151 (edited by UHD 20-12-2021 22:54:20)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Quaternions wrote:

unfortunately the interpolated video has striped artifacts outside of a 2048x2048 area

Did you use FP32 or FP16? What you write is a known problem for FP16:
https://github.com/hzwer/arXiv2020-RIFE/issues/188


Quaternions wrote:

3080 ti is very close to real time speed with cuda for 1080p

Thanks a lot for the testing and feedback on the new model. In fact the latest graphics cards are one step away from interpolating 1080p files in real time using RIFE. That is why our feedback to the RIFE developer is so important. Maybe something more can be squeezed out of RIFE?

I have written a request for help in optimizing RIFE for the latest graphics cards. You can find the details at this link: https://github.com/hzwer/arXiv2020-RIFE/issues/217

Now I need one or even better two or three people to check the GPU (CUDA) load during interpolation. I would be very grateful if you could do a simple test including CUDA load:


And here are the details of the test:

Fixed test parameters:

SVP & RIFE filter for VapourSynth (PyTorch)
re-encoding with x2 interpolation
RIFE model: 4.0
scale=1.0

Variable test parameters:

Math precision: FP16 and FP32

Test results:

re-encoding speed [FPS]
CUDA utilisation [%]

Video file:

original demo video from the creator of RIFE at: https://github.com/hzwer/arXiv2020-RIFE
720p (1280x720), 25FPS, 53 s 680 ms, 4:2:0 YUV, 8 bits
direct link: https://drive.google.com/file/d/1i3xlKb … sp=sharing


If the CUDA load is not displayed then you can check my post here: https://www.svp-team.com/forum/viewtopi … 493#p79493 and especially lwk7454's reply here: https://www.svp-team.com/forum/viewtopi … 497#p79497


If, in addition, you could find the time to do a similar test with some 1080p file then all the better and more data for the RIFE developer to analyse.


Quaternions wrote:

All in all very cool, I will definitely be watching anything 720p with rife

Once we can achieve x2 real-time interpolation for 1080p files, we will be able to simultaneously interpolate x3 real-time 720p files. So it's worth testing and giving precise data to the RIFE developer. When he sees how close it is I think he will find a way to optimize RIFE even more.

152 (edited by UHD 20-12-2021 23:52:58)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

dlr5668 wrote:

Yep. I use my PC as server for hyper-v machines so I cant change GPU driver or install CUDA toolkit

4.0 is also better under GPU load. 3.8 dropped my encode fps to 16 and 4.0 maintains 30-35 with less VRAM. Crazy magic

Thanks for the new information dlr5668. I just don't know where the drop in fps to 30-35 comes from? Is the earlier result some kind of GPU boost?

I also have a question, would you have the ability to do the above test I ask for Quaternions in post above? I am keen to know the performance information in fps combined with GPU load (CUDA)

I don't think you need to install the CUDA toolkit. It is probably enough to disable hardware scheduling, as lwk7454 wrote about: https://www.svp-team.com/forum/viewtopi … 497#p79497

You could probably find that the 3D load will equal the CUDA load. That's what I think, after what lwk7454 described, but I'm not sure. Unfortunately, it's hard to write something when I don't have a proper graphics card myself yet.

I don't know if lwk7454 is reading this thread and will be able to do some more testing.That's why it would be good if you could do a similar test that I'm writing about for both the 3.8 and 4.0 models. The idea is to be consistent in the data and see how the computing power utilization as a percentage changed between these models on the same graphics card.  The load on the 3090 and 3070 Ti cards can be quite different.

153 (edited by UHD 20-12-2021 23:52:17)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

lwk7454 wrote:

Here are the test results:

Parameters:
Test-Time Augmentation: Enabled [sets RIFE filter for VapourSynth (PyTorch)]
re-encoding with x2 interpolation
RIFE model: 3.8
scale=1.0
Encoder: NVIDIA NVENC H.264

720p, FP16
FPS: 63.5
Cuda: 56%

720p, FP32
FPS: 69.8
Cuda: 58%

1080p, FP16
FPS: 26.9
Cuda: 62%

1080p, FP32
FPS: 28.1
Cuda: 66%

I've also tried TTA Disabled for comparison, just 1 test:
720p, FP32
FPS: 25.1
Compute_1: 100%
Cuda: 15%


lwk7454, could you do the same tests for the latest 4.0 model?

The links to the files you tested are in this post of mine: https://www.svp-team.com/forum/viewtopi … 525#p79525

Unfortunately dlr5668 deleted the 1080p file he used to test, but I guess any 1080p file should give similar results.

I will of course forward the results to the RIFE creator here:
https://github.com/hzwer/arXiv2020-RIFE/issues/217

154 (edited by UHD 21-12-2021 00:04:16)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

dlr5668 wrote:

I also reuploaded image. Almost good to distribute with svp installer like regular RIFE wink You can remove all .lib and *train* files from pytorch + apply NTFS compress to reduce unpacked

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


I think it's worth it...

Quaternions wrote:

4k runs at 11.6fps which is much faster than the 1.9fps I got without cuda

and this...

https://github.com/HolyWu/vs-rife/blob/ … t__.py#L20

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

UHD wrote:
dlr5668 wrote:

Yep. I use my PC as server for hyper-v machines so I cant change GPU driver or install CUDA toolkit

4.0 is also better under GPU load. 3.8 dropped my encode fps to 16 and 4.0 maintains 30-35 with less VRAM. Crazy magic

Thanks for the new information dlr5668. I just don't know where the drop in fps to 30-35 comes from? Is the earlier result some kind of GPU boost?

You could probably find that the 3D load will equal the CUDA load. That's what I think, after what lwk7454 described, but I'm not sure. Unfortunately, it's hard to write something when I don't have a proper graphics card myself yet.

I don't know if lwk7454 is reading this thread and will be able to do some more testing.That's why it would be good if you could do a similar test that I'm writing about for both the 3.8 and 4.0 models. The idea is to be consistent in the data and see how the computing power utilization as a percentage changed between these models on the same graphics card.  The load on the 3090 and 3070 Ti cards can be quite different.

Playing a game or background virtual machine uses GPU

156 (edited by Quaternions 22-12-2021 05:20:39)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

I was able to get the CUDA graph to show up by disabling GPU hardware scheduling, the 3D reporting is very different from when GPU hardware scheduling was turned on but here is my results:
1080p half
https://i.imgur.com/K4m4F6I.png
1080p full
https://i.imgur.com/iZQaEkH.png
4k full (thanks for pointing out that it fixes the artifacts)
https://i.imgur.com/rCGoipz.png
demo clip full
https://i.imgur.com/e4ySsnu.png

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

RTX3000 rife cuda load should be at 50% max since

One of the key design goals for the Ampere 30-series SM was to achieve twice the throughput for FP32 operations compared to the Turing SM. To accomplish this goal, the Ampere SM includes new datapath designs for FP32 and INT32 operations. One datapath in each partition consists of 16 FP32 CUDA Cores capable of executing 16 FP32 operations per clock. Another datapath consists of both 16 FP32 CUDA Cores and 16 INT32 Cores. As a result of this new design, each Ampere SM partition is capable of executing either 32 FP32 operations per clock or 16 FP32 and 16 INT32 operations per clock

158

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

dlr5668 wrote:

Playing a game or background virtual machine uses GPU

OK, thanks, now I understand.

dlr5668 wrote:

RTX3000 rife cuda load should be at 50% max since

Here are the results showing 66% maximum CUDA load for the 3090 and the 3.8 model:
https://www.svp-team.com/forum/viewtopi … 526#p79526

If you could do similar tests on your card for model 3.8 and 4.0, we would have a confirmation how the CUDA load has changed compared to the old model on the same graphics card. If in fact in addition to the performance increase the CUDA usage has decreased, it means that the potential for further optimization of the 4.0 model is even greater than for the 3.8 model and that the bottleneck is not CUDA, but something else entirely, as I wrote about here:  https://github.com/hzwer/arXiv2020-RIFE/issues/217

Then we could add the results of such tests to the above post and maybe the RIFE developer will find a way to optimize. I think we should all care about that.

And here are the details of the test:

Fixed test parameters:

SVP & RIFE filter for VapourSynth (PyTorch)
re-encoding with x2 interpolation
scale=1.0

Variable test parameters:

Math precision: FP16 and FP32
RIFE model: 4.0 and 3.8

Test results:

re-encoding speed [FPS]
CUDA utilisation [%]

Video file:

original demo video from the creator of RIFE at: https://github.com/hzwer/arXiv2020-RIFE
720p (1280x720), 25FPS, 53 s 680 ms, 4:2:0 YUV, 8 bits
direct link: https://drive.google.com/file/d/1i3xlKb … sp=sharing

159

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Quaternions wrote:

I was able to get the CUDA graph to show up by disabling GPU hardware scheduling, the 3D reporting is very different from when GPU hardware scheduling was turned on but here is my results:

Thanks a lot for the tests smile

Would you be able to repeat these tests or at least one test - demo clip full (FP32) on 3.8 model with CUDA graph? I don't know if you have this version of RIFE.

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Instructions simplified (https://www.svp-team.com/forum/viewtopi … 695#p79695) cause Python 3.9 and Vapoursynth R57 are now installed/updated with SVP.

161

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Chainik wrote:

Instructions simplified (https://www.svp-team.com/forum/viewtopi … 695#p79695) cause Python 3.9 and Vapoursynth R57 are now installed/updated with SVP.

Thanks smile

What about the following parameter?
https://github.com/HolyWu/vs-rife/blob/ … t__.py#L20

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

if you think that asking the same question ten times will change anything - you're wrong wink

163 (edited by UHD 22-12-2021 15:13:26)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Chainik wrote:

if you think that asking the same question ten times will change anything - you're wrong wink

I know, but interpolating x10 with SVP&vs-rife on this 85" 4K 240Hz anti-reflection TV panel: https://www.displayspecifications.com/en/news/2ff1668 will change everything big_smile

164 (edited by UHD 23-12-2021 23:42:18)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Chainik wrote:

if you think that asking the same question ten times will change anything - you're wrong wink

Of course, if you perceived my posts as time pressure then I sincerely apologise.

I am keen to provide feedback to the RIFE developer as soon as possible on the new 4.0 model in terms of performance and CUDA core loading on the latest graphics cards. I posted my request to the RIFE developer when the 3.8 model was available: https://github.com/hzwer/arXiv2020-RIFE/issues/217 However, we now have the 4.0 model which, as the tests posted here show, is a colossal leap in performance.

I am afraid that if there is no feedback, the RIFE developer will close the thread, concluding that since he added a new model and there are no comments, the issue is closed. So far we have tests done by Quaternions, which shows that paradoxically the 4.0 model is not only faster, but also probably less demanding on CUDA cores than the 3.8 model: https://www.svp-team.com/forum/viewtopi … 727#p79727

I write probably, because we have the test results of the 720p RIFE demo file  showing the following load on the CUDA cores:

Cuda: 58%, FPS: 69.8, FP32, Model 3.8, NVIDIA GeForce RTX 3090
https://www.svp-team.com/forum/viewtopi … 723#p79723

Cuda: 46%, FPS: 91.3, FP32, Model 4.0, NVIDIA GeForce RTX 3080 Ti
https://www.svp-team.com/forum/viewtopi … 727#p79727

Because the tests were done on two different computers and two different graphics cards, we can't be entirely sure of a similar conclusion when testing different models on the same hardware. This is why I asked 3 people on this thread to test both models or completing the test of the missing model.

That said, the 3080 Ti and 3090 cards are pretty close in performance and the 12% difference in CUDA load between the two different RIFE models is likely to be confirmed on tests on the same hardware.  This would mean that despite the optimisation of the 4.0 model, the potential for further optimisation has paradoxically increased even more as the CUDA load has decreased, which as you can see is not the bottleneck here.

Now the most important thing: 91.3 fps indicates that we are very close to x3 interpolation of 720p files in real time! Perhaps this is already possible if x3 interpolation uses some synergy effect whether better parallel processing, lower VRAM bandwidth needs, or better use of CUDA processing power. Without testing we don't know that. And for testing we need your help Chainik and adding this parameter https://github.com/HolyWu/vs-rife/blob/ … t__.py#L20 to work with SVP.

With full x2 and x3 test results we will give a full feedbeck to the RIFE developer and maybe he can find a way to make it even more effective by making better use of CUDA cores and maybe even Tensor cores.

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

We already hit limits. Now you need better hardware or reduce model complexity. No need to report anymore

https://i.imgur.com/xLN32Su.png

166

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Hi, first post. I've been using SVP for a while and thought I'd look into RIFE after seeing it in the installer.

I'm eager to test it out, but to clarify first, I notice the wiki and SVP itself say it can only be used for 2X interpolation, but some of the posts in this thread (I think) seem to suggest this limitation is no longer in effect, SVP just needs to be updated to account for it. Is this correct?

I have a project I've been planning for a while now and if there's a major new feature like this on the horizon I'll put it off longer. Does anyone know when this feature might be added to SVP? Or might I be better off trying to install standalone RIFE and using that for now?

Thanks

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Chainik wrote:

Instruction simplified: Python & Vapoursynth updated via SVP's maintenance tool

=== RIFE / PyTorch installation ===

0. Switch to Python 3.9 and Vapoursynth R57

===> Run SVP's update tool!

1. Dependencies installation:
--------------
download https://bootstrap.pypa.io/get-pip.py into SVP 4\mpv64

run cmd as Administrator
cd SVP 4\mpv64

run

python get-pip.py

run

python -m pip install torch==1.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
python -m pip install vsrife

will download 3+ GB and unpack it into 6+ GB of data big_smile

2. SVP modification
---------------
replace SVP 4\script\base.py and SVP 4\script\generate.js, restart SVP

in the "RIFE AI engine" video profile set
- "Test-Time Augmentation" (TTA) = Disabled -> use RIFE/ncnn/Vulkan
- TTA = enabled -> use RIFE/torch/CUDA

"Math precision" = fp16 or fp32
"GPU device" -> choose device in case there're several CUDA devices in the system.

"AI model" and "GPU threads" change nothing.

Thank you very much. I am not an advanced user, I just followed this easy explanation and now  I have a RIFE that works, because the one built in RIFE in SVP did not work for me, it just gave me black screens in every other frames.

I want to ask which RIFE model I now use with the default setting that came with the explanation? I have tried all RIFE model with the program "Flowframes" and know that 3.8 is much better than all the others. Model 4 is fast, but comes with much more artifacts than 3.8 for example. I hope that I am now using 3.8.

I like SVP much better than Flowframes when using RIFE, because with SVP the interpolation starts instantly, so I not have to wait for the program to identifying scen changes, extracting every frame in the movie, and then interpolating those frames on disk so I end upp with some million pictures on disk for every movie I want to interpolate. That is to much hammering on the SSD.

I also want to know if anybody have find a way to make RIFE in SVP work with 4k and not only 1080p, because I know that RIFE is able to handle that, because in Flowframes it works. When I try to use RIFE in SVP, it produces vertical lines in the video when transcoding 4K material, but only half of the screen gets this vertical artifacts.

