Search options (Page 4 of 21)

removed all 3, VRR works, but same problem as vulkan now so black levels are crushed & banding.

removed drm-vrr-enabled=yes and d3d11-sync-interval=1 as they're useless it seems.

Testing only with:

hwdec=d3d11va *Prevents VRR but black levels are good, and no banding.

Changed to d3d11va-copy *VRR works but same problem as above crushed blacks & banding

nvdec, cuda, etc all same thing, do not work properly.

hwdec=d3d11va provides the best black levels and no banding, but breaks VRR.

Question, regarding RIFE I assume the 4070, 4070 Super, 4070 Ti Super, etc are all better than the RTX 3090 with regards to RIFE yes?

I'm asking because I can sell my RTX 3090 used for around $1100 in Australia, and just buy an RTX 4070 Super or 4070 Ti Super for example on special for around the same price.

3090 can no longer handle 4K at 2x with the latest RIFE models sometimes, if the 4070 and higher can handle it, I am willing to do the switch, despite having half the vRAM.

Are you both updating your vsmlrt.py or not?

Go here: https://github.com/AmusementClub/vs-mlrt/releases

Currently, the latest version is v14.test3, scroll down and open Assets under the latest version. Download scripts.v14.test3.7z, which is 11.7 KB (that's very small and should take 1 second to download, don't download the big files).

Open it and keep it on the side, go to C:\Program Files (x86)\SVP 4\rife

Now drag and drop the vsmlrt.py from inside scripts.v14.test3.7z, into the folder above and replace the original file.

It should now be updated. Perhaps you're both using an outdated version of vsmlrt.py or an old nVidia driver. Your nVidia driver should also be up-to-date.

Generally you want to update vsmlrt.py every time they update it when you are always testing the latest versions.

If you want Dolby Vision, HDR10, 10Plus, and SDR to all work properly (with HDR enabled under Display setting in Windows Settings), I have these configurations saved. Remember, Dolby Vision + RIFE do not work together still.

The order of the shaders if you want to use them is also important, last year with the original post at the top, the order was completely wrong.

But as @dawkinscm said you don't need to use shaders, also can remove the configs they suggested removing. I will do more testing in the coming weeks probably and try to remember to update in the future as well.

Not sure which TV you are using, or which monitor, but for target-peak=800 in the code below, I have 800 because that's the maximum NITS of brightness for the LG C2 42", C3 would be the same, G3 would be different because it uses MLA technology and can go higher. QD-OLED can go higher as well, depending on the product.

ontop
fullscreen=yes
d3d11-exclusive-fs=yes

volume=100
volume-max=100
video-sync=audio
hr-seek-framedrop=no

fbo-format=rgba16hf

vo=gpu-next
gpu-api=d3d11
hwdec=d3d11va
gpu-context=d3d11
drm-vrr-enabled=no
d3d11-sync-interval=1

target-peak=800
target-trc=pq
target-prim=bt.2020

tone-mapping=spline
tone-mapping-mode=luma
gamut-mapping-mode=perceptual
tone-mapping-param=bt.2390
tone-mapping-max-boost=2.0

dither=error-diffusion
error-diffusion=burkes
dither-depth=10
spirv-compiler=auto

target-contrast=inf
target-colorspace-hint=yes
hdr-compute-peak=yes

scale=ewa_lanczos
cscale=ewa_lanczos
dscale=ewa_lanczos
tscale=ewa_lanczos

osd-level=1
osd-bar-w=25
osd-color=0.5
osd-bar-align-x=0
osd-bar-align-y=-1

glsl-shader="C:\Users\Username\AppData\Roaming\mpv\Shaders\FSRCNNX_x2_8-0-4-1.glsl"

glsl-shader="C:\Users\Username\AppData\Roaming\mpv\Shaders\KrigBilateral.glsl"

glsl-shader="C:\Users\Username\AppData\Roaming\mpv\Shaders\SSimDownscaler.glsl"

Biggest improvement in artifact masking I have seen in a long time, a must upgrade. I am testing 4.14 (v2)

No Time to Die (007 movie), 3 minutes onwards, when the masked man walked behind the curved glass door and opens it, then the curtain in the background behind the woman on the sofa.

From 14 to 17 minutes the fast action sequence now has no more artifacts in it.

The building time does take longer, around twice as long as before, so be patient at the start.

Sadly I can't test my usual full aspect ratio IMAX movies or TV shows that take the whole screen in 4K, the RTX 3090 is not strong enough.

So I am hoping 4.14 Lite will have the same level of artifact masking, because it is brilliant.

Yep, thanks. As suspected. It won't be connected to SVP or RIFE performance lift.

I have this setup for 4K Profiles:

And this setup for 1080p Profiles:

It never switches to Automatic for me with these settings.

Thank you!

This should make things a lot easier

Interesting on Github under 4.9 v2 notes it says this:

The full quote:

EDIT:

Tried to test 4.9 v2 from 2 days ago, changed the name to 4.4 as usual, but it doesn't build. In the beginning of the building process, it breaks and stops.

EDIT 2 - adding end of crash info

[11/04/2023-08:48:47] [W] Could not read timing cache from: C:\Users\Username\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.4.onnx.3840x2144_fp16_trt-8502_cudnn_I-fp16_O-fp16_NVIDIA-GeForce-RTX-3090_cf15df92.engine.cache. A new timing cache will be generated and written.
[11/04/2023-08:48:47] [E] Error[4]: [network.cpp::nvinfer1::Network::validate::3100] Error Code 4: Internal Error (input: for dimension number 1 in profile 0 does not match network definition (got min=11, opt=11, max=11), expected min=opt=max=7).)
[11/04/2023-08:48:47] [E] Error[2]: [builder.cpp::nvinfer1::builder::Builder::buildSerializedNetwork::751] Error Code 2: Internal Error (Assertion engine != nullptr failed. )
[11/04/2023-08:48:47] [E] Engine could not be created from network
[11/04/2023-08:48:47] [E] Building engine failed
[11/04/2023-08:48:47] [E] Failed to create engine from model or file.
[11/04/2023-08:48:47] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v8501] # C:/Program Files (x86)/SVP 4/rife\vsmlrt-cuda\trtexec --onnx=C:/Program Files (x86)/SVP 4/rife\models\rife\rife_v4.4.onnx --timingCacheFile=C:\Users\Username\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.4.onnx.3840x2144_fp16_trt-8502_cudnn_I-fp16_O-fp16_NVIDIA-GeForce-RTX-3090_cf15df92.engine.cache --device=0 --saveEngine=C:\Users\Username\AppData\Roaming\SVP4\cache\Program Files (x86)/SVP 4/rife\models\rife\rife_v4.4.onnx.3840x2144_fp16_trt-8502_cudnn_I-fp16_O-fp16_NVIDIA-GeForce-RTX-3090_cf15df92.engine --shapes=input:1x11x2144x3840 --fp16 --tacticSources=-CUBLAS,-CUBLAS_LT --useCudaGraph --noDataTransfers --inputIOFormats=fp16:chw --outputIOFormats=fp16:chw

RIFE 4.9 needs to be added to SVP, maybe even as a third option @Chainik

I already added it as a replacement to 4.4 so I can do some testing over the last few hours. There are some great improvements in artifact masking, and unlike version 4.8 (which was a lot more demanding), version 4.9 for example works with my RTX 3090 at 4K without dropping any frames in most scenarios (GPU utilisation though is at 100%, but I do not drop any frames with it, except in some content), so it is still more demanding than version 4.6

Tested a few scenes I have saved, all have improved:

Farm field at the start of Equalizer 3 with the panning shot in the air over the car. The fences flickered before, now they do not flicker.

Season 8 Episode 1 of Game of Thrones, the first few minutes after the intro there are a lot of spikes, arrow helmets, trees, and other factors that introduce artifacts all combined with many panning camera shots, all have improved. Little to no artifacts in those scenes.

Some scenes in the movie Air have improved as well.

Beginning of Tenet there are improvements to fast movements, strings from guitars being broken, etc

However, extremely challenging scenes still show visible artifacts. It's not perfect, but I think 4.9 is a big enough improvement over 4.6 that it warrants being added into SVP as an update. But it shouldn't replace 4.6, because it is more demanding, so preferably as its own option as a 3rd selectable.

New nVidia driver, 545.84 WHQL

I assume this doesn't impact us, right? Would be nice if we can get a 2x performance lift.