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. |
26th January 2014, 17:45 | #21961 | Link |
Registered User
Join Date: Jan 2013
Posts: 18
|
Madshi, just to report my experience:
OS: Win 8.1 GPU: AMD 6570M (rebranded 5730M) CPU: i7 2630QM Drivers: 13.12 video used: NTSC 720x480i Mpeg2 upscaled to 1366x768 Code:
86.11: average stats rendering ~15.00ms present ~0.16ms max stats (5s) rendering ~19.70ms present ~0.35ms GPU usage: ~26% 87.1b: average stats rendering ~5.63ms present ~0.18ms max stats (5s) rendering ~6.96ms present ~0.29ms GPU usage: ~25% NNEDI3: no problems, just gpu too weak to be usable even with 16 neurons Debanding: no problems - gpu: ~31% For me, no problems or bugs so far. Thank you for your great work. |
26th January 2014, 18:48 | #21962 | Link | |
Registered User
Join Date: Sep 2006
Posts: 2,197
|
Quote:
0.8611: without deinterlacing: 50ms with deinterlacing: 35ms + deint: 24ms 0.871b without deinterlacing: 51ms with deinterlacing: 60ms + deint: 10ms edit: well I forgot to disable deinterlacing and stuff in the intel HD settings before I did this comparison. after doing so, rendering times for 0.871b for deint are ~3.30ms
__________________
Laptop Lenovo Legion 5 17IMH05: i5-10300H, 16 GB Ram, NVIDIA GTX 1650 Ti (+ Intel UHD 630), Windows 10 x64, madVR (x64), MPC-HC (x64), LAV Filter (x64), XySubfilter (x64) (K-lite codec pack) Last edited by Thunderbolt8; 26th January 2014 at 19:13. |
|
26th January 2014, 19:05 | #21963 | Link | |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Quote:
Look at Reclock's frame rate detection on VFR videos for example. It takes quite a few seconds for it to detect the VFR framerate changes during playback, and smaller VFR segments go completely unnoticed since they do not have enough frames to accurately average a fps value. Such a delayed detection implementation would not be up to madVR's quality standards, not to mention that smooth motion likely wouldn't take kindly to turning on and off frequently. Last edited by cyberbeing; 26th January 2014 at 20:21. |
|
26th January 2014, 19:41 | #21964 | Link |
*****
Join Date: Feb 2005
Posts: 5,643
|
madshi, can you add a command-line parameter to madHcCtrl.exe that directly opens the settings window and also exits the process when closing the settings window (if the process wasn't already running)?
__________________
MPC-HC 2.1.7.2 |
26th January 2014, 20:28 | #21965 | Link | |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Quote:
Speaking of which, there is a new issue in 0.87.1b with offline setting editing which now causes madHcCtrl.exe to continue running in the background even after the settings window is closed. @clsid You can already do this with the following, but it has the 5 second delay bug: Code:
madHcCtrl.exe editLocalSettings The version without the 5 second delay is the following, but unfortunally madHcCtrl.exe doesn't automatically close with this version: Code:
madHcCtrl.exe editLocalSettingsDontWait Last edited by cyberbeing; 26th January 2014 at 20:43. |
|
26th January 2014, 20:55 | #21966 | Link | |||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Noticed the bug with not closing myself. Should be fixed in v0.87.2. The correct parameter to use should be "editLocalSettingsDontWait". |
|||
26th January 2014, 20:58 | #21968 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
madVR v0.87.2 released
http://madshi.net/madVR.zip Code:
* fixed: D3D9 device wasn't freed properly, resulting in stability problems * fixed: using shared NV12 surfaces resulted in DXVA performance drop * fixed: opening settings without connected decoder took 5 seconds * fixed: chroma Jinc keyboard shortcut activated NNEDI3 * fixed: chroma NNEDI3 keyboard shortcut was not listed in the settings |
26th January 2014, 21:05 | #21969 | Link |
Registered User
Join Date: Mar 2007
Posts: 934
|
Nice work, the performance regression when deinterlacing is fixed for me. Comparable stats to 0.86.11. It looks like I can activate AR for chroma now too, although the GPU is dangerously close to 100% when using this for the hardest scenario (1080i at just under 1080p resolution).
The tray icon doesn't appear any more though, even after clicking the "Show tray icon" button. :/
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7 Last edited by DragonQ; 26th January 2014 at 21:15. |
26th January 2014, 21:16 | #21973 | Link | ||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
You get the "split" queue in the OSD when "copyback" is used. It shows how many frames were already converted to a pixel shader compatible format by "splitting" the NV12 surfaces via copyback. |
||
26th January 2014, 21:22 | #21974 | Link |
Registered User
Join Date: Mar 2007
Posts: 934
|
Don't leave me hangin', I'm gonna need an explanation of what the problem was. I may not be as proficient as yourself but I am a programmer too.
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7 |
26th January 2014, 21:28 | #21975 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
The changelog already explains it: The slowdown was caused by asking D3D9 to allocate *shared* surfaces for DXVA processing. Sharing is necessary to make OpenCL perform well (with AMD and Intel GPUs, NVidia on the other hand doesn't like sharing at all). But it seems that using shared surfaces slows down DXVA decoding/processing a lot under specific circumstances. Which doesn't make much sense to me. But that's the way it seems to be. I'm not sure if this only affects Intel or not. Could be an Intel only problem.
|
26th January 2014, 21:38 | #21976 | Link |
Registered User
Join Date: Mar 2006
Posts: 58
|
Sorry for the noob question, but I read about the number of frames presented in advance being increased to 8 (from 4?) for exclusive mode. I'm looking at the "exclusive mode settings" tab and "how many video frames shall be presented in advance" is set to "4". Saw that on 0.87.1; this is with 0.87.2. GPU queue size is set to "8". Am I looking in the wrong place? Do I need to reset the settings? Normally I just unzip the new version, overwriting without question.
I didn't do nearly the extensive testing as others, but for my usual use on a Radeon HD 6670 (and less used 5670), I didn't notice any performance hit. My default settings have deband on (default setting), Smooth motion FRC on (only if there would be motion judder without it...), and Jinc 3 + AR from chroma and image upscaling. Pretty sure "trade quality for performance" is default (3 checked). nnedi @ 32 neurons was OK if I used lesser scaling (I left deband and Smooth motion on). OpenCL error diffusion proved a little much, though I didn't try disabling debanding and Smooth motion, or going really low with upscaling. I'll likely stick with what I normally use, unless I find time to do some serious testing. I used a 23p SD (probably recorded from DVR using S-Vid to Hauppauge; lo-fi style ;-) file and a 1280x544 23p DVD encode. I have some copy free MPEGs from Comcast (so, interlaced), but haven't checked those yet. core i5, 8 GB RAM, always software decoding. 1920x1080 60 Hz LG LED TV. |
26th January 2014, 21:51 | #21979 | Link |
Registered User
Join Date: Feb 2004
Posts: 399
|
madshi, I'm uncertain if it's still expected at this point with 0.87.2 but I'm still getting a blackscreen with all videos with barebone settings (using nvidia 332.21).
__________________
XP SP3 / Geforce 8500 / Zoom Player |
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
Thread Tools | Search this Thread |
Display Modes | |
|
|