madVR - high quality video renderer (GPU assisted)

[Asmodian]

How would a benchmark differ from simply trying it? In your examples you don't need a benchmark tool (what would it do?), set the setting and see what the rendering time is. The complexity of the possible settings is the only issue and if you pick what "128 neuron doubling" means for all the other settings, source resolution, destination resolution, etc. it is easy to "benchmark" any GPU you own at "128 neuron doubling".

Like any GPU benchmark the results are highly dependent on the settings used and different users/sites use different settings so it is hard to compare results between users. Maybe a tool that played a stock video using a collection preset options with a particular player and at a specific resolution? It could report the average/min/max rendering times at each setting. This would be easy to do now, no special tool needed, but maybe review sites would include madVR performance if such a tool existed. However, it sounds like a lot of work, certainly more work than the benefit justifies before version 1.0 of madVR.

madVR is changing fast so it isn't time for standard benchmark tools yet.

[QBhd]

^^

Well put. Let's put this slightly off topic to bed.

QB

[Arm3nian]

Originally Posted by Asmodian

How would a benchmark differ from simply trying it? In your examples you don't need a benchmark tool (what would it do?), set the setting and see what the rendering time is. The complexity of the possible settings is the only issue and if you pick what "128 neuron doubling" means for all the other settings, source resolution, destination resolution, etc. it is easy to "benchmark" any GPU you own at "128 neuron doubling".

I can't test a gpu's performance in madVR if I don't have the gpu. It will help when buying a new card because I can see what scores others have gotten. Users can also post stock and overclocked results. I don't want it to be an e-peen tool, just something to view relative performance. Nothing like that currently exists. The best you can do is calculate how much faster a certain gpu is from another by using non madVR benchmarks, then apply that percentage to your current gpu's render clocks and get an estimated projection. Not very accurate.

Originally Posted by Asmodian

Maybe a tool that played a stock video using a collection preset options with a particular player and at a specific resolution? It could report the average/min/max rendering times at each setting. This would be easy to do now, no special tool needed, but maybe review sites would include madVR performance if such a tool existed.

Yes this is precisely what I envision. If madVR is included in benchmarks carried out by popular sites it will gain more publicity and attract more users, including enthusiasts.

Originally Posted by Asmodian

However, it sounds like a lot of work, certainly more work than the benefit justifies before version 1.0 of madVR.

madVR is changing fast so it isn't time for standard benchmark tools yet.

I agree, there are lots of new things happening currently. When it is a bit more finalized then a benching tool can be of use.

[e-t172]

Originally Posted by Asmodian

How would a benchmark differ from simply trying it?

Reading a number off a publicly available table (or better, a chart) is much easier than having to manually try dozens of combinations. The former takes seconds, the latter takes hours.

[huhn]

Originally Posted by e-t172

Reading a number off a publicly available table (or better, a chart) is much easier than having to manually try dozens of combinations. The former takes seconds, the latter takes hours.

and what is the point if you never really judge the picture quality of the scaler yourself?

[Arm3nian]

Originally Posted by huhn

and what is the point if you never really judge the picture quality of the scaler yourself?

Nearly all of the settings, including needi3 and the newely implemented ones have a positive correlation between quality and GPU usage.

[e-t172]

Originally Posted by huhn

and what is the point if you never really judge the picture quality of the scaler yourself?

• Being able to know which ones you can cross out from the start because they won't be fast enough.
• When you know what the scaler looks like, but you don't know how fast it will be because you're using a new GPU.
• Sometimes we trust other people to be the experts on figuring out what scalers are the best or even worth considering. Not everybody has dozens of hours to spend trying lots of combinations of settings across a wide variety of samples. They just want something that is most likely to look best on typical content considering their hardware.

There are tons of reasons why having easily accessible performance data for various option combinations is a very good idea. I suspect it would also help madshi know what is worth optimizing.

[Asmodian]

Originally Posted by e-t172

Reading a number off a publicly available table (or better, a chart) is much easier than having to manually try dozens of combinations. The former takes seconds, the latter takes hours.

I meant in the context of generating this theoretical chart. A benchmark tool does not create said chart, someone still has to run it on each card.

The desire behind this request is for someone to test madVR with a wide collection of video cards in a systematic way and post the results. A special tool isn't required for that and it wouldn't even make the process much easier.

Originally Posted by huhn

and what is the point if you never really judge the picture quality of the scaler yourself?

This is to use madVR like 3DMark, simply as a way to compare the performance of two video cards, not to pick settings.

[Arm3nian]

Originally Posted by Asmodian

The desire behind this request is for someone to test madVR with a wide collection of video cards in a systematic way and post the results. A special tool isn't required for that and it wouldn't even make the process much easier.

