|
|
Unfortunately, I can't reproduce this issue here, which makes it very hard to do anything about it. Have you tried both 32bit and 64bit versions? |
|
|
Reinstalling the whole K-Lite Codec Pack fixed the issue, but through applying my own preferred settings again (and stumbling onto the same problem) I've managed to trace it down to a single setting. The problem only occurs when under madVR Settings->Rendering->Dithering the "Error Diffusion - Option 2" is selected (regardless of other modifiers). |
|
|
Strange. I've no idea why that would cause problems for you. Hopefully it's not a show stopper and you can use a different dithering method? |
|
|
I can confirm, that it has to do something with the latest AMD Adrenaline driver 24.9.1 (my GPU: 7800xt). It worked on beta 206b + Adrenaline 24.8.1 like usual, but with the update to 24.9.1 something broke. I am amazed, that OP could track it down to one option. I can confirm, that Error Diffusion 1 works and Error Diffusion 2 is buggy. |
|
|
No idea why. |
|
|
this is nearly for sure a madVR issue. i have "only" a couple of sec for this issue. why is madVR the issue. because 0.9217 works just fine. what is actually happening newer mpc-hc version report that a 25 FPS video is played and we get a screen when it reports the actual video. it looks like the first attachment. my best and blind guess is that it takes that long to check the clock or if it is a capture card. should be easy to check if first version with a timer and the first version with a capture card detection combined with that last version that has none. a copied version with a two week time should more then enough for me to check and rule that out. i don't use error diffusion BTW. and use a 4060 so as different as it gets. image.png (54,779 bytes) |
|
|
I'm confused about your report. You're talking about a 25 FPS video and a new MPC-HC version, and then a clock and capture card. I don't understand how all this goes together. What does 25 FPS have to do with anything, and how is the MPC-HC version involved in that? E.g. does this only occur with 25 FPS video? Or does it only occur with the latest MPC-HC version but not with older versions? |
|
|
the video is not 25 fps but it is reported as 25 when the bug happens (seems to be the fallback FPS). as soon as madVR starts rendering frames it will be 23.98 (they round to two digits). it only happens with new madVR versions and not with the old versions. happens in mpc-be and HC both with internal decoder i just used 1.7.9 32 bit same issue (does not report 25 HZ it is to old for that). other video renderer work completely fine. making this unlikely to be a video player issue. this is an SDR video so the HDR changes shouldn't matter much here too. the biggest changes are the timed versions and capture card detection. i can not do a regression check because they will not render anything and i don't know the number of them. but these would be the version i would start checking to rule this out and to be a rough idea where a change was made to create this bug. why clock and capture card. because there is no reason for the older version to check the system clock and or if it is a capture card. image-2.png (50,954 bytes) |
|
|
So it happens with one specific video, only? Can you upload a sample of that video, so that I can reproduce it here? |
|
|
happens with every video and for quite a while. it's only 20-41 frames on my system. youtube-dl.exe -f 614 "https://www.youtube.com/watch?v=7PfH9VeoPTU" youtube-dl.exe -f 251 "https://www.youtube.com/watch?v=7PfH9VeoPTU" just put https://www.youtube.com/watch?v=7PfH9VeoPTU in mpc-be |
|
|
I don't have youtube-dl.exe. Can you provide the download somewhere for me? If I download it by using a different tool, the bug might not happen? |
|
|
mpc-be doesn't need it. it has something similar build in. option -> playback -> online media... happens with every video anyway. i tried the mp4/webm directly. m2ts blu ray remuxes. literally everything. the example file is just trash i had lying around. let me look up a file host. https://filebin.net/6a1jjyd49h1imj6y |
|
|
That file works for me. It seems that madVR works fine for most people. There must be a reason why you're having problems? Is it only on one specific PC? Or is it maybe only with the internal decoders but not external decoders? I've also not heard about fps being misreported. Does this also happen always? Or maybe also only with specific decoders? |
|
|
decoder doesn't matter i even tried nvdec which has trouble even decoding the vp9 video. i tried old lav version and so on on win 11 24h2 with a 1060 (don't ask) it works as usual. |
|
|
So it's just 1 specific PC where it doesn't work? Did this problem just start with the latest build? Or did it happen with the previous (newer) builds, as well? |
|
|
it's been happen for a while could be many years. i only have 20-41 frames madVR dropping frames that's not that unusual even in the past. it's just the first time i saw someone to report multiply seconds. i only have 2 windows Pc. and there is no report of win 11 yet. |
|
|
Well, if it's just one PC causing trouble, then it could be anything, hard to say. The other reports from this thread have a very specific reason: Namely newer AMD drivers breaking one specific type of error diffusion. That's a completely different issue. |
|
|
found it. this happens with the legal stuff is unzipped my stable 09217 didn't had this. this should mean madVR is still compiling shader at the start and does this everytime. does error diffusion need shader compiling and is not precompiled? 18 sec sounds like quite something. could you not dump a shader cache with these shader like mpv doesn't this doesn't fix error diffusion but that's something. |
|
|
Good to hear you found it. It has been many years since I looked at the shader compiling when unzipping the legal stuff. Do you really need that? Hopefully not? It's not needed for error diffusion, for sure. |
|
|
more important is to know if that's the issue with error diffusion that it is delayed to work for some reason. i don't need that and it is technically still working and that's the important part. i will put my 5700 XT back into a PC temporary and try to reproduce the issue. completely different architecture so don't expect anything here. does error diffusion still use one of these compute APIs openCL, directcompute or what ever something that can explain a delayed answer from the driver? the error is literally the same audio works but there is no image for a duration. |
|
|
Unzipping the legal stuff has absolutely nothing to do with error diffusion. As I said, that's a completely different issue. Error diffusion uses DirectCompute. Clearly, AMD broke something in the latest driver. |
- Anonymous
