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.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 21st July 2019, 04:38   #56901  |  Link
tp4tissue
Registered User
 
tp4tissue's Avatar
 
Join Date: May 2013
Posts: 706
Quote:
Originally Posted by ryrynz View Post
I think the best target is 4K. videos a@ 1080@24fps, 4k@24fps and 4K at 60fps and HDR tone mapping for 4K@60fps possibly as well.
Will wait a couple of weeks for you to create the thread and we'll go from there.
Problem is variable performance within the same video. It's hard to pick the exact read spot for the video because it's unstable.

There also isn't enough resolution in the timer to tease out differences between cards that are close.
__________________
Ghetto | 2500k 5Ghz
tp4tissue is offline   Reply With Quote
Old 21st July 2019, 04:49   #56902  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
i have an idea for a rough test.

there is nothing worse then over doing the settings and making it to complicated and it doesn't have to be very accurate.
huhn is offline   Reply With Quote
Old 21st July 2019, 04:50   #56903  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Yeah. That's why we'll just use screenshots of graphs from the log data and videos that are long enough to give good averages. I'm thinking about 2 mins. If you want to discuss feel free to msg me.

Actually let's just start a discussion thread first then once we've defined parameters then we'll make a new one for results.
ryrynz is offline   Reply With Quote
Old 22nd July 2019, 14:37   #56904  |  Link
mclingo
Registered User
 
Join Date: Aug 2016
Posts: 1,348
I was trying to think of a dirty way of creating a tool or at least something that was portable, something we could all use, something that forces us all to use the same settings at least, other than just using stock settings. This idea might be possible but its dirty as hell but could be refined if someone knows how to create scripts and compile exe's.

someone could create an extracting exe to do something like the following

Creates a folder c:\MADVR TEST
Runs a script which queries the current location of MADVR install and renames is MADVR.old
extracts a proconfigured MADVR install to c:\MADVR TEST\MADVR
extracts a proconfigured portable MPC-HC program to c:\MADV TEST\MPC
extracts a video file to c:\MADVR TEST
Runs a script which installs MADVR
Runs a scripts which runs the video file so user can take a screen shot
When movies closes a script renames MADVR.old back to MADVR and installs it again.
Script delete c:\MADVR TEST

Would this work though, would you be able to run MPC portable without setting the decoder or configure anything in MADVR?
__________________
LG OLED EF950-YAM RX-V685-RYZEN 3600 - 16GBRAM - WIN10 RX 5700 - https://www.videohelp.com/software/madVR/old-versions
mclingo is offline   Reply With Quote
Old 23rd July 2019, 00:07   #56905  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
I wouldn't put too much effort into it right now as there's a chance madshi might re-release his benchmarking version sometime in the near future. The goal imo is to simply determine average performance, not to make a quick n dirty tool providing more accurate and comparable results. We've made do for this long.. Just need to decide on content, settings, tools are reporting method.
ryrynz is offline   Reply With Quote
Old 23rd July 2019, 09:20   #56906  |  Link
HillieSan
Registered User
 
Join Date: Sep 2016
Posts: 176
👍 sounds good.
HillieSan is offline   Reply With Quote
Old 23rd July 2019, 13:23   #56907  |  Link
mclingo
Registered User
 
Join Date: Aug 2016
Posts: 1,348
I hope madshi does re-release the tool, I would say its more essential now than it has ever been due to the sheer number of cards and driver version available, it would be good to compare results from different driver versions as well as cards !
__________________
LG OLED EF950-YAM RX-V685-RYZEN 3600 - 16GBRAM - WIN10 RX 5700 - https://www.videohelp.com/software/madVR/old-versions
mclingo is offline   Reply With Quote
Old 23rd July 2019, 14:41   #56908  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
benchmark by doing as many frames as possible is not reliable test too each shader has an overhead of about 0.30 ms so a high end card would loose more performance to that then a try level card.
huhn is offline   Reply With Quote
Old 23rd July 2019, 17:45   #56909  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
If you log with a monitoring tool (I've read madVR doesn't like GPU-Z and the like but I've personally never witnessed any difference in performance when it's running) and multiply the GPU clock by the load percentage, then divide that number by the max boost clock (which can usually be found in reviews, or just max out your madVR settings to find it), you can get an estimation of the relative work that is needed from the GPU for a certain madVR config.
It's not a 100% accurate evaluation but it's closer than "this GPU can probably do X scaling algo at Y source and Z output resolution"
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 23rd July 2019, 17:49   #56910  |  Link
HillieSan
Registered User
 
Join Date: Sep 2016
Posts: 176
Quote:
Originally Posted by huhn View Post
benchmark by doing as many frames as possible is not reliable test too each shader has an overhead of about 0.30 ms so a high end card would loose more performance to that then a try level card.
I don’t understand your logic. The performance of the NGU algorithm translates to a benchmark of many frames as possible for certain “ideal” settings. I can imagine that an expensive graphics card may not be the winner. Let the tool determine the best cards for the NGU algorithm (=madvr) and settings. Just my 2 cents.
HillieSan is offline   Reply With Quote
Old 23rd July 2019, 18:44   #56911  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
the 2 finalsteps currently take about 2 ms.
to make math simple if you manage to reach 250 FPS with the giving setting you are currently loosing 50% of the performance to the finalsteps this is not an accurate representation of the card.
that a problem in game benchmarking too but it is far far less of an issue there we have way more steps then the finalsteps.
Quote:
Originally Posted by el Filou View Post
If you log with a monitoring tool (I've read madVR doesn't like GPU-Z and the like but I've personally never witnessed any difference in performance when it's running) and multiply the GPU clock by the load percentage, then divide that number by the max boost clock (which can usually be found in reviews, or just max out your madVR settings to find it), you can get an estimation of the relative work that is needed from the GPU for a certain madVR config.
It's not a 100% accurate evaluation but it's closer than "this GPU can probably do X scaling algo at Y source and Z output resolution"
if my card jumps between 1000 and 700 mhz the load is going crazy i would get number apart by more then 100 % for the same load making the matrix pretty inaccurate at best.
huhn is offline   Reply With Quote
Old 23rd July 2019, 18:44   #56912  |  Link
Bryanhoop
Registered User
 
Join Date: May 2011
Posts: 8
My GPU can't keep up with NGU doubling on low when playing low-resolution 50fps PAL content (Antiques Roadshow), but is fine playing basically all other content at even NGU high.

Is there a way to modify the upsampling algorithm when playing >=50fps content?
Bryanhoop is offline   Reply With Quote
Old 23rd July 2019, 18:50   #56913  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
check out the render profiles on page one 3rd post.
huhn is offline   Reply With Quote
Old 23rd July 2019, 19:28   #56914  |  Link
el Filou
Registered User
 
el Filou's Avatar
 
Join Date: Oct 2016
Posts: 896
Quote:
Originally Posted by huhn View Post
if my card jumps between 1000 and 700 mhz the load is going crazy i would get number apart by more then 100 % for the same load making the matrix pretty inaccurate at best.
It keeps changing that much? That's interesting, on Adaptive mine always stabilizes after less than a minute on some clock/load combination, even with light stuff.
Maybe rendering and queue settings influence this.
__________________
HTPC: Windows 10 22H2, MediaPortal 1, LAV Filters/ReClock/madVR. DVB-C TV, Panasonic GT60, Denon 2310, Core 2 Duo E7400 oc'd, GeForce 1050 Ti 536.40
el Filou is offline   Reply With Quote
Old 24th July 2019, 04:26   #56915  |  Link
Shiroe
Registered User
 
Join Date: May 2019
Posts: 8
Is there a madVR setting that increases playback FPS? When using LAV Filters Megamix mid-tier (NGU medium) preset, PotPlayer's OSD information displays this:

https://i.postimg.cc/Zb7hVpst/zuikaku-S9-YJb-KMTj-M.jpg

One line reads:

"Output: P010, 2048x1080(1.9:1/1.78:1), FPS: 23.976->59.94"

If I use another video renderer, e.g., EVR-CP, the FPS would read:

"23.976->23.976"

Just in case, madVR's OSD information displays this:

https://i.postimg.cc/Bs7mQM1f/zuikaku-c6-Xopp-Q0o2.jpg

If a madVR setting is increasing playback FPS, is that setting affected by the Windows version?

The reason I ask is using LAV Filters Megamix's mid-tier (NGU medium) preset on Windows 10 to watch anime gave me 60 FPS. I switched to Windows 8.1 for reasons, and the same preset gave me 30 FPS although now it is giving me 60 FPS, even though I made no changes.
Shiroe is offline   Reply With Quote
Old 24th July 2019, 06:13   #56916  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Your refresh rate is 60Hz (you are asking your display to show 60fps), madVR in your particular case is blending frames AKA Smooth Motion (Seen enabled in your OSD)
If you are not blending frames this way say with another renderer, then you have judder as a result of having a 3:2 pulldown as frames are displayed in a 3-2 pattern.
Keep Smooth Motion on and enjoy, madVR is the best renderer available for your video content so stick with it.
ryrynz is offline   Reply With Quote
Old 24th July 2019, 08:00   #56917  |  Link
HillieSan
Registered User
 
Join Date: Sep 2016
Posts: 176
Quote:
Originally Posted by huhn View Post
the 2 finalsteps currently take about 2 ms.
to make math simple if you manage to reach 250 FPS with the giving setting you are currently loosing 50% of the performance to the finalsteps this is not an accurate representation of the card.
that a problem in game benchmarking too but it is far far less of an issue there we have way more steps then the finalsteps.

if my card jumps between 1000 and 700 mhz the load is going crazy i would get number apart by more then 100 % for the same load making the matrix pretty inaccurate at best.
I see. I remember that on my RX480 madvr did it’s job fine when I keep the latency below 15 ms (not sure if it was exactly 15 ms). Any swing of the clock didn’t matter as long as it stays around 15 ms. Is this useful for a benchmark? I had to keep the error dithering simple (I think ordered setting) to allow NGU high for most movies. This helped to drop the latence under the 15 ms. This latence was my reference for performance tuning.

Last edited by HillieSan; 24th July 2019 at 08:34.
HillieSan is offline   Reply With Quote
Old 24th July 2019, 09:05   #56918  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Like Nev said the benchmark runs the card at full speed so clock speeds won't be a concern. Mostly we just want to test/compare NGU and HDR algorithm performance.
ryrynz is offline   Reply With Quote
Old 24th July 2019, 10:27   #56919  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by ryrynz View Post
Your refresh rate is 60Hz (you are asking your display to show 60fps), madVR in your particular case is blending frames AKA Smooth Motion (Seen enabled in your OSD)
Or switch refresh rate of the display to 23p if that's possible and turn smooth motion off.
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config
chros is offline   Reply With Quote
Old 24th July 2019, 12:08   #56920  |  Link
Shiroe
Registered User
 
Join Date: May 2019
Posts: 8
Quote:
Originally Posted by ryrynz View Post
Your refresh rate is 60Hz (you are asking your display to show 60fps), madVR in your particular case is blending frames AKA Smooth Motion (Seen enabled in your OSD)
If you are not blending frames this way say with another renderer, then you have judder as a result of having a 3:2 pulldown as frames are displayed in a 3-2 pattern.
Keep Smooth Motion on and enjoy, madVR is the best renderer available for your video content so stick with it.

So that is why I had judder when using video renderers other than madVR. madVR also has a setting to change automatically display modes. Can that setting replace Smooth Motion, and is it better or worse than Smooth Motion, which is to ask does Smooth Motion have downsides?
Shiroe is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 21:47.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.