madVR - high quality video renderer (GPU assisted)

[madshi]

re-downloaded the latest madVR, ran restory default settings.bat again, opened madTPG and the screen turned gray again (not due to 16-235 range)...
I intentionally made the madTPG screen in the latest version slightly gray, as long as no remote control software (like HCFR or ArgyllCMS is connected). That's nothing to worry about.

At first Windows shows me the Security Warning. I press RUN. Then madTPG appears and it has a valid black screen.
It should automatically turn gray if you start it.

Then, I get madHcCtrl.exe Security Warning. As soon as I press RUN - madTPG screen goes from black to gray
That was probably a coincidence. madHcCtrl.exe starting should have no big affect on madTPG.

and then I get Windows Access problem...
Can I see a screenshot of that? I thought that came from ArgyllCMS trying to contact madTPG? Now it sounds as if it had nothing to do with Argyll?

[madshi]

Yes, I cannot do that with the recent builts.

This issue is with madTPG running on its own and with HCFR (I'm not sure how to try while ArgyllCMS are doing automated measurements)...
I suppose this might have to do with the 3dlut file you've loaded. Can you upload the 3dlut somewhere for me to look at? And then please create a new bug report in the madVR bug tracker (link see first post of this thread) and add a link to the uploaded 3dlut file. Thanks.

[Werewolfy]

Did I reply to this one? I think not, but I'm not sure. Just to be safe:

Unfortunately the log doesn't tell Direct3D changes the display mode. Which means that Direct3D does that through an API which the API Monitor didn't hook/watch... :( So back to square one: I can probably only do anything about this if I'm able to reproduce it on my PC, which currently I can't.

No you didn't reply earlier but there was a lot of comments following mine ;)

I will test next versions of API Monitor to see if it will watch that API one day... And I hope that someone else will have the same problem, maybe it will help to find where this problem come from. Fortunately I don't have any problem with Overlay mode, no tearing or frame drops.

[madshi]

P.S:

I can't unpress both of these buttons (at least one of the buttons will be blue no matter what I try). I suspect it has do with the logic controlling the buttons... when disable VideoLUTs is grey (ie the VideoLUT is enabled), and disable 3dlut is blue (ie disabled), pressing the disable 3dlut button toggles the state of Both buttons.
Oh, it seems Asmodian was right about this. See his post here:

http://forum.doom9.org/showpost.php?p=1673587&postcount=24974

Basically if you enable your 3dlut then the 3dlut decides about which GPU gamma ramps are used - because the GPU gamma ramps must fit to the 3dlut if you want to get correct results. 3dluts created by ArgyllCMS have the matching GPU gamma ramps attached. So if you use an ArgyllCMS 3dlut enabled in madVR, your GPU gamma ramps are already decided, too.

Now you're probably wondering why older madTPG builds behaved differently to the latest madTPG build? The answer to this is the following: Older madTPG builds didn't properly recognize whether the GPU gamma ramps attached to the 3dlut were linear/empty or whether they actually contained some sort of modification. So older madTPG builds thought that there were useful GPU gamma ramps attached to the 3dlut, so the GPU gamma ramps button was enabled. The latest madTPG build now detects that the GPU gamma ramps attached to your 3dlut are actually linear/empty. As a result the "disable GPU gamma ramps" button is enforced, to signal that this specific 3dlut only works properly with disabled GPU gamma ramps.

[MistahBonzai]

HD7770 Calayst Beta

http://i.imgur.com/X4blITZ.png

HD7850 Catalyst 14.1 - OpenClSpeedTest Results (http://www.mediafire.com/view/79bk5i5e7etn12v/Speed_Test.PNG)

[Mr Alpha]

I'm having a weird issue with MadVR. Every 5-10 second the render time will spike up to 270ms. This is on Win8.1 with a GTX Titan and the latest Nvidia drivers. Has anyone seen something similar or know what is going on?

[huhn]

I'm having a weird issue with MadVR. Every 5-10 second the render time will spike up to 270ms. This is on Win8.1 with a GTX Titan and the latest Nvidia drivers. Has anyone seen something similar or know what is going on?

if madvr start spiking i reset everything to default and setup all again this normally fix the issue at least for me.

