BM3D (VapourSynth) versus MVTools

DJATOM · 4th July 2021, 18:10

CUDA based BM3D (V-BM3D) is fast enough on decent GPU and probably faster than mvtools2 with small temporal radius. You might want to consider using it as faster denoising backend.
Some numbers if you interested:
1080p, radius=3, fast=True - ~21.85 fps
1080p, radius=3, fast=False - ~14.17 fps
1080p, radius=2, fast=True - ~38.93 fps
1080p, radius=2, fast=False - ~19.69 fps
1080p, radius=1, fast=True - ~63.21 fps
1080p, radius=1, fast=False - ~31.49 fps
1080p, radius=0, fast=True - ~159.75 fps (that's a spatial mode)
1080p, radius=0, fast=False - ~79.24 fps (that's also a spatial mode)
"fast" option squeezes more fps with the same quality at expense of flooding PCIe lanes, probably my card can preform better, but due to SSD attached to CPU lanes it works at x8 mode (still enough for me).

ChaosKing · 4th July 2021, 19:21

ahh finally some numbers for bm3dcuda. Could you also post some numbers for the bm3dCPU version? I wonder how big the speed gap is.

DJATOM · 4th July 2021, 19:41

radius=0 - ~114.46 fps
radius=1 - ~65.87 fps
radius=2 - ~46.11 fps
radius=3 - ~36.03 fps
radius=4 - ~29.26 fps
I'm measuring with 12 cores CPU and overclocked RAM (yet no turbo boost enabled).

ChaosKing · 4th July 2021, 20:54

Didn't expect that. It's even faster than your 2070. I guess a ~8 cores cpu could be about the same speed as the 2070 then.

tormento · 5th July 2021, 07:16

Quote:

Originally Posted by DJATOM

CUDA based BM3D (V-BM3D) is fast enough on decent GPU and probably faster than mvtools2 with small temporal radius. You might want to consider using it as faster denoising backend.

Thanks for your suggestion, I did not know that BM3D had a CUDA version. Where can I find a Windows build, better with HBD support?

However, there are a few CUDA denoiser around already. What we need is a MVTool on CUDA or OpenCL.

ChaosKing · 5th July 2021, 07:48

Quote:

Originally Posted by tormento

Thanks for your suggestion, I did not know that BM3D had a CUDA version. Where can I find a Windows build, better with HBD support?

However, there are a few CUDA denoiser around already. What we need is a MVTool on CUDA or OpenCL.

https://github.com/WolframRhodium/VapourSynth-BM3DCUDA

Scroll to version 2.1 for cuda version.

tormento · 5th July 2021, 07:50

Quote:

Originally Posted by ChaosKing

Scroll to version 2.1 for cuda version.

Thanks but it is for VS not AVS+.

kedautinh12 · 5th July 2021, 08:34

In the present, BM3D only in VS

WolframRhodium · 5th July 2021, 09:22

The numbers from DJATOM looks interesting.

The GPU performance aligns well with the estimation that

Code:

fps * (2 * radius + 1) ≈ constant

, since the calculation is dominated by block matching.

The CPU performance is unexpected to me. Maybe the cache plays a nontrivial role.

Anyway, I have no experience in writing an AVS+ plugin. Will take a look if people are interested.

kedautinh12 · 5th July 2021, 09:31

Yeah, i'm waiting many years for avs+ ver

real.finder · 5th July 2021, 10:51

Quote:

Originally Posted by WolframRhodium

Anyway, I have no experience in writing an AVS+ plugin. Will take a look if people are interested.

you can make the plugin work in both https://github.com/HomeOfAviSynthPlu...on/neo_DFTTest (this is backport of vs DFTTest and now it's work on both)

WolframRhodium · 5th July 2021, 12:22

Quote:

Originally Posted by real.finder

you can make the plugin work in both https://github.com/HomeOfAviSynthPlu...on/neo_DFTTest (this is backport of vs DFTTest and now it's work on both)

Thanks, I am aware of the DualSynth implementation. I need to read more about the parallelism mechanism of AVS+, thought.

Reel.Deel · 8th July 2021, 00:55

Quote:

Originally Posted by WolframRhodium

Anyway, I have no experience in writing an AVS+ plugin. Will take a look if people are interested.

+1 ... definitely some interest here.

Dogway · 8th July 2021, 09:27

Obviously super interested in this. I think it can easily double current scripts speed that make use of mvtools. Probably for implementation reference you can check source of AVS+ 3.7.1 CUDA.

real.finder · 22nd July 2021, 05:42

WolframRhodium, any news about avs+ port?

@Moderators, can you move BM3D posts to new thread?

tebasuna51 · 22nd July 2021, 07:18

Quote:

Originally Posted by real.finder

@Moderators, can you move BM3D posts to new thread?

Moved. It is ok now?

real.finder · 22nd July 2021, 10:45

Quote:

Originally Posted by tebasuna51

Moved. It is ok now?

Yes. It's fine. Thanks

WolframRhodium · 22nd July 2021, 13:43

I'm sorry about being lazy on avs+ port... Will consider it this weekend.

