51 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

So this is a little weird. Before I got v4.15 Lite to be "almost" as good as v4.15/60fps by running the lite version at 72fps. But now, for the Alita battle drome scenes the Lite/72fps version actually handles artefacts a little better than base 4.15/60fps. I've made chages to my Nvidia Low latency settings which might make a processing speed difference. But I think maybe v4.15 has regressed a little with TRT 8.5 because I don't remember seeing these artefacts before but I have no actual evidence for this. I'm not prepared to do any new TRT testing until maybe 10.0.0.1 because even running at 72fps, the Lite version uses between 30 and 40% less GPU.

Go to forum: Using SVP

52 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

flowreen91 wrote:

If you set a low Scene change threshold like 1-15 then it will compare the two images and it might come to the conclusion that the difference between the two source images is big enough that it will consider it a scene transition.

On a related note, while I have been using the features of later vsmlrt versions I also set my SCT to 100.  After reseting everything to default (but still using the latest vsmlrt), I found I had to turn SCT on for best overall smoothness.The default value of 10 works best although 12 works better for one particular instance. YMMV.


flowreen91 wrote:

1. Both work, MPV slightly faster.

I would say MPV is significantly faster when stressed.


flowreen91 wrote:

3. Search "resize" in settings and specify your exact resolution downscale value. Try to add various shaders to regain visual quality.

This is interesting. I've tried downscaling before but the quality is clearly worse than mpv's class leading scaling. I think there is an option to change the scaling from bicubic to something better but it's been a while. BTW Is the number being used in your attachment 2560x1440 resolution?


flowreen91 wrote:

4. 2 threads+, enabled. Use which one you have better results.

I leave 2 threads on by default but tbh with my card I see no real difference between 1 and 4 threads.


flowreen91 wrote:

5.  Lite = uses less resources,

Except v4.14 "lite" which depending on your card can actually use more GPU resources.

Go to forum: Using SVP

53 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

OK so I'm using the latest vsmlrt script because it supposedly properly detects the TRT version. When I did that I notice that I loose some performance with the 9.2 test versions. So I reset everything back to default including the default SVP TRT version (8.5.1) and got the performance back. So where I am now is similar to where I was before. Rife 4.15v2 lite should be the new SVP default version because it uses similar GPU to v4.9 but has better artefact handling. But Rife v4.15v2 is still the best version overall because it removes or reduces every artefact I've ever come across.

Go to forum: Using SVP

54 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

abraxas wrote:

But - your milage may vary, I suggest doing some testing yourself. Especially if you are using VR.

Already have. HAGS is a problem for the VR streaming app I use.

Go to forum: Using SVP

55 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:
aloola wrote:
unrealit wrote:

Hi!

Sorry for the stupid Question!

But , How can i get Benchmark Values?

And what is HAGS?

What do you mean?

Thanks smile

use the script here.
https://github.com/AmusementClub/vs-mlrt/discussions/19

HAGS = Hardware Accelerated GPU Scheduling
you could google it for more info.

In RIFE with HAGS on performance might drop up to 25%.

HAGS ON is 25% perf drop? I think it would increase performance, no?

The reviews I've read of the feature suggests that any difference it makes to games is minimal at best. But it seems that for non games it can make things worse. I think I learned about turning it off a year ago or so from a chat I was having about VR performance issues.

Go to forum: Using SVP

56 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:

Nice, thank you! Do you guys test with pre-rendered RIFE or real-time RIFE rendering.

I've never even tried pre-rendered Rife. Real time Rife ALWAYS!

Xenocyde wrote:

Interestingly enough, the short clip is yielding different results on 4.16 V2 Lite. The long clip shows faint artifacts on the bottom left side of the net, but the short clip shows no artifacts in that area, although I can see faint artifacts on the right side of the net now.

Maybe try a longer clip of 20 seconds or so?

Go to forum: Using SVP

57 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:
dawkinscm wrote:
Xenocyde wrote:

How do you guys activate Rife in a short clip like that one I linked above? I play the clip but Rife does not activate. I have the long clip of 47 minutes, and I can see more interference artifacts in the goal net on 4.15 V2 towards the left bottom side. Those artifacts are barely visible on 4.16 V2 Lite.

As you have found, only clips above a certain duration will cause SVP to trigger. When I had this issue some time back, I found the setting for the video duration threshold and I think I set it to zero which means SVP will run on any length of video clip. I can't remember if it's code or an actual SVP setting because it's not in the SVP manual.

Welp, if anyone knows how to reduce that threshold to 0, please let me know.

OK found it. In the settings it's under setup and it's called min_duration.

Go to forum: Using SVP

58 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:

How do you guys activate Rife in a short clip like that one I linked above? I play the clip but Rife does not activate. I have the long clip of 47 minutes, and I can see more interference artifacts in the goal net on 4.15 V2 towards the left bottom side. Those artifacts are barely visible on 4.16 V2 Lite.

As you have found, only clips above a certain duration will cause SVP to trigger. When I had this issue some time back, I found the setting for the video duration threshold and I think I set it to zero which means SVP will run on any length of video clip. I can't remember if it's code or an actual SVP setting because it's not in the SVP manual.

Go to forum: Using SVP

59 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:
dawkinscm wrote:

Thanks but it says I need to request access.

Try this.

Thanks for this. I had a look with Rife v4.15 v2 lite and even slowed it down to make sure. I saw no fast movement artefacts when panning or anything else.

Go to forum: Using SVP

60 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:
dawkinscm wrote:

Fair enough. All the fast motion artefacts I see are improved with v4.15 but left untouched by v4.16. Can you could post a clip? It might be useful to  add it to my test collection.

Here it is.

Thanks but it says I need to request access.

Go to forum: Using SVP

61 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:
dawkinscm wrote:

I don't know about Anime but for live action both versions of 4.15 are much better for fast movement artefacts. Also mpv has nothing to do with Rife AI generated artefacts because they are "AI generated" artefacts.

I'm testing with live action, never tested anime. I have a specific scene from a BBC documentary abut the 1990 football cup in Italy where the camera pans from right to left and the football net usually shows interference artifacts. This scene looks a bit better on 4.16 V2 Lite compared to 4.15 V2.

Fair enough. All the fast motion artefacts I see are improved with v4.15 but left untouched by v4.16. Can you could post a clip? It might be useful to  add it to my test collection.

Go to forum: Using SVP

62 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Xenocyde wrote:

Tried 4.16 V2 Lite with the new MPV update a few days ago and it looks like the fast movement interference artifacts are almost gone now.

I don't know about Anime but for live action both versions of 4.15 are much better for fast movement artefacts. Also mpv has nothing to do with Rife AI generated artefacts because they are "AI generated" artefacts.


scb wrote:

I noticed there is only a 'Lite' version available for 4.16. Does anyone know if that's how things will be from now on, just 'Lite' versions?

They were planning to use the "lite" from now starting from v4.16. But v4.15 is much better so that might have been a miscommunication. Either way it looks like they want to start using "lite" models.

scb wrote:

What does Lite even mean in practice?

With the exception of 4.14 lite, they use less GPU resources than the non "lite" models.

Go to forum: Using SVP

63 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

SHTH34D wrote:

Does this showing up in the command line mean I need to change something?

"Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32."

Yep just ignore it. Internal stuff.

Go to forum: Using SVP

64 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:

Please reatach my reedited comment again, dont reply to it, will be deleted smile

I deleted my reply so i don't have anything to reattach. Don't worry about it smile

Go to forum: Using SVP

65 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:
dawkinscm wrote:
RickyAstle98 wrote:

In my case problem not with artefacts, frames are mismatched or delayed a bit!

Is that for build_optimization or running at 72fps.

My RTX 4070 doing perfect smooth 4.4-4.9 models, 4.10+ models not smooth, with or without build optimization or running at 72fps or dropped frames...
Update : opt level 1 helps 4.10 to 4.14 models for smoothness...

That's weird but I am interested to see the rest of your findings. But I don't like the fast action artefacts that models below 4.15 produce. Running at 72fps there are some slow pan issues on a couple of movies but for now I will stick with v4.15 lite/72fps.

Go to forum: Using SVP

66 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:
dawkinscm wrote:
RickyAstle98 wrote:

Hehe!

Update:
I think my initial impressions were correct. For certain artefacts, Rife v4.15 lite 72fps works almost as well as Rife 4.15 60fps. For others they are about the same with both running at 60fps. So Rife v4.15 lite 72fps is not quite as smooth at 60fps but it minimises certain artefacts while being  smooth enough to not be a problem.

In my case problem not with artefacts, frames are mismatched or delayed a bit!

Is that for build_optimization or running at 72fps?

aloola wrote:

has anyone tried the option builder_optimization_level=5?

BTW Thanks for the reminder. The last time I tried this I remember seeing warnings generated, possibly because it was running out of workspace. Now I get no more errors and I think it does help a little.

Go to forum: Using SVP

67 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:
dawkinscm wrote:
RickyAstle98 wrote:

But what TRT doing in my case? No drops no stutters but not smooth (any after 4.10), even with new optimization level, ah yes wait >
Builder optimization level 5 (default 3), adds another 3 minutes to build engine cache lock and only 3% performance!