It's not practical to do it manually. It would take time to create all the profiles, run them one by one, record them, then write it in the spreadsheet. Plus you need the same source material. Countless of other factors can result in errors and is overall too much of a hassle. No one would participate. You need to automate the process with a script. Press one button and it runs the bench and uploads the results to the spreadsheet.

[James Freeman]

Originally Posted by Arm3nian

You guys are missing the point. A benchmark would not be used to find optimal settings. It would be used to show what a certain gpu is capable of.

Originally Posted by Asmodian

How would a benchmark differ from simply trying it?
madVR is changing fast so it isn't time for standard benchmark tools yet.

Arm3nian,

You can always create a "madVR GPU Benchmark Thread" in which you'll define certain parameters with a certain madVR version as a base.
People could post their results with the GPU's they own there, and not contaminate madVR official thread with such nonsense.

As I already said.
Just keep the rendering time under frame time.
1/"frame time" = maximum rendering time before dropped frames.

Cheers.

[JarrettH]

Benchmark? It's either realtime or the video is choppy Maybe the benchmark should be for how low your rendering time can go.

[Arm3nian]

Originally Posted by James Freeman

Arm3nian,

You can always create a "madVR GPU Benchmark Thread" in which you'll define certain parameters with a certain madVR version as a base.
People could post their results with the GPU's they own there, and not contaminate madVR official thread with such nonsense.

Read my previous on why this wouldn't work.

Originally Posted by James Freeman

As I already said.
Just keep the rendering time under frame time.
1/"frame time" = maximum rendering time before dropped frames.

Three pages and you still have no idea what we're talking about. If anyone should stop the nonsense, it's you.

Originally Posted by JarrettH

Benchmark? It's either realtime or the video is choppy Maybe the benchmark should be for how low your rendering time can go.

Benchmark results show the most demanding setting you can run real time.

[ryrynz]

Might be an idea showing GPU usage on the OSD when GPU-Z is active.

[Nevilne]

Potplayer has GPU usage built-in fyi

[wanezhiling]

Originally Posted by Nevilne

Potplayer has GPU usage built-in fyi

Yes.
Not support Intel GPU though

[madshi]

Originally Posted by 6233638

Now that we have a 10-bit output, I tried using CalMAN with madTPG to create a LUT for my display, and I'm running into a few issues:

1. When double-clicking to exit FSE mode, the screen just turns black rather than returning to windowed mode.
2. madTPG doesn't work very well if you have a single display setup.

It's very awkward to switch between CalMAN to set up the reads and then quickly switch over to madTPG and make it fullscreen on the same monitor.
Ideally what would happen is that madTPG would run hidden in the background and then enter an always-on-top state or FSE mode as soon as it is being told to display patterns.
And once that series of patterns is completed, return to being hidden in the background.

Ideally there would also be a way to pause for 5 seconds at the initial switch to FSE mode, but I assume that's something which would have to be fixed on CalMAN's side of things.

Hi there,

the latest Calman release now has some improvements when running Calman on the same PC as madTPG. The changes have been inspired by your "complaint"... So it would be awesome if you could try the new build and maybe provide feedback here:

http://www.spectracal.com/forum/view...hp?f=94&t=5729

Stacey Spears has actively asked for feedback about this. Would be great if you could tell him/us if it works better now, or if there's still something that could be improved.

[XRyche]

Originally Posted by madshi

Hi there,

Subtle and Polite LOL

[madshi]

Had actually meant to send it as a PM. Ooops...

[ikarad]

Originally Posted by viewer

Source: H.264 8bit 4:2:0 http://forum.doom9.org/showthread.ph...13#post1727413

a. SuperRes off
b. SuperRes on (passes: 2 strength: 0.75 softness: 0.10)
c. SuperRes on (passes: 2 strength: 1.00 softness: 0.00)

\ a b c
1) chroma upscaling: Catmull-RomAR
2) chroma upscaling: Bicubic75AR
3) chroma upscaling: SoftCubic50AR
4) chroma upscaling: Lanczos3AR
5) chroma upscaling: JincAR


\ a b c
1) chroma upscaling: Bilateral
2) chroma upscaling: super-xbr100AR
3) chroma upscaling: super-xbr150AR
4) chroma upscaling: Nedi
5) chroma upscaling: NNEDI32


\ a b c
1) chroma upscaling: Catmull-RomAR
2) chroma upscaling: JincAR
3) chroma upscaling: Bilateral
4) chroma upscaling: super-xbr150AR
5) chroma upscaling: NNEDI32


It seems like no need Chroma SuperRes.

What are the diffrences because i don't see anything?

[ryrynz]

Originally Posted by ikarad

What are the diffrences because i don't see anything?

Haha. You need to look closer.. or chop the picture up and compare in an image viewer, this is why I asked for individual images, placing them together like that doesn't make comparing them very easy.

BTW feel free to edit quotes BTW, you didn't need to quote his entire post.