[turbojet]

Quote:
Originally Posted by turbojet View Post
madshi: Switching to pc levels fixes overlay but now window and fse are crushed.
You mean if you switch madVR to *TV* output levels, correct?

Correct, tv levels. Currently can't have correct levels on window/fse and overlay at the same time unfortunately, can on 3 other displays I tested. I prefer window but ivtc+smoothmotion is a blended mess in window although the gpu is only about 50% loaded, switching to overlay fixes it. Is there anything that can be done outside of switching hardware that you can think of?

Quote:
Originally Posted by turbojet View Post
You can't use it with jrmc until it or madvr adds pixel shader support. To use it with mpc-hc copy the script (check my sig) into a txt file, rename it .hlsl and put it in mpc-hc's shaders directory. In mpc-hc right click>shaders>select shaders.

The defaults are horrible, I'd suggest pattern 3 and adjusting strength between 0.5-2.0, increasing clamp improves really fine detail without artifacts if adjusted correctly, I use 0.5.
Someone needs to start a LumaSharpen thread. I have so many questions, but this probably isn't the appropriate place to ask.

I've been using 1.5 sharpness, .05 clamp, and pattern 3. I'd really like to understand what makes this pattern better or what the optimal settings are in general and the best way to test.

The developer, ceejay.dk, mentioned he was going to start a thread but it looks like that's not going to happen. Pattern 3 uses more than twice the taps (17) the other patterns do so maybe that's responsible for being able to use higher strengths without artifacts.

While irrelevant now I have a feeling lumasharpen will be relevant to madshi at some point. On that note, madshi, do you plan on posting pixel shaders at some point? Would like to use debanding on i3 laptop, it can't handle madvr with 720p60/1080i but evr works.

Quote:
Originally Posted by agustin9 View Post
That's how composition works, but I don't think it bothers madvr
This is correct, it is an issue with Windows 7 that is not present in Windows 8.1.

I understand a v-sync is forced at the composition rate by Aero so you might try the "disable desktop composition" option if you are experiencing issues caused by the composition rate.

Removing composition in windows 8 and still looking good with all the features is about the only improvement worth anything in windows 8 over 7 but it's big enough to consider upgrading imo. Too bad you have to spend a bunch of time tweaking windows 8 to get it desktop usable and there's still some bugs.

[Ceremony]

I'm having a weird issue with MadVR. Every 5-10 second the render time will spike up to 270ms. This is on Win8.1 with a GTX Titan and the latest Nvidia drivers. Has anyone seen something similar or know what is going on?make sure this spike isnt caused by some powerstate change of the card. just overwrite it and force P0 and check, whether that fixes it?

[nevcairiel]

Removing composition in windows 8 and still looking good with all the features is about the only improvement worth anything in windows 8 over 7 but it's big enough to consider upgrading imo. Too bad you have to spend a bunch of time tweaking windows 8 to get it desktop usable and there's still some bugs.

You do realize that Windows 8 still has desktop composition, and in fact they even force it on all the time, there is no way to turn it off anymore (like there was in 7) ? :)
Microsoft did improve the entire DWM (Desktop Window Manager, the compositor) though.

[turbojet]

Maybe it's aero then, madvr doesn't report composition rate which means no more dropped frames because composition and display rate differ, multiple displays with different refresh rates work fine, chrome no longer breaks when changing refresh rates, remote access tools no longer have to change the look to connect, dwm no longer uses gpu. All these things are true when composition is disabled in windows 7 and in windows 8...

[QBhd]

Wow... my results seem to be vastly different with my R9 270x:

http://i.imgur.com/aKE7CLZ.png

I just don't understand some of these differences?

QB

[madshi]

Correct, tv levels. Currently can't have correct levels on window/fse and overlay at the same time unfortunately, can on 3 other displays I tested. I prefer window but ivtc+smoothmotion is a blended mess in window although the gpu is only about 50% loaded, switching to overlay fixes it. Is there anything that can be done outside of switching hardware that you can think of?
Well, it seems that only madVR Overlay gives you the best image quality. According to what you're reporting, all other combinations (madVR windowed, madVR FSE, non-madVR-Overlay) result in NVidia stretching madVR's output behind madVR's back - probably in 8bit without dithering.

To be honest, I don't want to add any additional options to madVR just to work around an NVidia driver issue which only occurs with svideo, considering that nobody really uses svideo today, anymore.

I think you have 2 options: (1) Either disable FSE mode and always use Overlay mode. This made actually work fine, if you're lucky. Or (2) Enable FSE mode and always use FSE mode. This is not as comfortable, of course, and it comes with the disadvantage that NVidia's level manipulations will probably introduce banding artifacts. So I'd recommend to try (1).

Wow... my results seem to be vastly different with my R9 270x

I just don't understand some of these differences?
I don't understand it, either. The interop cost seems to vary greatly. Not sure why it's sometimes "extremely high" and sometimes only "very high". In any case it's clear that there's a serious performance problem which AMD should fix.

[QBhd]

I know... even comparing seiyafan's 270x results to mine is vastly different... Very odd behavior

QB

[turbojet]

I understand not wanting to add a new option in the gui for deprecated output only and agree. I was thinking it was an issue with madvr's overlay since it's correct with other overlay rendering.

Using overlay all the time instead of ~35% of the time it's needed is an option but not a preferred one as it comes with some disadvantages in my case, changes player gui color and disables taskbar preview. Another option would be using 'levels=tv' which I'd probably prefer.

However, would allowing the output range to be changed with profiles be a possibility? This would lead to another solution of nvidia's refresh rate levels issue over hdmi.

[madshi]

I was thinking it was an issue with madvr's overlay since it's correct with other overlay rendering.
Well, there are probably 2 different point of views. But *my* point of view is that if you use madVR you want the max image quality. And you can get that only by stopping the GPU from stretching the rendered video behind the video renderer's back. That is how I define "correct", and based on this definition madVR's overlay is the only "correct" mode you have. By this definition the other overlay renderers produce "incorrect" results.

However, would allowing the output range to be changed with profiles be a possibility? This would lead to another solution of nvidia's refresh rate levels issue over hdmi.
The only proper solution to nvidia's refresh rate levels issue over hdmi is to use the madLevelsTweaker, or similar tools.

It seems you're only looking at whether black & white are sent correctly, and you're ignoring everything else. Of course having black & white correct is the most important thing of all. However, avoiding banding artifacts is important, too. So I will not support half-ass solutions which produce banding artifacts. Switching output levels via profiles would be such a half-ass solution because it would result in some modes producing banding artifacts. Well, maybe there is no other solution for svideo, but for HDMI there certainly is.

That said, currently profiles are not supported in madVR's device setup. This may change in a future version, but definitely not in v0.87.x. And I'm not sure yet if I'll allow profiles for the "properties" page. Then again, it probably wouldn't harm, so I might allow it, after all.

[XMonarchY]

madshi, I have no idea where the problem is, but I do know that madVR 0.87.6 madTPG window works perfectly fine with dispcalGUI and raw ArgyllCMS 1.6.3 commands, but madTPG 0.87.7 shows gray screen after I press Run when Windows Security for madHcCtrl.exe shows up upon launching the file for the first time in whichever folder. I'm sorry for the long run-on sentence...

Here is a screenshot - https://www.dropbox.com/s/cj2hy5mccdl18up/madVR.jpg and it happens regardless of:
-drivers installed (happens in 327.23 too),
-dithering method chosen,
-display device chosen,
-upscaler chosen,
-overlay enabled or not,
-any settings actually...

The "Window Access Failed" shows up in both - raw ArgyllCMS DOS commands and dispcalGUI (any version). I don't know if its madVR, but it seems to be so on my PC... since 0.87.6 and older work perfectly fine. I really have no clue... I also have Windows 8.1 beta Escrow Update 17031 installed, but I had it installed for a long time and it caused no problems before... I doubt its the cause of the issue. I take great care of my PC - there is no junk, all is set properly, accurately, and everything always worked until 0.87.7...

[Mr Alpha]

if madvr start spiking i reset everything to default and setup all again this normally fix the issue at least for me.The spikes still remain after resetting everything to default, it just shrank to 135ms instead of 270ms.

