madVR - high quality video renderer (GPU assisted)

[bugmen0t]

Originally Posted by madshi

If you disable the display mode switcher then madVR is not responsible for the refresh rate.

For me madVR does switch to 60p in FSE mode with display mode switcher off since v0.86.7 and before it switched to 59p.

[mark0077]

ah great thanks, I guess it reads the original frame rate rather than rate coming out from last filter? Its seeing the 60.000fps output from SVP as the original 24.000fps . Its not an issue, I will try to make a custom windows resolution to see can I get around it.

[madshi]

Originally Posted by bugmen0t

For me madVR does switch to 60p in FSE mode with display mode switcher off since v0.86.7 and before it switched to 59p.

It probably has in v0.86.7, but it should not in v0.86.8. The fix I implemented in v0.86.8 is now only supposed to be applied if the display mode switcher is turned on.

Originally Posted by mark0077

ah great thanks, I guess it reads the original frame rate rather than rate coming out from last filter?

If SVP sends 60.000fps but tells madVR that it's sending 24.000fps then complain to the SVP devs. Or if SVP is run via AviSynth (I don't know if it is) then it's probably an ffdshow problem. In any case, it's not madVR's fault.

[AGKnotUser]

86.8 creates a log file every time I start mpc-hc. Any way to turn that off? I couldn't see anything in the setup.

[bugmen0t]

Originally Posted by madshi

It probably has in v0.86.7, but it should not in v0.86.8.

It does so with v0.86.8, too. However I don't see a problem with that.

EDIT: Was able to make it switch: Played a PAL DVD in FSE first, still at 60p. After that, if I load a BluRay it will switch to 59p in FSE. Quit MPC-HC. Restart it and load the BluRay again and it switches back to 60p (as it would have played at 60p without playing the DVD first). Really strange. (All with display rate changer off, of course.)

EDIT 2: That even holds true if I turn the display rate changer on and set it to 1080p60! Still switches to 59p if a PAL DVD was played before.

[madshi]

Originally Posted by AGKnotUser

86.8 creates a log file every time I start mpc-hc. Any way to turn that off? I couldn't see anything in the setup.

When did you download v0.86.8? The first time I uploaded (some time yesterday) it was broken the way you report. However, I've fixed that in the meanwhile. Just re-download and you should be fine.

http://madshi.net/madVR.zip

Originally Posted by bugmen0t

It does so with v0.86.8, too.

Strange.

[AGKnotUser]

Originally Posted by madshi

When did you download v0.86.8? The first time I uploaded (some time yesterday) it was broken the way you report. However, I've fixed that in the meanwhile. Just re-download and you should be fine.

http://madshi.net/madVR.zip


Strange.

That did the trick. Thanks.

[mark0077]

Originally Posted by madshi

If SVP sends 60.000fps but tells madVR that it's sending 24.000fps then complain to the SVP devs. Or if SVP is run via AviSynth (I don't know if it is) then it's probably an ffdshow problem. In any case, it's not madVR's fault.

Ah right ok np. LAV seems to also incorrectly report its output rate. Eg without ffdshow / SVP, if I use LAV to deinterlace a 25i to 50p, it still says 25fps. I'll post on LAV thread.

[jmonier]

I'm getting strange results when using madTPG. I do not see any of this when going through the madVR pattern generator. I have verified it when using HCFR. There may also be problems with ArgyllCMS but I can't be sure at this point.

It's most obvious when I attempt to measure the uncalibrated response of my LG. I get reasonable results when I go through madVR. I get much worse results when I go through madTPG. The worst of these is a gamma of 2.7@10% going down to 1.4@90%. (The gamma when going through madVR is relatively flat.)

The states for "disable VideoLUT" and "disable 3dLUT" do not affect things at all.

I do have a 3dLUT generated from ArgyllCMS according to madshi's latest workflow. Except for gamma, it seems fairly good when I measure through madVR. (The gamma is approximately the shape of a BT.1886 curve but has values about 0.1 higher than the HCFR BT.1886 curve at it's high end.)

