Drakko01 wrote:

Any combination I try with LSFG +Rife its not given me better viewing experiences  that Rife x3 or x4 on my tv, maybe I doing something wrong or some setting/program or parameter in my config its interfering.

Looks like LSFG doesnt capture your frames properly, enable Draw FPS and you will see (captured frames / displayed frames) for example!

If you see 120/120 means LSFG detects your monitor refresh rate instead!

Close overlays like Steam, Discord, NVIDIA, everything!

Open video through mpv and press F (to go fullscreen) then Ctrl+Alt+S hotkey!

The behavior might look like > 48 / 120 (RIFE x2 for example)

Drakko01 wrote:
RickyAstle98 wrote:
Drakko01 wrote:

Hi! I recommend those who wanna experiment, the  new release https://github.com/AmusementClub/vs-mlrt/releases? for me work really well , not much as performance freak, but I KINDA improve image for me or maybe placebo effect.

Actually watching 15.9 footages, I see something changed too, anyways thats 50% placebo!

For performance freaks: 15.9 library - 1.2% performance drop!

Thanks for your testing! I mentioned am not a performance guy, because i'm more interesting of less errors and more quality, i take less performance for better quality anytime, a compromise that many of us do with a newer rife models, but anyways , im think there is some difference, less microstutters, more stable image, i can't quite figured out what exactly or if 15.9 its the reason,and honestly i dont wanna touch anything because a like the way it works now!

Less microstutters - placebo, quality - 50% placebo (atleast for me)

I tested RIFE x2 + LSFG 180FPS target (LSFG was updated and can generate frames without exact multipliers with consistent framepacing)!

LSFG from RIFE frames 48>180 looks almost the same as RIFE 24x7 and uses 6 times less resources!

Drakko01 wrote:

Hi! I recommend those who wanna experiment, the  new release https://github.com/AmusementClub/vs-mlrt/releases? for me work really well , not much as performance freak, but I KINDA improve image for me or maybe placebo effect.

Actually watching 15.9 footages, I see something changed too, anyways thats 50% placebo!

For performance freaks: 15.9 library - 1.2% performance drop!

10700K/4070 (HAGS OFF)

!TRANSCODING SPEEDS!

720p60>120 (4.4v2 / 1 GPU thread / 8 CPU threads) - 318 fps!
720p60>120 (4.4v2 / 1 GPU thread / 8 CPU threads / No Turbo) - 300 fps!

720p60>120 (4.4v2 / 1 GPU thread / 12 CPU threads) - 322 fps!
720p60>120 (4.4v2 / 1 GPU thread / 12 CPU threads / No Turbo) - 300 fps!

720p60>120 (4.4v2 / 2 GPU threads / 16 CPU threads) - 430 fps!
720p60>120 (4.4v2 / 2 GPU threads / 16 CPU threads / No Turbo) - 420 fps!

720p60>120 (4.4v2 / 3 GPU threads / 16 CPU threads) - 432 fps!

oriento wrote:

i DDUed to 572.70 and 572.16 (the lowest compatible with rtx 5000), and both have the same cpu usage at 100%. Indeed the gpu usage on taks manager is incorrect, but it's still less than 70%. HAGS still disabled
https://i.imgur.com/X8KCFsm.png

https://i.imgur.com/h237SJf.png

My task manager reports 36% overall usage! Your task manager reports 100% with 80% mpv usage!

EDIT1: I should say, you supposed to have 60FPS transcoding speed for realtime playback without issues!
EDIT2: I recomend to take Afterburner + SVP usage reports for exact CPU usage!
EDIT3: Try to rescale and pick Image Comparison method 25% please!

Thats reported or noticeable drops?

Doesnt matter why you purchased 5090 since you have CPU bottleneck!

Something weird with your CPU performance, asked a friend to test exact same SVP4 Pro package with same settings!
Your PC: 8700K/5090 | My PC: 10700K/4070 | His PC: 9700/4070S
He cant go over 50% CPU usage with 4K transcoding!

EDIT4: I recomend you Lossless Scaling program as alternative, 8 times less demanding than RIFE but 4 times less accurate, great solution for 4K realtime over 100FPS without drops until CPU upgrade!

I watched full movie Matrix Ressurections 3840x1600 RIFE 4.9v2 x2 mode, 2 GPU threads, GPU average temp 66C degrees, mpv reports 200 dropped frames out of 2h28min movie, no stutters, 4070 without overclock offsets! Maximum realtime CPU usage 34% w/Afterburner reports, 32% SVP program, 30% NVIDIA overlay, 40% Task manager, 98% average GPU usage, 180 watts average!

oriento wrote:

thanks for your help
When transcoding using nvenc (which uses mpv), I have 100% cpu and 35% gpu usage.
I have even lower performance than before reinstalling svp yesterday

when using svp to "open file via mpv", i have the same. 100% cpu and 35% gpu, and stuttering sad

https://i.imgur.com/OlyZVSD.png

https://i.imgur.com/B1Ee6fv.png

You cant see over 70% GPU usage because of processor bottlenecks, weird things happens!