I also noticed that the spikes only happen in full screen exclusive mode.

[Eiffel]

Oh, it seems Asmodian was right about this. See his post here:

http://forum.doom9.org/showpost.php?p=1673587&postcount=24974

Now you're probably wondering why older madTPG builds behaved differently to the latest madTPG build? The answer to this is [...]

Thanks madshi and Asmodian! I'm starting to understand what's going on (I'd never have thought that madVR was that smart :cool:).

I now need to investigate why my latest 3dlut doesn't behave as expected (possibly a user error?).

[seiyafan]

This is on my R9 270x:

D3D9 StretchRect: 2832 fps
D3D9 HLSL PixelShader: 2133 fps
OpenCL copy: 3050 fps
OpenCL kernel: 3612 fps
OpenCL copy interop: 400 fps
OpenCL kernel interop: 462 fps


The difference is even greater!!!!!!!

I just ran it again, and D3D9 StretchRect drops to 2394 fps...hmmm..........not consistent.

One minute later, it dropped to 2211 fps....

2189 fps the third test and still dropping...

fifth test, went back up to 2856 fps, magic ain't it?

[Allamar]

This may have been asked before - but I am using a AMD RR9 270X connected to a 4K monitor with MST. Whenever I run madVR with MPC-HC (lav decoder) and enter full screen exclusive I get a black screen. The audio continue playing - and the movie picked up fine if I switch back to windowed mode. Anyone see this happen? I'm using windows 8.1 and the latest non beta catalyst drivers.

[jkauff]

madshi,

Please consider adding something like this to Settings:

https://dl.dropboxusercontent.com/u/29526622/Profiles.png

[Qaq]

@madshi
I'd like to measure my 7750 OCL performance and share the results but couldn't get your tool runing. Some VC package seems missing or just wasn't found (I use to disable lots of system services for media playback). What application exactly needs to be installed to run your tool? And if its really missing can it affect OCL performance somehow?

[StinDaWg]

I'm trying to sort out my issues with nnedi image doubling with my 7770. With 720p24 content upscaled to a 1920x1080 screen, I can't get any of the image doubling settings to play smoothly. The best I can get is ~42-45 ms rendering with 16 neurons, but this is not good enough and it just drops frames. I have tried every trade for performance option and setting combination imaginable.

Meanwhile, I've been chatting with someone on another forum who also has a 7770, and he claims he can do 720p24 on a 1080p screen with nnedi doubling 32 neurons/16 chroma double, error diffusion 1 with no issues. These same settings give me ~92ms rendering time on my card.

Can anyone with a 7770 test this out and tell me what you get? I find it hard to believe two 7770 cards can be this dramatically different. I have a lot of 720p24 content and I just want to be able to do nnedi image doubling/32 neurons without issue. I only have a few days left to return my card so I need to figure out if this an issue on my end or if I need to return the card and get something more powerful so I can do nnedi. I've been so impressed with the image quality, this has become a must have feature for me.

[MistahBonzai]

Meanwhile, I've been chatting with someone on another forum who also has a 7770, and he claims he can do 720p24 on a 1080p screen with nnedi doubling 32 neurons/16 chroma double, error diffusion 1 with no issues. These same settings give me ~92ms rendering time on my card.

Can anyone with a 7770 test this out and tell me what you get?

HD7850 Catalyst 14.2 beta. All the very latest MPC-HC/MadVR/LAV filters... i7-3770, 3.4GHz, Win7 64-bit - works fine.

No can do :( See screen crop (http://www.mediafire.com/view/k4vdj1c244cida8/720p-24fps.png). Drops frames steadily. ~51.25ms (no trade quality for performance options checked - no smooth motion - low stress scaling excepting image doubling.)

[bugmen0t]

What does dropping frames because of 'no free vsync slot' mean? Is this a performance problem, something to be expected to happen sometimes or a bug in madVR (FRC & ReClock in use)?

