximized", not "restored" as it did with older versions or with D3D11 disabled
2. I think somebody already mentioned it but there is a problem with the new logic for showing the seek bar. Whenever I place the mouse over the taskbar on my primary monitor the seek bar shows up on the secondary. Also in my multi-monitor configuration
-------------
| |
| 1 |
| |
------------- -------------
| |
| 2 |
| |
-------------
whenever my mouse is on the secondary monitor (even at the top of it) the seek bar is always visible.
Rain1
28th April 2011, 01:46
Ok so I go into exclusive mode with D3D11 and when I come out of it I lose the minimize, maximize and close (X) buttons in the upper right hand corner :O
+1
Does the D3D11 path actually work with GPU's without D3D11 hardware support? (like GTX 260)
I want to know that as well /
So far, I find out that (playing 1080p23.976 on 60Hz monitor):
* If D3D11 was used: got glitch (other settings does not matter)
* If 3rd flush option is anything other than "flush & wait": got glitch (other settings does not matter).
* If D3D11 was not used & 3rd flush setting = flush & wait: no glitch
(Automatic display mode matching didn't seem to work for me, so i left it empty)
dansrfe
28th April 2011, 02:36
I know you're not truly obligated to fix this, madshi, however with the Windows XP qdvd.dll used in Windows 7 the exclusive mode seekbar is not shown. I hope you can fix it :) Thanks!
Ziddy76
28th April 2011, 02:44
Using .60 and sometimes the MadVR tray disappears after going into fullscreen and back.
6233638
28th April 2011, 02:49
I know you're not truly obligated to fix this, madshi, however with the Windows XP qdvd.dll used in Windows 7 the exclusive mode seekbar is not shown. I hope you can fix it :) Thanks!MadVR is also incapable of detecting the framerate with this as well, once you switch title.
I was actually just going to ask if it could either be fixed, or if we could have the option to:
1. Have madVR treat an unknown framerate as 24fps. (I'm not entirely sure what it does in that scenario, but I did have to switch to 24Hz manually)
2. Have an option to have madVR return to a selected display mode after playback rather than simply the previous mode. I had to set my display to 24Hz manually, and would prefer it if madVR always switched back to 1080p60 after exiting. As it is, I then had to switch over to 60Hz manually after playback.
But this seems like really low priority stuff, if Madshi even considers it something he wants to address at all.
neoufo51
28th April 2011, 02:50
I have a question:
I want to use MadVR fulltime but I'm missing a simple to use feature on it that MPC-HC has built-in with EVR:CP. On that renderer's settings, I usually enable color management and set to "absolute colorimetric" to achieve the most accurate looking image on my display. How can I achieve the same effect with MadVR?
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.