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. |
3rd March 2013, 18:41 | #17881 | Link |
Registered User
Join Date: Jan 2008
Posts: 589
|
I would apply the 3D-LUT independently (i.e. measure without 1D-LUTs and apply without 1D-LUTs by using the "disable gamma ramps" option). Doing everything in one processing step is better in terms of precision than doing it in two separate steps (especially when there are integer conversions in the middle).
|
3rd March 2013, 19:59 | #17882 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
I have a questions people. Do anyone know which GPU would be powerful enough to use with smooth motion + frame blending up to 30 fps content without dropped frames? I hear the GTX 260 can't save us this time.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
3rd March 2013, 20:04 | #17883 | Link |
Registered User
Join Date: Oct 2012
Location: Akron, OH
Posts: 491
|
Now that madVR includes a Smooth Motion option, how does that affect the use of SVP? I asked the same question on the SVP forum, but I'm wondering if I still need SVP or if madVR has made it obsolete?
|
3rd March 2013, 20:08 | #17884 | Link | |
Audiophile
Join Date: Oct 2006
Posts: 353
|
Quote:
SVP actually adds extra frames and makes it more smooth. |
|
3rd March 2013, 21:40 | #17885 | Link | |
Registered User
Join Date: Jan 2007
Posts: 130
|
Quote:
__________________
Win 7 x64, AMD Phenom II X6 1075T, nvidia 430GT, mpc-be x86 |
|
3rd March 2013, 23:12 | #17886 | Link | |
Registered User
Join Date: Sep 2012
Posts: 174
|
Quote:
but if you want motion interpolation (more fps) then you'll still have to use SVP... |
|
3rd March 2013, 23:40 | #17887 | Link | |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Quote:
If the person is using stock (non-OC) CPU and GPU clocks, has low DPC latency, and is unable to troubleshoot the issue by using less intensive scaling settings, or tweaking madVR's numerous other options, they should file a bug report on madshi's bugtracker. Overall, I don't think you should even try to make any sort of broad determination about Smooth Motion until madshi is satisfied that he's fixed all the critical bugs, and enables by default in a future release. At the moment, I'd say the feature is still in the user feedback and testing phase. |
|
4th March 2013, 00:38 | #17888 | Link |
Registered User
Join Date: Jan 2009
Posts: 1,210
|
To be frank, I think FRC is awesome. Hopefully with some more improvements it will become a standard, nice alternative to using SVP and avisynth for realtime playback. FRC definitely looks closer to SVP than just refresh rate multiple and reclock, at least if running at < 60Hz (48Hz for 24fps).
|
4th March 2013, 01:36 | #17889 | Link |
Registered User
Join Date: Mar 2007
Posts: 934
|
I plan on using a cheap monitor as a second TV in a year or so (connected to a PC running MediaPortal in client mode). It only does 60 Hz AFAIK though so MadVR's smooth video option will be vital. That gives the MediaPortal devs a year to add MadVR support. Not very likely but you never know.
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7 |
4th March 2013, 02:19 | #17890 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
BTW, talking about 8bit LUT's, 60Hz-only monitors and all I just found out that LG's 170€ M2232D / 200€ M2432D 1080p serie is "ISF ccc" certified so you get all the colorimetry options a colorfreak could ever crave, has a MPEG4 DVB-T tuner onboard and also supports 24/30/50/60Hz
The 27" version uses a 600:1 CR IPS panel and these are 1K:1 TN of course, but bang/bucks looks huge! Their built-in dynamic contrast algorithm looks extremely impressive to me(not as artificial as Sammy's DNIE) and the 24p demo BD in the store looked as smooth as it gets.....I'll prolly get ahold of the 24" version tomorrow I'll lose a hell lot of native CR compared to CRT but I'll gain perfect geometry/sharpness, perfect colorimetry(without a nasty 8bit LUT), most impressive dynamic CR and if I'm lucky clouding won't as much of an issue on 20" LCD's as it on 40"s. Also, 24p should be a lot less blurry on a 5ms TN panel than on a +16ms MVA. BTW, just for the record I updated the XLS gamut mapping spreadsheet with 6 digits after the coma for the standards gamuts white points instead of the stock 4 digits. I'm attaching it to this post. Last edited by leeperry; 4th March 2013 at 05:06. |
4th March 2013, 13:46 | #17891 | Link | |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Quote:
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
|
4th March 2013, 14:10 | #17892 | Link |
Registered User
Join Date: Dec 2008
Posts: 496
|
Of course it does. Frame blending is extremely expensive. But why would you use FRC and frame blending together, I don´t get it. Please give us a technical explanation what you´re trying to achieve with that.
