Search options (Page 2 of 8)
No improvement in realtime performance across all models, almost the same transcoding speed (0.8% slower), but new heavy models runs 5% faster now!
RTX 4070
dawkinscm wrote:RickyAstle98 wrote:Part of my message got lost. I was asking how are you actually running LSFG? I have not seen anything except for a steam package.
I will upload quick video YT guide, I cant record generated frames from LSFG overlay!
https://youtu.be/8cIXy2IuMAs
Oh I see. It's part of the new Nvidia app.
Third-party software, not part of new NVIDIA App actually!
Part of my message got lost. I was asking how are you actually running LSFG? I have not seen anything except for a steam package.
I will upload quick video YT guide, I cant record generated frames from LSFG overlay!
https://youtu.be/8cIXy2IuMAs
dawkinscm wrote:RickyAstle98 wrote:raider10 wrote:
Do you do it with lossless scaling?
As you can see, I did RIFE x3 (1st action) on movie, then (LSFG3 x2) ontop!
I'm probably missing something obvious but how are you applying it? I only see a Steam implementation, nothing else.
The LSFG3 x3 alone: looks choppier and less accurate than RIFE x3 mode, takes 8 times less resources!
Thats better solution for 4K videos for example! For HFR monitor owners SPECIFICALLY!
Usage: when your GPU able to do 3x RIFE with 4K input, theres enough GPU headroom to RIFE x2 + LSFG3 x3 ontop without 100% usage!
For 144Hz users (1080p output / 4070 power levels): RIFE x3 + LSFG x2 = looks very smooth (24>144 / GPU usage 70% ~120W)
For 240Hz users (1080p output / 4070 power levels): RIFE x2.5 + LSFG3 x4 = looks very smooth and accurate (24>240 / GPU usage 80% ~130W)
For 360Hz users (1080p output / 4070 power levels): RIFE x3 + LSFG3 x5 = looks very smooth and accurate (24>360)
All tests done with 4.4v2 model (old) for GPU usage local measurement, newest models almost 2 times heavier, the multiplier decision, gentlemens, is up to you!
raider10 wrote:RickyAstle98 wrote:LSFG3 vs LSFG2 on any multipliers - smoother and better! Quality increased, same fast transitions now look more accurate, not just a fluid!
Im using RIFE x3 + LSFG x2 for 24>144 movies conversion, usually GPU usage looks like RIFE x4 with 6x real FLUIDITY rate!
On videos with LSFG3 only, looks pretty normal, conversion from 24FPS is less accurate, but smooth enough!
Do you do it with lossless scaling?
As you can see, I did RIFE x3 (1st action) on movie, then (LSFG3 x2) ontop!
The LSFG3 x3 alone: looks choppier and less accurate than RIFE x3 mode, takes 8 times less resources!
Thats better solution for 4K videos for example! For HFR monitor owners SPECIFICALLY!
Usage: when your GPU able to do 3x RIFE with 4K input, theres enough GPU headroom to RIFE x2 + LSFG3 x3 ontop without 100% usage!
Vequa wrote:Can someone please post updated instructions on installing and using V2 rife models?
I tried to get it done based on this thread but all the significant information is either from 2023 with dead links, doesn't work or is impossible to understand for us common folk.
I attempted to add the "4.25 heavy" v2 model with and without replacing "vsmlrt.py". It appears in the "AI model" dropdown menu but no matter what I do I can't get it to work.
Rename the 4.25 heavy to 4.30 for example, worked for me without replacing vsmlrt.py a long time ago!
LSFG3 vs LSFG2 on any multipliers - smoother and better! Quality increased, same fast transitions now look more accurate, not just a fluid!
Im using RIFE x3 + LSFG x2 for 24>144 movies conversion, usually GPU usage looks like RIFE x4 with 6x real FLUIDITY rate!
On videos with LSFG3 only, looks pretty normal, conversion from 24FPS is less accurate, but smooth enough!