10700K/4070OC with 4K realtime output gives me 36% CPU usage maximum (2 threads)

EDIT1: Try GPU acceleration!
EDIT2: Try performance boost!
EDIT3: Try HAGS options!
EDIT4: Fresh Windows 10/11?
EDIT5: Try older GPU drivers!

oriento wrote:

for me reducing the number of "gpu threads" to 1 reduce gpu and cpu load, but it's dropping a lot of frames

Then I cant help, but you have CPU bottleneck here, your GPU cant render more frames!

oriento wrote:

i have a rtx 5090 and a intel i7 8700k (will upgrade soon). When playing 4K videos, i have 100% cpu usage and 40% gpu. It's very stuttering. Is the cpu is really too old (it's supposed to run mostly on gpu ?!), or do I have something wrong ?

edit: reinstalled svp and reconfigured it, now cpu usage is at 60% and no dropping smile
however, at x3, the cpu usage goes to 100% and it's dropping. This is because my cpu needs to be upgraded or something is wrong ?

edit2: it reset madvr, that's why it's using less power ^^'

edit3: actually i still have big problems i upgraded to madvr beta, and it stuttered. Switched to mpc video renderer, and still stutter due to 100% cpu


Something wrong because my 10700K have 20% usage maximum during 1080p24>72 realtime playback!

With 4K the CPU usage doesnt increase exponentially, 30% average usage!

RTX 5090 stronger than 8700K (you cant use full GPU potential - CPU usage increase)

Try to decrease RIFE threads number, because 2 threads increase my CPU usage!

Chainik wrote:

please update now

Thanks! It worked now!

@Chainik

What happens after recent updates? Transcoding now broken!

VapourSynth encountered a fatal error: Filter SVSmoothFps declared the size 1280x736, but it returned a frame with the size 1280x720 (WTF)

Same happens for any declared sizes with returned frame sizes! I understand that triggers because of RIFE_SC method but previously did not!

osudahiuhsdfg wrote:
RickyAstle98 wrote:
osudahiuhsdfg wrote:

Does this look good for RIFE with a 5080? I get dropped frames at 48fps sometimes but if it's overclocked it seems fine. Any settings that would improve performance enough to permit 60fps? I'm using madVR so it's demanding.

https://i.postimg.cc/SRVW71gm/111.png

1) Use v2 models (15% faster without quality difference)

2) Use multipliers instead fixed refresh rate (30>48 have double performance penalty than 30>60 for example)

Hi where are v2 models? I don't see anything about that in the settings on that page. Also, I would love to do 60FPS but as I mentioned I can barely do 48fps.

You can download v2 models here: v1/v2 packages!

https://github.com/AmusementClub/vs-mlr … nal-models

Newest models like 4.25 heavy, 4.25 twice as much demanding than older legacy 4.4 - 4.9 models, 4.9 model is very acceptable for any videotypes, 4.15v2 lite for fast action movies!

Barely 48 fps but you used FIXED refresh rate which isnt multiples, when you tried 25>48 or 30>48 RIFE engine do more work before frames divided to output!

If you hit barely 48 fps with exact multiple of x2 for example, thats GPU issue, my 4070 can run nearly 1800p video (4.9v2) with x2 mode on 1 thread, which is enough for 2160p x2 conversions with 2 threads!

With v2 models you basically no need GPU overclock, also you supposed to use TRT boost option which gives another 5% extra performance, since engine build for exact container resolution!

osudahiuhsdfg wrote:

Does this look good for RIFE with a 5080? I get dropped frames at 48fps sometimes but if it's overclocked it seems fine. Any settings that would improve performance enough to permit 60fps? I'm using madVR so it's demanding.

https://i.postimg.cc/SRVW71gm/111.png

1) Use v2 models (15% faster without quality difference)

2) Use multipliers instead fixed refresh rate (30>48 have double performance penalty than 30>60 for example)

deleted

deleted

deleted

deleted

narkohol wrote:
RickyAstle98 wrote:
narkohol wrote:

Captain America The Winter Soldier - 2160 HDR - Triskelion - (Heavy artifacts on exterior windows and interior straight line patterns)
155MB
https://drive.google.com/file/d/1XQYN6p … sp=sharing

Cap America test scene, no artefacts, LSFG 3.0 x6!

You talking about using *just* LSFG? with no SVP RIFE at all?

Just tried with LSFG 3.0 x2 and x4, and certainly there are artifacts.

With SVP RIFE too, and also DmitriRender and hardware MEMC.
No solution is perfect in these areas:

https://i.postimg.cc/HxYHzrXv/image.png
https://i.postimg.cc/Hx6mhTtY/image.png

And yes thats not real scene, thats rendered scene (across the building)!

narkohol wrote:
RickyAstle98 wrote:
narkohol wrote:

Captain America The Winter Soldier - 2160 HDR - Triskelion - (Heavy artifacts on exterior windows and interior straight line patterns)
155MB
https://drive.google.com/file/d/1XQYN6p … sp=sharing

Cap America test scene, no artefacts, LSFG 3.0 x6!

You talking about using *just* LSFG? with no SVP RIFE at all?

Just tried with LSFG 3.0 x2 and certainly there are artifacts.

With SVP RIFE too, and also DmitriRender and hardware MEMC.
No solution is perfect in these areas:

https://i.postimg.cc/HxYHzrXv/image.png
https://i.postimg.cc/Hx6mhTtY/image.png

These places always ARTIFACTING with anykind of interpolation, but LSFG handles the transitions more stable, I dont know how, thats not fast paced scene, RIFE TRT trying to duplicate some areas across sections, sometimes with far more worst results than LSFG provide, but RIFE always been smoother on low framerates conversion!

narkohol wrote:

BTW, in case somebody is interested, these are my top test scenes. The first two are specially good for testing microstuttering/judder.

Top Boy S01E03 - 1080p - Intro  - (Ultra long dolly)
31MB
https://drive.google.com/file/d/16PiODq … sp=sharing

Unbelievable S01E01 - 2160p HDR - Intro (Ultra long panning)
65MB
https://drive.google.com/file/d/1SWNDEi … sp=sharing

Vikings - 1080p - (Text Artifacts)
34MB
https://drive.google.com/file/d/1KlDEw2 … sp=sharing

Captain America The Winter Soldier - 2160 HDR - Triskelion - (Heavy artifacts on exterior windows and interior straight line patterns)
155MB
https://drive.google.com/file/d/1XQYN6p … sp=sharing

Cap America test scene, no artefacts, LSFG 3.0 x6!

Chainik wrote:

> it broke rife 4.25 heavy models

yeah, doesn't work for me too

I have 4070 too, just rename 4.25 heavy model to 4.30 and everything works!

Santr wrote:
RickyAstle98 wrote:

The 4.4v2 model best on performance, but overall, quality remains questionable! Try everything from 4.4v2 to 4.9v2 or switch to 4.15v2 lite (4.9 perf)!

Tried all of them but it seems current RIFE that comes with SVP isn't quite there to handle it. Both 4.15l and 4.16l fall just 15 fps short of hitting the 120 target.
Now I'm eagerly awaiting for the update as that looks like just what's needed to get there.

This is strange, my 4070 handles 4.15l v2 1440p 60x2 without issues...

raider10 wrote:

Today I received my RTX 5080 and SVP has stopped working, while with my previous RTX 4070 it works perfectly. I have seen that some components need to be updated, but the application does not show any updates. Will SVP be updated? Can anyone describe step-by-step how to do it manually if necessary? Sorry, I don't speak English, I'm using Copilot to translate.

1) Download this

2) Unpack only vsmlrt-cuda folder from archive to ...\SVP 4\rife

Santr wrote:

Is there a model that can do 2x with a 1440p60 video on a 4080S? I know I can downsize with SVP but I'd like to avoid doing that.

The 4.4v2 model best on performance, but overall, quality remains questionable! Try everything from 4.4v2 to 4.9v2 or switch to 4.15v2 lite (4.9 perf)!

_hazoR_ wrote:

Managed to get a 5090 day1, would be really happy to be a betatester to all your experiments XD
Coming from a 1070ti and honestly can't wait to test properly RIFE (tested it with a 4060 max and the results were amazing)

It should come this week (or next one). Any suggestion?

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

Also dont forget update to v15.8: Blackwell support, latest TensorRT!

oriento wrote:

@Chainik, is SVP updated with tensort that is compatible with rtx 5XXX ?

Actually not, but 15.8 supports 50 series!

Chainik wrote:

NVOF is not an "AI" thing at all.
"SVP RIFE TensorRT" can use NVOF for a scene change detection only.

How to trigger NVIDIA Smooth Motion works through SVP? Thats new AI thing that working with optical flow, according to NVIDIA ofcourse!