Last edited by iSunrise; 4th March 2013 at 14:13. |
4th March 2013, 14:23 | #17893 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
I'm sure madshi said that disabling the option "don't user linear light for smooth motion frame bleding" in rendering options would make the quality go up. Unless I misread that, it does make it look a little better here. But again, I can't use it because of the frame drops.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
4th March 2013, 15:19 | #17894 | Link | |
Registered User
Join Date: Dec 2008
Posts: 496
|
Quote:
Enabling or disabling that option on a GTX580 with 2D clocks (Core 405MHz/Memory 162.0MHz/Shaders 810MHz) doesn´t make any difference in utilization here. It stays the same at 50% GPU Load and 38% Memory Controller Load. Utilization is at 30% GPU Load and 20% Memory Controller Load without smooth motion. I am not sure why you are having problems with that. Your GPU should be fast enough. Check your GPU utilization with GPU-Z and post a screenshot here. Last edited by iSunrise; 4th March 2013 at 15:24. |
|
4th March 2013, 16:10 | #17895 | Link | |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Quote:
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
|
4th March 2013, 16:24 | #17896 | Link | |
Registered User
Join Date: Dec 2008
Posts: 496
|
Quote:
Last edited by iSunrise; 4th March 2013 at 16:27. |
|
4th March 2013, 16:32 | #17897 | Link | |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Quote:
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
|
4th March 2013, 17:50 | #17898 | Link |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Doing a bit more testing with Smooth Motion right now, it would appear 0.86.1 caused a minor performance regression with smooth motion + windowed mode + high framerate video on my GT440 DDR5.
With 0.86.1 my GT440 DDR5 bottlenecks with 50fps content in windowed mode. Backbuffer queue drops to 0-4/4, Render Queue drops to 2-4/16 when the bottleneck occurs along with dropped frames. With 0.86.0 my GT440 DDR5 is fine with 50fps content, but bottlenecks with 60fps content in windowed mode. Backbuffer queue fluctuates between 0-4/4 & 3-4/4, Render Queue drops to 4-6/16 when the bottleneck occurs along with dropped frames. Setting Windowed Mode "after intermediate render step" to flush & wait seems to counteract the dropped frames by preventing the Backbuffer Queue from falling to 0. madVR 0.86.1 (Backbuffer 1-3/4) is still slower than 0.86.0 (Backbuffer 2-3/4) even with this change. Yet setting flush and wait there creates a different bottleneck to my Render Queue on high framerate video when not using Smooth Motion, so it's not really a practical workaround. Fullscreen Exclusive Mode is unaffected by this issue. For whatever reason, Fullscreen Exclusive GPU load is 15% lower compared to Windowed when scaling the SmoothMotion59.ytp patten with Smooth Motion enabled. Nothing really noteworthy about this other than my GPU just being too weak, and quick to become bottnecked as the number of flushes performed by madVR in the render pipeline increase. |
4th March 2013, 18:20 | #17899 | Link |
Registered User
Join Date: Dec 2012
Location: Neverland, Brazil
Posts: 169
|
Oh yeah, I know this is almost totally unrelated to madVR but I really need some help with this.
When I use ReClock and I enable the "PAL SpeedDown (force to 24 fps)" option (screenshot), my audio gets in some sort of slow motion and it's really annoying. I've already set madVR to treat 25p as 24p, and I've tried numerous ReClock options to see if I could fix that but nothing. The only way to not make that happen is to manually set the speed to "24p" in the media stream when the video is being played (screenshot). It's the same window where you set "PAL SpeedDown (force to 24 fps)". For reference, I have a 60hz screen.
__________________
madVR scaling algorithms chart - based on performance x quality | KCP - A (cute) quality-oriented codec pack |
4th March 2013, 20:10 | #17900 | Link |
Registered User
Join Date: Jan 2008
Posts: 589
|
I'm not sure what the fuss is all about regarding FRC performance. I just tested FRC on a Intel HD 4000 with a 1080p video (and a 1080p screen @ 60Hz) and it just works, with no frame drops. And that's with the "trade quality for performance" option disabled (so maximum quality).
|
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|