Whether or not the 3dLUT is enabled in madVR, I get the same bad results above (crazy gamma, etc.) when attempting to measure it via madTPG from HCFR. In other words, I can find no combination in which the 3dLUT affects things when I use madTPG.

[cyberbeing]

Originally Posted by ryrynz

so that 10% GPU usage really is in the more extreme situations.

It is indeed, but that 10% is actually a bit deceptive, since it actually bottlenecks my GPU slightly and makes my queue unstable once GPU load hits ~50% if I also had something like AR enabled, while if I disable it my queues are stable up to ~95% load. This is why I keep it disabled, but overall I don't care what the default is, as long as there is still a trade-quality option to toggle it off. In motion, I cannot even see the difference between enabled/disabled anyway.

My GT440 DDR isn't that weak, as it's ~2x as powerful as the Intel HD4000 iGPU, it's just that there are certain shader operations like linear light conversions which have a rather bad performance:quality trade-off compared to faster GPUs. Could I use much heavier settings than I do with 23.976 content, sure, but I value stability over all else. I usually tune my madVR settings so they never exceed 99% load or drop frames even under the most extreme cases, which I may only experience ~1% of the time. And yes, this usually means I'm usually sitting at ~20% GPU load normal case (Spline36 AR Luma + Catmull-Rom Chroma + 3DLUT).

[madshi]

Originally Posted by jmonier

I'm getting strange results when using madTPG. I do not see any of this when going through the madVR pattern generator. I have verified it when using HCFR.

I've just tried on my LCD computer monitor, and HCFR produces the same results for me, regardless of whether I use MPC-HC+madVR for test pattern generation or whether I use madTPG. Did you test both in the same mode (windowed vs. fullscreen exclusive)?

[jmonier]

Originally Posted by madshi

I've just tried on my LCD computer monitor, and HCFR produces the same results for me, regardless of whether I use MPC-HC+madVR for test pattern generation or whether I use madTPG. Did you test both in the same mode (windowed vs. fullscreen exclusive)?

I always use madVR fullscreen exclusive. I did use madTPG fullscreen (and I'm assuming that is exclusive as well). I am using both on a secondary monitor (I move madTPG to that monitor) for what that's worth.

In HCFR I always do a Measures/Configure before each run (because of the bug previously reported). The only other thing I can see in HCFR is "Blank Screen during Measure" which I left checked.

I'm using ZoomPlayer rather than MPC-HC but, of course, it's working fine there.

Did you check whether the "Disable VideoLUT" and "Disable 3dLUT" have any effect? What exactly are they supposed to do (since I don't see any effect)?

Do the madVR settings have any effect on madTPG? If not, how would madTPG know where to find the 3dLUT?

I will check further on HCFR. Right now, I'm doing an ArgyllCMS calibration using madVR. The numbers I'm seeing from dispcal (while in progress) are significantly different (and probably better) than I saw when doing it using madTPG.

EDIT: I only entered the 3dLUT path into REC709 and left the others blank. Is that right?

One more thing: ArgyllCMS says it has no access to the VideoLUT. Is that normal?

And another thing: I have an AMD/ATI card. I'm wonder if there could be a driver issue.

[madshi]

Originally Posted by jmonier

I always use madVR fullscreen exclusive. I did use madTPG fullscreen (and I'm assuming that is exclusive as well). I am using both on a secondary monitor (I move madTPG to that monitor) for what that's worth.

In HCFR I always do a Measures/Configure before each run (because of the bug previously reported). The only other thing I can see in HCFR is "Blank Screen during Measure" which I left checked.

I'm using ZoomPlayer rather than MPC-HC but, of course, it's working fine there.

Did you check whether the "Disable VideoLUT" and "Disable 3dLUT" have any effect? What exactly are they supposed to do (since I don't see any effect)?

Do the madVR settings have any effect on madTPG? If not, how would madTPG know where to find the 3dLUT?