RAGEdemon wrote:This is probably a stupid question however, with AI software, I have noticed that on a 4090, RT is faster than CUDA but not by much.
A 4090 has 128 RT cores and >16000 coda cores.
Is it true that with SVP, when RT cores are being used, CUDA cores are not really being used, and vice versa?
In that case, half the power of the GPU is wasted.
Would it be possible to implement both TensorRT AND CUDA simultaneously to double the performance?
Thats waste of time, recently tested RIFE low/mid workload while playing game, the game doesnt use RT cores for example, lost about 65% peak performance ingame, thats was Counter-Strike 2 game!
In both cases lost averages about 60% performance drop, which means over half the power of the GPU is wasted!
oriento wrote:i'm with 4.25v2, you're using 4.4v2 ?
The difference between 4.4 and 4.25 twise as much by performance, but without artefacts! Thats was example!
oriento wrote:thanks, for those with rtx 4090, what fps do you get when transcoding 4k video ?
The RIFE model dependant!
Older the model - better performance, newer backwards!
My RTX 4070 can handle 4K source, transcoding 24>48 sequence at 48FPS speed - 4.4v2 performance!
narkohol wrote:RickyAstle98 wrote:Something weird happens, did 1080p test now, 6 delayed frames, then tried 1440p test, got 16 delayed frames!
Man, I used 4.25 heavy, thats model twice heavier than 4.6 (by performance)!
What's your GPU model? Nvidia drivers version and Windows version?
The 11 months old (bought 11 months ago) RTX 4070 GPU model, 565.90 graphics driver, Windows 10 19045.5011 OS build!
narkohol wrote:RickyAstle98 wrote:Only 2 delayed frames for 24 seconds with VSync jitter always sit under 0.200 (59.940Hz display mode / GSync off)
Only 5 delayed frames for 24 seconds with VSync jitter always sit under 0.320 (60Hz display mode / GSync off)
Ahhh... now I see what it's happening here.
I am talking about 1440p output. 720p or 1080p are also fine here. This was with Model 4.22_lite v2, I think:
but 1440p:
So, nowhere near maxing the GPU % usage, but very noticeable stuttering/hitching/dropped/delayed frames.
This is with Model 4.6, much lower GPU % usage, not even 50%, but still a lot of stuttering/hitching/dropped/delayed frames:
I don't get why it doesn't work smoothly for 1440p, when it's not using even 50% of the GPU.
And going back to 1080p resolution in 2024 is just too big of a sacrifice. The trade-off of resolution/smoothness at 1080p doesn't pay off...
Something weird happens, did 1080p test now, 6 delayed frames, then tried 1440p test, got 16 delayed frames!
Man, I used 4.25 heavy, thats model twice heavier than 4.6 (by performance)!
narkohol wrote:RickyAstle98 wrote:I have smooth 60FPS playback!
Default SVP settings and RIFE engine, mpv player...
You mean 'default' with no scaling at all and no special settings in MPV??
MPV doesn't count Dropped frames properly. Even when you have a stutter fest playback it always says 'Dropped Frames: 0 (decoder) 0 (output)'
I had to use --hr-seek-framedrop=no --framedrop=no --video-sync=display-tempo to display 'Delayed' frames, and then:
Only 2 delayed frames for 24 seconds with VSync jitter always sit under 0.200 (59.94Hz display mode / GSync off)
Only 5 delayed frames for 24 seconds with VSync jitter always sit under 0.320 (60Hz display mode / GSync off)
narkohol wrote:RickyAstle98 wrote:I have smooth 60FPS playback!
Default SVP settings and RIFE engine, mpv player...
You mean 'default' with no scaling at all and no special settings in MPV??