00032189 Render Queue.decoded: 16, ivtced: 0, subtitled: 0, uploaded: 8, dxvaProcessed: 0, split: 0, rendered: 8, backbuffered: 0, presented: 7, PresentQueueCount: 9, PresentQueuePresentCount: 8
[...]
00032189 Render NextVSync: 1927, PresentQueueCount: 8, PresentQueuePresentCount: 7, PresentQueue.id, vsync, delay, presentId [0] 1526, 1927, 1, 0 [1] 1545, 1928, 1, 0 [2] 1528, 1929, 1, 0 [3] 1547, 1930, 1, 0 [4] 1530, 1931, 1, 0 [5] 1549, 1932, 1, 0 [6] 1532, 1933, 1, 0 [7] 1532, 1934, 1, 0 [8] 0, 0, 0, 0 [9] 0, 0, 0, 0 [10] 0, 0, 0, 0 [11] 0, 0, 0, 0 [12] 0, 0, 0, 0 [13] 0, 0, 0, 0 [14] 0, 0, 0, 0 [15] 0, 0, 0, 0 [16] 0, 0, 0, 0 [17] 0, 0, 0, 0 [18] 0, 0, 0, 0 [19] 0, 0, 0, 0 [20] 0, 0, 0, 0 [21] 0, 0, 0, 0 [22] 0, 0, 0, 0 [23] 0, 0, 0, 0 [24] 0, 0, 0, 0 [25] 0, 0, 0, 0 [26] 0, 0, 0, 0 [27] 0, 0, 0, 0 [28] 0, 0, 0, 0 [29] 0, 0, 0, 0 [30] 0, 0, 0, 0 [31] 0, 0, 0, 0
[...]
00032189 Render CVSync::TimeLeft(self: 0D2B8FE8, vsync: 1946) -> + (318694 microseconds)
00032189 Render CVSync::TimeLeft(self: 0D2B8FE8, vsync: 1928) -> + (18706 microseconds)
00032189 Render drop, nextVSync 1927, timeLeft 18706, frame vsyncs 1526: (1925.50-1925.50 1926 276940000), 1545: (1928.02-1928.02 1928 277360000), 1528: (1928.02-1928.02 1929 277360000), 1547: (1930.53-1930.53 1930 277780000), 1530: (1930.53-1930.53 1931 277780000), 1549: (1932.99-1932.99 1932 278190000), 1532: (1932.99-1932.99 1933 278190000), 1551: (1935.51-1935.51 1935 278610000), 1534: (1935.51-1935.51 1936 278610000), 1553: (1938.02-1938.02 1938 279030000), 1536: (1938.02-1938.02 1939 279030000), 1555: (1940.48-1940.48 1940 279440000), 1538: (1940.48-1940.48 1941 279440000), 1557: (1943.00-1943.00 1943 279860000), 1540: (1943.00-1943.00 1943 279860000), 1542: (1945.52-1945.52 1946 280280000), 1544: (1948.04-1948.04 1948 280700000), 1546: (1950.50-1950.50 1950 281110000), 1548: (1953.02-1953.02 1953 281530000), 1550: (1955.54-1955.54 1956 281950000), 1552: (1958.00-1958.00 1958 282360000), 1554: (1960.46-1960.46 1960 282770000), 1556: (1962.98-1962.98 1963 283190000)
00032189 Render LastPresentVSync: 1933, freeNum: 1934 [1551] 1935, 1935, 1936 [1534] 1936, 1936, 1937 [1553] 1938, 1938, 1939 [1536] 1939, 1939, 1940 [1555] 1940, 1940, 1941 [1538] 1941, 1941, 1942 [1557] 1943, 1943, 1944
00032189 Render drop frame 1557, no free vsync slot, plannedVSync: 1943, next frame plannedVSync: 1943

[StinDaWg]

HD7850 Catalyst 14.2 beta. All the very latest MPC-HC/MadVR/LAV filters... i7-3770, 3.4GHz, Win7 64-bit - works fine.

No can do :( See screen crop (http://www.mediafire.com/view/k4vdj1c244cida8/720p-24fps.png). Drops frames steadily. ~51.25ms (no trade quality for performance options checked - no smooth motion - low stress scaling excepting image doubling.)
So I guess he's full of it, since your much more powerful 7850 can't even do it?

[aufkrawall]

The only proper solution to nvidia's refresh rate levels issue over hdmi is to use the madLevelsTweaker, or similar tools.

If one's using DP, only a forced custom resolution via CRU actually works, afaik.

[wolfman2791]

madshi,

Please consider adding something like this to Settings:

https://dl.dropboxusercontent.com/u/29526622/Profiles.png Just put your preferred settings as the first one on each profile... the first one gets loaded automatically.

[madshi]

I have no idea where the problem is, but I do know that madVR 0.87.6 madTPG window works perfectly fine with dispcalGUI and raw ArgyllCMS 1.6.3 commands, but madTPG 0.87.7 shows gray screen after I press Run when Windows Security for madHcCtrl.exe shows up upon launching the file for the first time in whichever folder. I'm sorry for the long run-on sentence...
Is it possible you have an old madHcNet32.dll flying around somewhere? Maybe in the ArgyllCMS folder? Please search your harddisk for madHcNet32.dll, madHcNet64.dll, madVR.ax, madTPG.exe and madHcCtrl.exe. All of these files should only exist once on your PC - in the madVR folder. If they're anywhere else, delete those.

This may have been asked before - but I am using a AMD RR9 270X connected to a 4K monitor with MST. Whenever I run madVR with MPC-HC (lav decoder) and enter full screen exclusive I get a black screen. The audio continue playing - and the movie picked up fine if I switch back to windowed mode. Anyone see this happen? I'm using windows 8.1 and the latest non beta catalyst drivers.
Is it possible that Direct3D switches to a display mode your TV can't handle? Are you using the madVR display mode switcher? I guess you could create and upload a (zipped) a debug log. Maybe I can then tell you whether madVR is rendering at all, or not.

Please consider adding something like this to Settings.
If you don't have profile rules setup, the first listed profile is always activated by default. You can change the order of profiles by drag&drop in the settings dialog. So simply move the profile you want to be used by default to the top of the list.

I'd like to measure my 7750 OCL performance and share the results but couldn't get your tool runing. Some VC package seems missing or just wasn't found (I use to disable lots of system services for media playback). What application exactly needs to be installed to run your tool? And if its really missing can it affect OCL performance somehow?
I don't know. The tool was really only meant for the AMD developers to use for testing. It wasn't meant as a benchmark madVR users should run. So I haven't taken any care whether it works fine on all GPUs (it crashes on NVidia GPUs), or whether it has some weird requirements or not.

And no, OCL performance is not affected, in any case.

What does dropping frames because of 'no free vsync slot' mean?
There can be several reasons, but the most usual one is that your display mode's refresh rate is slightly too low, so that madVR has to drop a frame once in a while to keep audio/video in sync.

[flashmozzg]

So I guess he's full of it, since your much more powerful 7850 can't even do it?

Well, he could have overclocked version with high amount of ram. That would help. Just ask him for a screen.

[leeperry]

I'd like to measure my 7750 OCL performance and share the results but couldn't get your tool runing. Some VC package seems missing
Same here, I get a SideBySide error msg....I apparently lack some VC++ runtime libraries.

[madshi]

I will test next versions of API Monitor to see if it will watch that API one day... And I hope that someone else will have the same problem, maybe it will help to find where this problem come from. Fortunately I don't have any problem with Overlay mode, no tearing or frame drops.
Can you please try this test build?

http://madshi.net/madVRrefreshRate.rar

Please upload the debug log you get, no matter if it works or not, I'd like to see it in any case. Thanks.

So I guess he's full of it, since your much more powerful 7850 can't even do it?
Not necessarily. As we've seen interop cost seems to vary greatly from one user to the next, seemingly even with the same GPU.

[StinDaWg]

Not necessarily. As we've seen interop cost seems to vary greatly from one user to the next, seemingly even with the same GPU.
Is there anything I can do to troubleshoot or try to figure out why my card is so slow compared to others? A 40 ms rendering difference from two 7770s with similar files and the same settings is a bit extreme don't you think?

[madshi]

Nobody knows why interop is so costly, and why it differs from one PC to the next. So no, I don't think there's anything you can do. We'll have to wait for AMD to fix the problem (hopefully).