Dogway · 22nd July 2021, 17:14

I think OP was replying to my post:

Quote:

Originally Posted by Dogway

Currently mvtools is the bottleneck of most scripts, I calculated that for a 2h feature we need a performance of at least 5fps to finish an encode in 10h, good value for an overnight encode. I think this is still a bit far when using refinemotion (a must in my book) -1080p @ 16-bit-.

Thank you WolframRhodium, it would actually be an achievement and a revolution for AVS since mvtools is the major bottleneck in most scripts.

lansing · 22nd July 2021, 20:42

I'm confused, so the 3900x is faster than the RTX 2070??

4th July 2021, 18:10	#1 \| Link
DJATOM Registered User Join Date: Sep 2010 Location: Ukraine, Bohuslav Posts: 356	BM3D (VapourSynth) versus MVTools CUDA based BM3D (V-BM3D) is fast enough on decent GPU and probably faster than mvtools2 with small temporal radius. You might want to consider using it as faster denoising backend. Some numbers if you interested: 1080p, radius=3, fast=True - ~21.85 fps 1080p, radius=3, fast=False - ~14.17 fps 1080p, radius=2, fast=True - ~38.93 fps 1080p, radius=2, fast=False - ~19.69 fps 1080p, radius=1, fast=True - ~63.21 fps 1080p, radius=1, fast=False - ~31.49 fps 1080p, radius=0, fast=True - ~159.75 fps (that's a spatial mode) 1080p, radius=0, fast=False - ~79.24 fps (that's also a spatial mode) "fast" option squeezes more fps with the same quality at expense of flooding PCIe lanes, probably my card can preform better, but due to SSD attached to CPU lanes it works at x8 mode (still enough for me). __________________ Me on GitHub PC Specs: Ryzen 5950X, 64 GB RAM, RTX 2070

4th July 2021, 19:21	#2 \| Link
ChaosKing Registered User Join Date: Dec 2005 Location: Germany Posts: 1,514	ahh finally some numbers for bm3dcuda. Could you also post some numbers for the bm3dCPU version? I wonder how big the speed gap is. __________________ AVSRepoGUI // VSRepoGUI - Package Manager for AviSynth // VapourSynth VapourSynth Portable FATPACK \|\| VapourSynth Database \|\| https://github.com/avisynth-repository Last edited by ChaosKing; 4th July 2021 at 19:23.

4th July 2021, 19:41	#3 \| Link
DJATOM Registered User Join Date: Sep 2010 Location: Ukraine, Bohuslav Posts: 356	radius=0 - ~114.46 fps radius=1 - ~65.87 fps radius=2 - ~46.11 fps radius=3 - ~36.03 fps radius=4 - ~29.26 fps I'm measuring with 12 cores CPU and overclocked RAM (yet no turbo boost enabled). __________________ Me on GitHub PC Specs: Ryzen 5950X, 64 GB RAM, RTX 2070

4th July 2021, 20:54	#4 \| Link
ChaosKing Registered User Join Date: Dec 2005 Location: Germany Posts: 1,514	Didn't expect that. It's even faster than your 2070. I guess a ~8 cores cpu could be about the same speed as the 2070 then. __________________ AVSRepoGUI // VSRepoGUI - Package Manager for AviSynth // VapourSynth VapourSynth Portable FATPACK \|\| VapourSynth Database \|\| https://github.com/avisynth-repository

5th July 2021, 09:22	#9 \| Link
WolframRhodium Registered User Join Date: Jan 2016 Posts: 143	The numbers from DJATOM looks interesting. The GPU performance aligns well with the estimation that Code: fps * (2 * radius + 1) ≈ constant , since the calculation is dominated by block matching. The CPU performance is unexpected to me. Maybe the cache plays a nontrivial role. Anyway, I have no experience in writing an AVS+ plugin. Will take a look if people are interested.

5th July 2021, 08:34	#8 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 864	In the present, BM3D only in VS

5th July 2021, 09:31	#10 \| Link
kedautinh12 Registered User Join Date: Jan 2018 Posts: 864	Yeah, i'm waiting many years for avs+ ver

8th July 2021, 09:27	#14 \| Link
Dogway Registered User Join Date: Nov 2009 Posts: 1,552	Obviously super interested in this. I think it can easily double current scripts speed that make use of mvtools. Probably for implementation reference you can check source of AVS+ 3.7.1 CUDA. __________________ [i7-4790K@Stock::GTX 1070] AviSynth+ filters and mods on GitHub

22nd July 2021, 05:42	#15 \| Link
real.finder Registered User Join Date: Jan 2012 Location: Mesopotamia Posts: 2,359	WolframRhodium, any news about avs+ port? @Moderators, can you move BM3D posts to new thread? __________________ See My Avisynth Stuff Last edited by real.finder; 22nd July 2021 at 05:44.

22nd July 2021, 13:43	#18 \| Link
WolframRhodium Registered User Join Date: Jan 2016 Posts: 143	I'm sorry about being lazy on avs+ port... Will consider it this weekend.

22nd July 2021, 20:42	#20 \| Link
lansing Registered User Join Date: Sep 2006 Posts: 1,512	I'm confused, so the 3900x is faster than the RTX 2070??

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode