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 7th January 2019, 21:18   #54141  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,406
Quote:
Originally Posted by Damien147 View Post
Which one of these three gpu load readings I can trust?Gpuz gpu load goes up and down as you see and the other two are stable.
madVR load is odd, once the buffer is full it runs as fast as possible for one frame and then waits for the next frame. This causes aliasing artifacts that change with the polling rate which can make it look like the load is fluctuating a lot.
__________________
madVR options explained
Asmodian is offline   Reply With Quote
Old 7th January 2019, 21:42   #54142  |  Link
Damien147
Registered User
 
Join Date: Mar 2011
Posts: 380
All three were different and got confused.
I guess the answer is that you can't measure it in % accurately then.I'll measure it in ''dropped frames''
Thanks.
Damien147 is offline   Reply With Quote
Old 7th January 2019, 21:44   #54143  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,406
If you set the polling rate really fast you can get a better idea of the real load, something like 100 ms.
__________________
madVR options explained
Asmodian is offline   Reply With Quote
Old 7th January 2019, 22:03   #54144  |  Link
Damien147
Registered User
 
Join Date: Mar 2011
Posts: 380
I see.No big deal,just curious because previous gpu years ago was giving me more stable readings without spikes in gpu z.Thanks again.
Damien147 is offline   Reply With Quote
Old 7th January 2019, 23:09   #54145  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,342
Depending on how the workload is structured, you would want an average load that takes into account bursts and idle periods. GPU-Z seems to not do that. Spiky 0/100 graphs are not very useful.

madVR is rather special in how it works load-wise, because it has practically no CPU limit, so when it can work it'll work 100%, but it has a hard frame rate limit, so it doesn't have endless work - ie. very spiky behavior. An ordinary 3D renderer would be CPU limited and/or capable of rendering an unlimited amount of frames, so it would either always cap the GPU, or be limited consistently by the CPU, but not simply by "time".
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is online now   Reply With Quote
Old 8th January 2019, 00:21   #54146  |  Link
Damien147
Registered User
 
Join Date: Mar 2011
Posts: 380
Ok,I'll stay with ''no dropped frames,all good''.Thank you too.
Damien147 is offline   Reply With Quote
Old 8th January 2019, 01:49   #54147  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
Quote:
Originally Posted by chros View Post
Definitely not, although I haven't seen your TV. Use passthrough then.
"Definitely not" but I don't really know for sure. Lol.

If configured right I would expect better results but a 1060/1070/2060 would really help a lot.
ryrynz is offline   Reply With Quote
Old 8th January 2019, 08:25   #54148  |  Link
madjock
Registered User
 
Join Date: May 2018
Posts: 259
Quote:
Originally Posted by ryrynz View Post
"Definitely not" but I don't really know for sure. Lol.

If configured right I would expect better results but a 1060/1070/2060 would really help a lot.
Haha, well I have mostly given up after trying all the different options, I could never really get much below 36ms at a base level, with obvious spiking and jumps into stutter levels or having it at dumb.

Just can't justify a Graphics Card more expensive than the TV I just got, so it is what it is.
madjock is offline   Reply With Quote
Old 8th January 2019, 10:14   #54149  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,646
That's a shame, with 2060 out now 1060's should start to lower in price so if you're okay with second hand jump on one, preferably the 6GB model.
ryrynz is offline   Reply With Quote
Old 8th January 2019, 10:49   #54150  |  Link
chros
Registered User
 
chros's Avatar
 
Join Date: Mar 2002
Posts: 2,323
Quote:
Originally Posted by zapatista View Post
Q3) when playing 4k files (with my limited settings ) my GPU reports 72 or 75% GPU load, and 93 % vram used for ex, but my i7 cpu is only at 20 or 25% load.
- is there a way i can offload some of the video tasks more to the CPU and then be able to enhance other settings in madvr that are GPU load dependent ?
i have tried different "video decoder" selections in mpc-hc, but it doesnt seem to alter this (am using dxva2-copyback as this was suggested by you in an earlier setup guide iirc)
1060 3gb GPU
What's checked in trade performance options? If nothing then check only "scale chroma separately..." option, it should improve things a lot.
Also use DXVA11 native in LAV filters not copyback, it's way faster.
Although the 3GB version of 1060 has less shaders than the 6GB one but I think it should handle 4k hdr well on 1080p screen. But you'll see