I will check further on HCFR. Right now, I'm doing an ArgyllCMS calibration using madVR. The numbers I'm seeing from dispcal (while in progress) are significantly different (and probably better) than I saw when doing it using madTPG.

EDIT: I only entered the 3dLUT path into REC709 and left the others blank. Is that right?

That sounds all quite weird to me. Could you do some more tests, trying to find out where the problem comes from? E.g. could you please try windowed mode instead of fullscreen mode for both madVR and madTPG and see if that makes a difference? Also please check with your primary monitor to check if it works ok there. Thanks.

[jmonier]

Originally Posted by madshi

That sounds all quite weird to me. Could you do some more tests, trying to find out where the problem comes from? E.g. could you please try windowed mode instead of fullscreen mode for both madVR and madTPG and see if that makes a difference? Also please check with your primary monitor to check if it works ok there. Thanks.

OK, here's what is happening:

If madTPG opens on the primary monitor and is then moved to the secondary monitor, everything is OK (including operation of the "Disable 3dLUT")

If madTPG opens on the secondary monitor (because it was closed on that monitor) you get the problems that I described.

Note that the 3dLUT needs to be enabled in madVR for it to be used in madTPG. I presume that this is by design (either explicitly or implicitly). Any more info on the use of the Disable buttons in madTPG would be appreciated.

One more note about HCFR: I had problems with it measuring the primary monitor until I turned off "Blank screen during measure". I don't think that had anything to do with the primary problem but I'm going to leave it off from now on.

[madshi]

Interesting. What happens if you run "dispwin -c -d2" in the situation where you get weird results in madTPG. Does that fix the issue? Or what happens if you call "dispwin -c -d2" in the situation where you get *good* results in madTPG. Does that reproduce the problem?

Edit: The buttons are supposed to do with their caption says. If you have a 3dlut configured for the display madTPG runs on, the "disable 3dlut" button is supposed to disable the 3dlut (only for madTPG, nor for madVR).

[jmonier]

Originally Posted by madshi

Interesting. What happens if you run "dispwin -c -d2" in the situation where you get weird results in madTPG. Does that fix the issue? Or what happens if you call "dispwin -c -d2" in the situation where you get *good* results in madTPG. Does that reproduce the problem?

I'm not sure that I understand. To begin with, what I call the secondary monitor is the one that is not the main or primary display and does not have the taskbar. That ends up being #1 on most of my systems. So, I think you mean -d1 rather than -d2.

Or did you want to clear the VideoLUT on the main display while madTPG runs on the secondary display?

Beyond that, did you want me to do this once madTPG is positioned on the secondary display (but before running HCFR), either by opening there (*bad* results) or being moved there (*good* results)?

[jmonier]

Beyond what I said above, I'm not sure that I understand how the VideoLUT is used in conjunction with madVR. I always use fullscreen exclusive so I'm primarily interested in that case.

Is it active when madVR is used?

Does it get loaded when the 3dLUT is used?

In your workflow, should dispcal.cal be loaded into the videoLUT or is it included in the 3dLUT?

Should dispcal.cal be loaded into the VideoLUT for use by players that can't use madVR?

[tobindac]

Program needs some kind of easy way to check what version you've installled.

Not that one can't just force a reinstall or check the changelog, but it's weird it doesn't have an About on the main interface.

edit: Oh, there is one, bottom-left corner. Easy to miss, too used to About messages.

[dansrfe]

Is there a tool out there that can selectively turn off displays in a multi-display setup? While watching a movie on one monitor at night, for example. It would be easier than turning off the monitor manually since the power button is in an awkward spot behind the monitors.

[tobindac]

Does anyone know if SVP's 60FPS affects madVR's output image quality? I LOVE 60FPS because it just makes everything lively (and I don't accept the "home video" arguments personally). I noticed madVR does properly show a nasty result on "Nearest Neighbor" and a properly nice result on other algorithms, but I wonder if I miss something.