[peplegal]

My miserably results from my HD6450 :(


http://i.imgur.com/T7BIYNb.jpg

[James Freeman]

Excuse me, whats the purpose of this benchmark? I think I've missed it.

[madshi]

The one and only purpose was for AMD developers to be able to reproduce the interop cost. It wasn't really meant as a benchmark for madVR users. Although I understand that it might be interesting for madVR users to try it.

[James Freeman]

madshi can you give us a hint what's next for madVR?
Are you working from a list or the ideas just float spontaneously?
...I'm getting addicted to this thread... :D

[Werewolfy]

Can you please try this test build?

http://madshi.net/madVRrefreshRate.rar

Please upload the debug log you get, no matter if it works or not, I'd like to see it in any case. Thanks.


It doesn't work but now I have dropped frames or repeated frames.

https://www.mediafire.com/?664asihei0sjs3k

[peplegal]

Excuse me, whats the purpose of this benchmark? I think I've missed it.

You kidding me ?

People LOVE benchmarks! You know, we like comparing everything, even though there is no good reason for that or the result itself doesn't say anything. :p


BTW...Madshi : I notice some CPU clock deppendency on those OpenCL tests, when I tweak Windows power management....Or, can be only a side effect of my weak VGA.

[madshi]

madshi can you give us a hint what's next for madVR?
No... :p

It doesn't work
Too bad. Log confirms old results. I'm officially out of ideas.

[James Freeman]

No... :p


No surprise.
It'd be like throwing a chicken into a lions cage.
Or a whale nibbling your bait with the line tied to your wrist.
Playing with matches in a hay barn?

:D:D

[bugmen0t]

There can be several reasons, but the most usual one is that your display mode's refresh rate is slightly too low, so that madVR has to drop a frame once in a while to keep audio/video in sync.
I doubt that. Since you fixed the 'full present queue slowed down rendering' bug I went back to default queue sizes and such frame drops can sometimes happen now. However making sure the present queue isn't completely full will make everything totally stable and there's never a frame drop then.

[huhn]

Same here, I get a SideBySide error msg....I apparently lack some VC++ runtime libraries.

i reinstalled all vc++ runtimes and it didn't worked, after that i tried "All in One Runtimes" after that i was able to use it.
my numbers are very very low too, compared to user who got a gpu like my r9 270.

[kasper93]

Well guys this benchmark is not meant to be used by us. But I do agree it always fun to compare results :D As for people who can't run it, it seems that madshi compiled this with VS2005 so you need to have VC++2005 redist which is old stuff :) In the end it's not important, just hope AMD will fix this and they do in this year not like in 2018... I don't have big hopes unfortunately. NV guys seems to be more flexible if it comes to such fixes. AMD guys are like "+15% in GTA 5" or "Fix random crashes in DX9 applications" and this all they put in release notes ;p

[jkauff]

Just put your preferred settings as the first one on each profile... the first one gets loaded automatically.
Good to know, that wasn't clear. I'm using a script, and I thought I needed to load the profile group with the hotkey to run the script.

[MS-DOS]

Seriously!?!? This is not funny, anymore. 402fps = 2.5ms rendering time. 13.22fps = 75.6ms rendering time. So on your PC interop cost per 1080p frame is 73.1ms. That's insane.

HD5870 13.12 here:

http://i2.imageban.ru/out/2014/03/16/7b339ce9cbefdc3fd404d6233f14733a.png

Sad but true. That's why I dream about NNEDI3 on DirectCompute.

[Niyawa]

I have a question about ordered dithering. madVR itself tells me that it's "a faster alternative to error diffusion" but oddly enough error diffusion is faster here (ordered dithering 8 ms, error diffusion 5 ms). I'd like some enlightenment if possible.

For the record, I have an old Intel 4500MHD.

[nevcairiel]

I have a question about ordered dithering. madVR itself tells me that it's "a faster alternative to error diffusion" but oddly enough error diffusion is faster here (ordered dithering 8 ms, error diffusion 5 ms). I'd like some enlightenment if possible.

For the record, I have an old Intel 4500MHD.

Error Diffusion requires a DX11 compatible GPU, and it notes as much in the options. Your GPU is not, so it'll fallback to some other mode - possibly random dithering, which may be faster.