MPV doesn't count Dropped frames properly. Even when you have a stutter fest playback it always says "0"
I had to use --hr-seek-framedrop=no --framedrop=no --video-sync=display-tempo to display 'Delayed frames', and then:
I dont need display-tempo/dislay-resample at all, GSync monitor, just did same test and see no delayed/mismatched or even dropped frames, perhaps thats hardware issue, not MPV related, I can clearly state that because my friends has 4070/4080s and have different behavior with RIFE engine with exactly the same settings, so thats impractical, OSD counts bad, see drops by my own, there was none!
narkohol wrote:RickyAstle98 wrote:One more thing, the video has some bad frames, trees sequence, with/without RIFE and with/without GSync with/without 60Hz mode and with/without scaling methods!
The skip/delay/mismatch frames happens from scaling methods also (when you downscaling/upscaling the source footage)!
It plays smooth with DmitriRender...
RickyAstle98 wrote:I have smooth 60FPS playback!
With which SVP settings and which player/video decoder and video renderer settings?
Default SVP settings and RIFE engine, mpv player...
narkohol wrote:I don't know why, but after trying a crap ton of different options and combinations I can’t get SVP RIFE to do smooth 60fps playback.
There are always some microstutters/hitching and Dropped/Skipped/Delayed frames when I use SVP with RIFE, even with low CPU and GPU % usages (I'm using scaling to QHD 1440p and black bars on a RTX4080 and Ryzen 5950x).
If I disable SVP there are no Dropped/Skipped/Delayed frames or stutters/hitching.
I tried MPV, PotPlayer, MPC-BE, VLC with all available video renderers and video decoders in each one, and the same happens in all of them.
All that I tried in detail here: https://www.svp-team.com/forum/viewtopic.php?pid=85781
TL;DR Can any of you play this file ( 30 seconds long dolly from Unbelievable 2160p DV HDR ) at 60fps perfectly fine with no Dropped/Skipped/Delayed frames and no visible hitches/stutters?
One more thing, the video has some bad frames, trees sequence, with/without RIFE and with/without GSync with/without 60Hz mode and with/without scaling methods!
The skip/delay/mismatch frames happens from scaling methods also (when you downscaling/upscaling the source footage)!
I have smooth 60FPS playback!
dawkinscm wrote:RickyAstle98 wrote:dawkinscm wrote:Are you seeing any difference between the old v4.25_heavy and the new smaller one? Have you guys tried the new 4.25_heavy with SC disabled?
The 4.25 lite (new) worse than most older models!
Performance? Can do 7x from HD source (24>168/720p)
Smoothness? Smoother than 4.15 model, but choppy sometimes!
Artefacts? A lot! Very noticeable difference against heavy!
The SC disabled!
RTX 4070
The QUALITY models
4.9/4.10/4.12/4.15/4.15l/4.18
The SMOOTH models
4.4/4.5/4.6/4.7/4.8/4.9/4.10/4.25h
What about 4.25h 100MB vs 4.25h 80MB?
I dont think theres the difference, downloaded since release, thats was 76 mb model!
dawkinscm wrote:Drakko01 wrote:Blackfyre wrote:Having said that, artefact masking is not as good as the original 4.25 that was released as a beta.
I am glad that was released as a beta, because that older one still remains the best quality version IMO and better than 4.18 too.
Exactly right, it kinda like the new 4.25_heavy, but reintroduce some artifacts the most models has for especific scene that 4.25 was fixed already, I hope see a combination of the two in the next model
Are you seeing any difference between the old v4.25_heavy and the new smaller one? Have you guys tried the new 4.25_heavy with SC disabled?
The 4.25 lite (new) worse than most older models!
Performance? Can do 7x from HD source (24>168/720p)
Smoothness? Smoother than 4.15 model, but choppy sometimes!
Artefacts? A lot! Very noticeable difference against heavy!
The SC disabled!
RTX 4070
The QUALITY models
4.9/4.10/4.12/4.15/4.15l/4.18
The SMOOTH models
4.4/4.5/4.6/4.7/4.8/4.9/4.10/4.25h
Grinchy wrote:Someone else with some kind of micro lags since the newest nvidia Driver Update? (4090 @ 4.15)
It feeled much smoother before the update.
