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. |
13th April 2009, 19:21 | #241 | Link |
Registered User
Join Date: Jan 2009
Location: UK
Posts: 403
|
madshi, aspect ratios are still not (quite) handled correctly. I've got a 704x576 encode with a 16/9 ar that madVR scales to 1918x1080, not 1920x1080. VMR9 and Haali both scale correctly.
Here's a sample. Thanks very much. |
13th April 2009, 20:17 | #245 | Link | ||||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Since you mentioned the Germany My Fair Lady DVD, here's that DVD, enlarged to my PC display's native resolution of 1680x1050: Left side is "SoftCubic50", right size "Lanczos8". The ringing you can see in the right side is not in the source at all. It's added by the Lanczos resampling algorithm. As you can see, I didn't have to search very long to find ringing. After all this is still the movie intro... Here are the full screenshots: SoftCubic50 Lanczos8 Quote:
Quote:
Quote:
Quote:
Quote:
Again (and again and again) motion smoothness is not a thing I have really looked into yet. Did your GPU stats (see Ctrl+J), especially "average gpu rendering time" get worse with 0.4 compared to 0.3? That's a thing I would have to look into. |
||||||
13th April 2009, 20:26 | #246 | Link |
MPC-HC Project Manager
Join Date: Mar 2007
Posts: 2,317
|
Hi
I just tested v4 (did not try the older builds) The bat file failed to install on my system (windows 7 32bit) The 1:1 results for media my system can handle are okay, but as soon as any scaling is applied the video becomes a slideshow. My specs: AthlonXP 2600+ @ 2ghz A7N8X Deluxe ATI HD2400pro AGP Windows 7 Catalyst 9.3 CPU% maxes out as soon a scaling is enabled 1:1 usage is between 30-80% EVR-CP cpu usage is 10% Screenshot of 1:1 stats: Screenshot of scaled to 1920x1080: due to no DXVA support i was unable to test anything above SD resolution (With overlay renderer my cpu is just fast enough for 1280x720) |
13th April 2009, 20:39 | #247 | Link | |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Did you get any specific error message?
Quote:
Wow, what a difference in gpu rendering times! Ok, the source seems to have a very small resolution. Of course that helps keeping rendering times down in 1:1 mode. "average gpu rendering time" is the most important and it's quite nice with 5ms in 1:1 mode. It needs to be smaller than the "movie frame interval" (40ms). However, in scaled mode your "average gpu rendering time" is going through the roof with almost 60ms! That's too much. It seems that your graphics card is not fast enough to do scaling to 1080p with full madVR quality options. You can turn down the quality a bit in the madVR settings dialog. Maybe that helps? Try lowering luma+chroma textures to 10bit. With a bit of luck that might already be enough to push "average gpu rendering time" down under 40ms. |
|
13th April 2009, 20:49 | #248 | Link |
Registered User
Join Date: Apr 2008
Posts: 1,106
|
Using version 4, with ffdshow in the chain doing de-interlacing (if necessary) I get DVD Macrovision failed error playing backed up dvd's from hdd.
This was the reason I could never use Haali also, could never get rid of this error. |
13th April 2009, 21:43 | #249 | Link | |
Registered User
Join Date: Jul 2008
Posts: 532
|
Quote:
The first sample (VC-1 in MKV) gives me the correct aspect ratio with the MPC-HC built-in decoder. I always got a wrong aspect ratio when the Microsoft decoder is used and it doesn't matter which splitter is used. The Arcsoft decoder doesn't want to connect even with the VC1tweak filter so I'm stuck to the MPC-HC decoder or the Microsoft one, but as you know the MPC-HC decoder is not working fine with your renderer for now. Now I got different results with the Beyonce TS sample depending on which combination of splitter + decoder is used. So I've done more tests: Splitter + Decoder = aspect ratio result MPC-HC + MPC-HC = wrong MPC-HC + Arcsoft = wrong MPC-HC + CoreAVC = OK MPC-HC + Divx7 = wrong Arcsoft + MPC-HC = OK but lot of stuttering (+ the bug of the decoder) Arcsoft + Arcsoft = OK Arcsoft + CoreAVC = wrong Arcsoft + Divx7 = doesn't connect I don't have Haali installed and I don't plan to install it. My config is actually working fine for every use I have except some troubles with MadVR. But I know it's a work in progress, I don't blame I only report. Good and impressive job anyway, thanks for that |
|
13th April 2009, 22:00 | #250 | Link |
Registered User
Join Date: Jun 2005
Posts: 630
|
@madshi: 0.4v High CPU to some extend and also still gets only the refresh rate from the original monitor.
update on the refresh rate issue: I found a way to reset it. Funny enough if you drop a new video to mpchc window then madVR will show proper Hz after reinitialisation. However, that doesn't happen if you just stop/restart the video. Needless to say that even if I reset the video, and then move to another monitor, it doesn't pick up the rate again. Since 0.3v the behaviour has slightly changed now. it seems the issue with secondary monitor is almost nailed!!! It seems now there's no difference between primary and secondary monitor in terms of CPU consumption, only thing matters is what monitor the video has been initialized in!!! The following scenario is now valid: open an empty window on primary, move to secondary, drop a video into it -- no core maximisation ;P move the window back to primary -- maximisation is present again If I revert the order of monitors in this scenario and do it again, same thing happens. To sum up, it is possible to use a workaround now, which is to force madVR to reinitialize on a different monitor. That apparently fixes now both refresh issue and CPU consumption. Only thing now is to make it work automatically ) |
13th April 2009, 22:03 | #251 | Link | ||
MPC-HC Project Manager
Join Date: Mar 2007
Posts: 2,317
|
Nope,it just flashes on the screen for a second and is gone.
Manually running the command from a admin command prompt works fine Quote:
High CPU consumption is caused by the OSD stats(ctrl-J) without ctrl-J cpu usage is always 35% regardless of scaling settings, retested with EVR-CP too and it's not as stale as with madvr, it hovers between 10 and 22% averaging around 15 EDIT: i also tested beliyaal's build and it gets an average of 25 Quote:
Last edited by tetsuo55; 13th April 2009 at 22:08. |
||
13th April 2009, 22:15 | #252 | Link |
Registered User
Join Date: Dec 2001
Location: Israel
Posts: 34
|
MadVR 0.4 crash + minor bug
Hi,
First, I'd like to thank you for your iniative creating MadVR. I wrote a VR for internal company use a year ago and I know the pains.... I came into 2 issues: When pressing CTRL+ALT+DEL (going into the screen where you can lock your computer or start task manager) and coming back to the desktop, the player crashes. Doesn't matter if in full screen or not. You should check you have a valid D3D device before the call to pd3dDevice->Present(). something like (similar code was found in direct show sample code from Windows SDK): Code:
HRESULT hr = pd3dDevice->TestCooperativeLevel(); //device not ready! if (hr != D3D_OK) { // If the device was lost, do not render until we get it back if (D3DERR_DEVICELOST == hr || D3DERR_DRIVERINTERNALERROR == hr) { return E_FAIL; } // Check if the device needs to be reset. if( D3DERR_DEVICENOTRESET == hr ) { // Reset the D3D environment DestroyD3D9(); hr = InitializeD3D9(); if (FAILED(hr)) return hr; hr = pd3dDevice->TestCooperativeLevel(); if (FAILED(hr)) return hr; } else { ErrMsg("D3D device reported an internal error!\n" "Quiting application.\n"); exit(hr); } } The other issue is minor, when MadVR is first launched, no setting are set in the DS property page. Maybe the default scaling algorithm should be set. My setup Vista 64 SP1 ZoomPlayer Last edited by ericgur; 13th April 2009 at 22:26. Reason: updated setup |
14th April 2009, 00:18 | #253 | Link | |
Registered User
Join Date: Nov 2008
Posts: 64
|
Quote:
has gone up to ~4ms from ~2ms before but "average gpu rendering time" is practically as good as with madVR 0.3. |
|
14th April 2009, 00:27 | #254 | Link | |||
Registered User
Join Date: Sep 2004
Posts: 1,295
|
Quote:
Quote:
For now all users should use the same gamma curves for input and output. This is a subject that should be discussed carefully, so let's wait I have the time for starting the cr3dlut thread... Quote:
|
|||
14th April 2009, 01:44 | #255 | Link | |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Quote:
I got a test rec709 MKV(made by Kazuya), I will put it through ddcc()/t3dlut()/MVR tomorrow : http://rapidshare.com/files/221027650/rec709.mkv.html anyway, I've watched a 2H movie w/ Reclock in 48Hz, the colorimetry was a bit off but it was hell smooth and very sharp, way to go! OTOH some BD's seem very hard to stabilize w/ CoreAVC+all my PP, but as soon as I turn CUDA on...hell breaks loose, I get tearing at the third quarter bottom of the screen(and reseeking does not help). do any of the stats look alarming? it plays fine in HR(RGB32). Last edited by leeperry; 14th April 2009 at 03:23. |
|
14th April 2009, 01:59 | #256 | Link |
In bytes we trust
Join Date: Oct 2002
Location: Torres Beach, Brazil
Posts: 279
|
an 9800GT Pci-E 2.0 @ 16x is not capable of output 60fps video at 1920x1200 screen?
wen using BOB on interlaced material, the playback is jerky. but if not viewing in fullscreen (resizing the window to 75% of display) the playback is smooth. this happens with madVR and haali renderer... only crappy overlay can play full-hd video at 60fps... |
14th April 2009, 03:51 | #258 | Link |
Solaris: burnt by the Sun
Join Date: Oct 2004
Location: /etc/default/moo
Posts: 1,923
|
I'm getting some weirdness with Dark_Shikari's 1080p Big buck bunny
click mitchel resizer i7 920 GTX 280 win7 had to register the ax myself as well
__________________
Last edited by Shinigami-Sama; 14th April 2009 at 03:53. |
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|