Quote:
Originally Posted by ryrynz View Post
"Definitely not" but I don't really know for sure. Lol.
I meant exactly this for a starter:

Quote:
Originally Posted by SamuriHL View Post
IMO you're better off with passthrough if you enable dumb tonemapping in the trade performance options.
Quote:
Originally Posted by ryrynz View Post
if you're okay with second hand jump on one, preferably the 6GB model
I second that, just buy a used 1060 6GB model (it has more shaders (=more powerful) than the 3GB variant) that's what I did. If noise level is important than go for MSI GTX 1060 Gaming X 6GB (although probably that's the most expensive card out of all of them).
__________________
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

Last edited by chros; 8th January 2019 at 10:54.
chros is offline   Reply With Quote
Old 8th January 2019, 10:55   #54151  |  Link
glc650
Registered User
 
Join Date: May 2003
Posts: 77
video card upgrade = dozen of presentation glitches every second

Hi,

I just upgraded from a Gigabyte Radeon HD 7870 to a XFX Radeon RX 580 GTS and now I get dozens of presentation glitches every second when playing any video if the HDMI Scaling slider is set above 0% in the AMD Radeon Settings GUI.

I tried updating the Radeon drivers to 18.12.2 (was on 17.12.1 with the old card). but still get the issue. I've also tried restoring madVR's default settings. And I tried it with the same madVR settings I used with the previous card. I also tried the previous version of madVR.

Never had this issue with the 7870 and I had the HDMI scaling slider set to 3% (which is where I also need it set for the new card).

Thanks,

>g.
glc650 is offline   Reply With Quote
Old 8th January 2019, 13:03   #54152  |  Link
actarusfleed
Registered User
 
Join Date: Jun 2009
Posts: 71
Did someone ever compared (only with madvr) a gtx 1070ti vs rtx 2070?

Some YouTube videos are about this but only on the videogame side...

I'm only interested to evaluate this by the madvr point of view...
actarusfleed is offline   Reply With Quote
Old 8th January 2019, 16:46   #54153  |  Link
tp4tissue
Registered User
 
tp4tissue's Avatar
 
Join Date: May 2013
Posts: 706
Quote:
Originally Posted by actarusfleed View Post
Did someone ever compared (only with madvr) a gtx 1070ti vs rtx 2070?

Some YouTube videos are about this but only on the videogame side...

I'm only interested to evaluate this by the madvr point of view...

Peeps are just cranking the upscaling to arbitrarily high setting to saturate their cards.

I'm not sure we can call that Performance..

at 2 meters from my tv, I can't even see the difference between bilinear chroma and nnedi..

I guess you could make a case for scaling 4K up to 8K tvs..

I have an old 7970, and it can do 4K jinc chroma, shader to SDR in 15ms..
__________________
Ghetto | 2500k 5Ghz

Last edited by tp4tissue; 8th January 2019 at 16:50.
tp4tissue is offline   Reply With Quote
Old 8th January 2019, 17:39   #54154  |  Link
tp4tissue
Registered User
 
tp4tissue's Avatar
 
Join Date: May 2013
Posts: 706
Quote:
Originally Posted by huhn View Post
do you have a screen that can do more than DCI P3? well no so you don't have to care the 3D LUT will do the gamut mapping anyway. just use the default setting.


clearly not 10000 so feel free to use the measured peak brightness.
i personally don't think this is a useful measuring anyway on an x900f the peak brightness for 10% is far higher than 2%.
You misunderstood my question huhn..

My question is,

How does changing the Sent meta data from rec2020 to dci-p3 change the image, if at all.

How does changing the peak nit part of the meta data affect the image..

If, there is a change at all, what does the tv DO with that meta data, does it pop into different modes.
__________________
Ghetto | 2500k 5Ghz

Last edited by tp4tissue; 8th January 2019 at 17:41.
tp4tissue is offline   Reply With Quote
Old 8th January 2019, 18:01   #54155  |  Link
SirSwede
Registered User
 
Join Date: Nov 2017
Posts: 69
Quote:
Originally Posted by tp4tissue View Post
Peeps are just cranking the upscaling to arbitrarily high setting to saturate their cards.