An other thing is that when using RIFE on 4k material, I can only use math precision 16 and not 32, because that made the 4 gb in the graphic card not being enough. How can I solve that without buying a new card? (HP support tells me that I can not put in another graphic card in my Pavillion All in one :-( ). I know there are some methods to solve this because math precision 32 in Flowframes, did not exceed the limits of my graphic card. The error report that I got gave me some indications that it may be a memory handeling method for solving this. I tried all options conserning GPU threads.

Remember that I am not an advanced user. Please talk to me as I am 7 years old. Lol. I managed to Install the Cuda capable RIFE in SVP but I have not understood what was really happening and why this can not be the default RIFE and so on. 6 gb pytorch is not normal? :-)))

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Ante85 wrote:
Chainik wrote:

Instruction simplified: Python & Vapoursynth updated via SVP's maintenance tool

=== RIFE / PyTorch installation ===

0. Switch to Python 3.9 and Vapoursynth R57

===> Run SVP's update tool!

1. Dependencies installation:
--------------
download https://bootstrap.pypa.io/get-pip.py into SVP 4\mpv64

run cmd as Administrator
cd SVP 4\mpv64

run

python get-pip.py

run

python -m pip install torch==1.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
python -m pip install vsrife

will download 3+ GB and unpack it into 6+ GB of data big_smile

2. SVP modification
---------------
replace SVP 4\script\base.py and SVP 4\script\generate.js, restart SVP

in the "RIFE AI engine" video profile set
- "Test-Time Augmentation" (TTA) = Disabled -> use RIFE/ncnn/Vulkan
- TTA = enabled -> use RIFE/torch/CUDA

"Math precision" = fp16 or fp32
"GPU device" -> choose device in case there're several CUDA devices in the system.

"AI model" and "GPU threads" change nothing.

Thank you very much. I am not an advanced user, I just followed this easy explanation and now  I have a RIFE that works, because the one built in RIFE in SVP did not work for me, it just gave me black screens in every other frames.

I want to ask which RIFE model I now use with the default setting that came with the explanation? I have tried all RIFE model with the program "Flowframes" and know that 3.8 is much better than all the others. Model 4 is fast, but comes with much more artifacts than 3.8 for example. I hope that I am now using 3.8.

I like SVP much better than Flowframes when using RIFE, because with SVP the interpolation starts instantly, so I not have to wait for the program to identifying scen changes, extracting every frame in the movie, and then interpolating those frames on disk so I end upp with some million pictures on disk for every movie I want to interpolate. That is to much hammering on the SSD.

I also want to know if anybody have find a way to make RIFE in SVP work with 4k and not only 1080p, because I know that RIFE is able to handle that, because in Flowframes it works. When I try to use RIFE in SVP, it produces vertical lines in the video when transcoding 4K material, but only half of the screen gets this vertical artifacts.

An other thing is that when using RIFE on 4k material, I can only use math precision 16 and not 32, because that made the 4 gb in the graphic card not being enough. How can I solve that without buying a new card? (HP support tells me that I can not put in another graphic card in my Pavillion All in one :-( ). I know there are some methods to solve this because math precision 32 in Flowframes, did not exceed the limits of my graphic card. The error report that I got gave me some indications that it may be a memory handeling method for solving this. I tried all options conserning GPU threads.

Remember that I am not an advanced user. Please talk to me as I am 7 years old. Lol. I managed to Install the Cuda capable RIFE in SVP but I have not understood what was really happening and why this can not be the default RIFE and so on. 6 gb pytorch is not normal? :-)))

4.0 works the best for me. I watched at least 10 hours with it

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

I made a PR that adds rational framerate multiplication:
https://github.com/HolyWu/vs-rife/pull/14
With this, SVP could set any fractional interpolation target.