For me (4070) the models from 4.11 to 4.25 looks choppy sometimes, I updated drivers and nothing happens, 4.15 has tendency to move frame forward and backwards for background objects, which makes the video not smooth, but 4.15 is awesome for action movies!
For me 4.25 heavy FIX smoothness issue from 4.11 to 4.25 models, my RTX 4070 runs these models choppy, not performance issue, no monitor issue, 4.25h now smooth as 4.10 without any artefacts!
What about performance? Well, 4.25 v2 heavy enough for 576p 24>192 conversion without drops (2 threads)
abraxas wrote:4.25_lite v2: max 5% less GPU usage than non-lite (4K HDR source material, frc.frame.resize=-19201080, fixed 60fps/SDR tone-mapped, upscaled to 4K with mpv/a couple of shaders) , no artifacts in my test scenes in The Martian, Jason Bourne, Rogue One, Big Bang Theory (1080p source, fixed 48fps) and downloaded YT-Video with different aspect ratio (fixed 48fps).
Smooth overall visual impression an 27" 4k monitor. Much better than 4.26x (which has many artifacts in my test cases).
Shaders:
hdebeband.glsl
SSimSuperRes.glsl
FSRCCNNX_x2_8-0-4-1.glsl
contrast.10.hk
I think your shaders somehow interfere RIFE algorithms, because 4.26 was better than 4.25L during my test sessions, a lot, but pixel crosstalk still an issue!
What do you mean upscaled back to 4K thats output resolution now? (resize and upscaling interfere RIFE methods - tested)
RTX 4070 (10 months old)
4.25 V1/V2 Lite test > worst (the pixels interpolating separately from main object) (not happens from 4.0 to 4.18)
Test parameters: mod8/mod32/mod64 inputs | builder optimization levels 2/3/4
scb wrote:RickyAstle98 wrote:NVOF : good (smoother than SVP)
SVP : normal
RIFE : good/normal (weird changes depends on threshold but perfect with newer models)
I think we're talking cross purposes here, the image comparison list in SVP has the following options
* Disabled
* SVP motion vectors
* NVOF motion vectors
* Image comparison
It's the differences between these four options that I'm looking for clarification on. Can you help?
Disabled - RIFE image comparison method (100% threshold) (same problems as image comparison with threshold >25%)
SVP motion vectors: smooth transitions between scenes but suffers with high speed camera movement (high frame repeat/frame blending)
NVOF motion vectors: smoother transitions between scenes than SVP but has wavy artefacts between scenes (low frame repeat/frame blending)
Image comparison: RIFE has weird scene transitions, overall smoothness better than others (high frame repeat/frame blending with threshold <5% / scene weirdness differs from model and threshold)
scb wrote:Chainik wrote:different scene change detection methods work differently 
I'm sure 
How would those differences be described, to allow an end-user like myself to know what might be reasons to pick one over the other?
NVOF : good (smoother than SVP)
SVP : normal
RIFE : good/normal (weird changes depends on threshold but perfect with newer models)
abraxas wrote:pensioner600 wrote:I checked it many more times 4.26. It's surprising that many people don't see her mistake. No, I stay at 4.18, that's the best for me.
If you have The Martian: at timestamp 18:33 and later, you can see bars between the bed and his workplace in the background. At 4.18 there are significant artifacts there. Not at 4.26, not at 4.22v2 lite, not at 4.9(!!). All the others I tested (4.15v2 lite, 4.16v2 lite, 4.18, 4.18v2) have artifacts.
Of course, this is only one scene, but for me these kinds of artifacts are the worst, because they catch my eye immediately.
For THIS test case, 4.26v2 and 4.9 are visually identical at normal speed.
Testing parameter: 4k source material, scaling to -19201080, upscaling with MPV and shaders to 4k LED screen.
Nope!
Posts found: 26 to 50 of 190