I'm not sure we can call that Performance..

at 2 meters from my tv, I can't even see the difference between bilinear chroma and nnedi..

I guess you could make a case for scaling 4K up to 8K tvs..

I have an old 7970, and it can do 4K jinc chroma, shader to SDR in 15ms..
Still on Bicubic 75 for Chroma here; never changed it. Looks splendid on my TV. Two metres here as well.

But, then again, it is madVR - after all.
SirSwede is offline   Reply With Quote
Old 8th January 2019, 19:47   #54156  |  Link
border.community
Registered User
 
Join Date: Dec 2018
Posts: 10
I am mostly playing UHD mkv rips and was wondering which madvr settings actually affect the picture since I am neither upscaling or downscaling.

Seems like only the settings under Devices, Processing and Rendering would make a difference, and Scaling Algorithms would not---but if I were to change NGU Anti-Alias to "High quality" it would obviously affect performance which begs the question: what are Scaling Algorithms doing to files that do not need to be scaled?
border.community is offline   Reply With Quote
Old 8th January 2019, 19:52   #54157  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
Chroma still needs to be scaled. It's not about resolution. Chroma data is stored in 4:2:0 and needs to be scaled to 4:4:4 or 4:2:2 depending on what you're sending to your display. However, whether you can tell the difference between NGU High AA or bicubic75 or not is an individual thing and only you can decide with your eyes if you can see a difference.
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 8th January 2019, 20:15   #54158  |  Link
border.community
Registered User
 
Join Date: Dec 2018
Posts: 10
Quote:
Originally Posted by SamuriHL View Post
Chroma still needs to be scaled.
Ah, thanks. I have a 1050ti and a C8 as well. Chroma upscaling is really the only thing I have turned on. Do you have anything else turned on, such as any of the Processing options?
border.community is offline   Reply With Quote
Old 8th January 2019, 20:58   #54159  |  Link
SamuriHL
Registered User
 
SamuriHL's Avatar
 
Join Date: May 2004
Posts: 5,351
I've got a 1060 6gb and I've got quite a few things enabled yes. I've had to change some of my settings recently because of new HDR tonemapping options that will be in the upcoming madvr release. It's a little more punishing than it is in the current release build. It's always a trade off of performance vs quality which is why you see so many people shelling out cash for 1080's and 2080's to make sure they have some breathing room. Everyone's configuration in madvr is going to be different based on what content they're viewing and what they're looking for in terms of quality.
__________________
HTPC: Windows 11, AMD 5900X, RTX 3080, Pioneer Elite VSX-LX303, LG G2 77" OLED
SamuriHL is offline   Reply With Quote
Old 8th January 2019, 21:51   #54160  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,903
Quote:
Originally Posted by tp4tissue View Post
How does changing the Sent meta data from rec2020 to dci-p3 change the image, if at all.
if your screen can do bt 2020 natively it would do something if it is close to DCI P3 it should do nothing at all.
Quote:
How does changing the peak nit part of the meta data affect the image..
Quote:
If, there is a change at all, what does the tv DO with that meta data, does it pop into different modes.
who knows literally what ever they want. HDR tonemapping doesn'T have a real one way is correct spec.
the difference should be the image dynamic until it hits it max brightness then it does what ever it wants.

as i said before there is no clear answer.
Quote:
Originally Posted by SamuriHL View Post
Chroma still needs to be scaled. It's not about resolution. Chroma data is stored in 4:2:0 and needs to be scaled to 4:4:4 or 4:2:2 depending on what you're sending to your display.
and in madVR case it will always be scaled to 4.4:4 and only RGB is send to the GPU driver that's why sending 4:2:2 or 4:2:0 is bad for the applied dithering and chroma quality the driver has to RGB -YCbCr convert which alone creates floatpoint numbers downsacle the chroma which again results in float point numbers. chroma sub sampling is just storing chroma in a lower resolution than luma so it's pretty much resolution.

and as a fun fact 8 bit RGB has more BPP as 10 bit YCbCr 4:2:2.

chroma is for most scenes not important up to a point that bilinear will be enough but as soon as an image get's chroma heavy or even based (red text on black background as an example) it shows significant differences especially on BDs where you are now literal using an good scaler on SD.

Last edited by huhn; 8th January 2019 at 21:55.
huhn 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 14:44.


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