Yeah I just tried it and if it is making a difference, it's at the margins and maybe pushing 4.15 lite closer to 4.15. But I will need to test with 4.15 to be sure. Also I'm not used to waiting this long for engine build anymore lol.

Hehe!

Update:
I think my initial impressions were correct. For certain artefacts, Rife v4.15 lite 72fps works almost as well as Rife 4.15 60fps. For others they are about the same with both running at 60fps. So Rife v4.15 lite 72fps is not quite as smooth as 60fps but it minimises certain artefacts while being  smooth enough to not be a problem.

Go to forum: Using SVP

68 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:

But what TRT doing in my case? No drops no stutters but not smooth (any after 4.10), even with new optimization level, ah yes wait >
Builder optimization level 5 (default 3), adds another 3 minutes to build engine cache lock and only 3% performance!

Yeah I just tried it and if it is making a difference, it's at the margins and maybe pushing 4.15 lite closer to 4.15. But I will need to test with 4.15 to be sure. Also I'm not used to waiting this long for engine build anymore lol.

Go to forum: Using SVP

69 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

RickyAstle98 wrote:

With TRT 10.0 my RTX 4070 struggles to 24>96 hd source (4.16 lite realtime), but easy 24>120 with TRT 9.2 hehe!
Maybe you know solution, why any models after 4.10 not smooth enough? But difference between 4.12 and 4.4 a whole pit, but SMOOTHNESS drop?
Maybe its frame pacing problem, just listen, 4.4v2 24>72 looks very smooth, but has a lot artefacts (no dropped frames), starting 4.10 everything looks like every 3th frame are mismatched (still no dropped frames)!

As I said, trt 10.0 is "one to avoid" because it cuts performance by around 30% while possibly doubling GPU usage.


aloola wrote:

has anyone tried the option builder_optimization_level=5?
The resulting engine may have better performance compared to an engine built with a lower optimization level.

They key phrase there is "may" have better performance. Yes I've tried this a couple of times. It's supposed to try to find more potential heuristics from what I understand but I didn't notice any real difference. But I can't remember if I have tried with my increased workspace so I might try again.

Go to forum: Using SVP

70 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Olombo wrote:

You do not miss that dropped framea with 60p then? No judder visible like this?

Nope. But YMMV.

Go to forum: Using SVP

71 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

Two bits of news. A while ago I said that increasing the frame rate can help with certain fast movement artefacts. Well initial testing suggests that since v4.15 lite is actually "lite",  I'm now able to run SVP at 72fps and it is helping with certain specific types of fast movement artefacts. It's dropping exactly 12 frames a second which makes sense but the helpful effect is not being lost. I still need to do more testing but this looks promising.

Also, Tensorrt v10.0.0 is available. I couldn't believe how bad the benchmarks look but the notes confirm that this is definitely one to avoid.

Go to forum: Using SVP

72 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

oriento wrote:

i'm seing artifacts in 4.15 lite that i don't see in 4.15

That's why I said:
Quality: Rife v4.15 > v4.15 lite > v4.16 lite.

Go to forum: Using SVP

73 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

abraxas wrote:

4.16 lite is really good, my first tests show the same quality with action movies as 4.15v2 - but 50% better performance. I could go up from x2 (48fps) to x2,5 (60fps) on a 4070 (about 90% GPU) for (nearly) 4K movies.

Rife v4.15 lite has been released and I think it will become the new SVP standard. v4.15 lite handles fast motion artefacts much better than 4.16 lite and is very close to v4.15 but with much better performance smile

Quality: Rife v4.15 > v4.15 lite > v4.16 lite
GPU usage: Rife v4.15 > Rife 4.15 lite = Rife 4.16 lite

Go to forum: Using SVP

74 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

abraxas wrote:

4.16 lite is really good, my first tests show the same quality with action movies as 4.15v2

For artefacts it may be a little better than 4.9 but it's definitely not the same as 4.15, not even close. But I have access to a large number of movies and over time I've worked out which ones stress the models. Also I'm watching on a big assed cinema sized screen big_smile So if you don't see any of the issues I and others can see with the "your" movies then it does not matter smile

Go to forum: Using SVP

75 Reply by dawkinscm

(1,566 replies, posted in Using SVP)

aloola wrote:

my point is 4.9 and 4.16 lite have the same performance but 4.16 lite has better quality. then there is no reason for people to use 4.9 even if live-action films are the same quality (no proof or testing yet)

I think your final point makes the most sense, so I did some testing using the same scenes I always test with. Quality wise they are very similar although 4.16 lite might have a slight edge. Rife 4.16 lite also uses about 2% less GPU resources than 4.9. So yeah it looks like 4.16 lite might be worth the upgrade smile

Go to forum: Using SVP