170 (edited by UHD 30-03-2022 21:32:37)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

A few hours ago:

RIFE ncnn Vulkan - Release 20220330:
https://github.com/nihui/rife-ncnn-vulk … g/20220330

"update ncnn with nvidia tensorcore optimization"

I'm extremely curious, how much of a performance boost can this give?

If it could manage to get performance close to vs-rife: https://github.com/HolyWu/vs-rife that would be something great smile

171

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

And since a few days we also have a new v4.1 model: https://github.com/hzwer/Practical-RIFE

Has anyone already tested the interpolation quality of the new model and its performance?

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

UHD wrote:

A few hours ago:

RIFE ncnn Vulkan - Release 20220330:
https://github.com/nihui/rife-ncnn-vulk … g/20220330

"update ncnn with nvidia tensorcore optimization"

I'm extremely curious, how much of a performance boost can this give?

If it could manage to get performance close to vs-rife: https://github.com/HolyWu/vs-rife that would be something great smile

Dont think it will be faster than cuda version

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

UHD wrote:

And since a few days we also have a new v4.1 model: https://github.com/hzwer/Practical-RIFE

Has anyone already tested the interpolation quality of the new model and its performance?

Good news that we can just replace 4.0 flownet file and it will work without errors. No need to wait vs rife update

174

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

Ante85 wrote:

I have tried all RIFE model with the program "Flowframes" and know that 3.8 is much better than all the others. Model 4 is fast, but comes with much more artifacts than 3.8 for example.

Try the 4.1 model, and if the result is worse than 3.8 then report this particular case to the RIFE developer, as here: https://github.com/hzwer/Practical-RIFE/issues/10


Ante85 wrote:

I like SVP much better than Flowframes when using RIFE, because with SVP the interpolation starts instantly, so I not have to wait for the program to identifying scen changes, extracting every frame in the movie, and then interpolating those frames on disk so I end upp with some million pictures on disk for every movie I want to interpolate. That is to much hammering on the SSD.

For millions of pictures it is best to use a RAM Disk, such as OSFMount:
https://chiaforum.com/uploads/default/original/2X/9/968639ce0ccef0ba373c7d9749a0909a9fc5ec57.png.


Ante85 wrote:

I also want to know if anybody have find a way to make RIFE in SVP work with 4k and not only 1080p, because I know that RIFE is able to handle that, because in Flowframes it works. When I try to use RIFE in SVP, it produces vertical lines in the video when transcoding 4K material, but only half of the screen gets this vertical artifacts.

4K and FP16 is a known issue: https://github.com/hzwer/arXiv2021-RIFE/issues/188


Ante85 wrote:

I managed to Install the Cuda capable RIFE in SVP but I have not understood what was really happening and why this can not be the default RIFE and so on. 6 gb pytorch is not normal? :-)))

Yes, it is normal. The RIFE ncnn/Vulkan version takes up much less space, but is much slower. At least it was. There is hope for a speedup, but it's not clear how big. See next post...

175 (edited by UHD 31-03-2022 23:10:26)

Re: New RIFE filter - 3x faster AI interpolation possible in SVP!!!

dlr5668 wrote:
UHD wrote:

A few hours ago:

RIFE ncnn Vulkan - Release 20220330:
https://github.com/nihui/rife-ncnn-vulk … g/20220330

"update ncnn with nvidia tensorcore optimization"

I'm extremely curious, how much of a performance boost can this give?

If it could manage to get performance close to vs-rife: https://github.com/HolyWu/vs-rife that would be something great smile

Dont think it will be faster than cuda version


In the case of the GeForce RTX 3090 graphic card, using CUDA Cores gives similar processing power for FP32 and FP16:

35.6 Peak FP32 TFLOPS (non-Tensor)
35.6 Peak FP16 TFLOPS (non-Tensor)

However, using Tensor Cores gives much more processing power than CUDA Cores and the difference between FP16 and FP32 is already double!

142/284 Peak FP16 Tensor TFLOPS with FP16 Accumulate
71/142 Peak FP16 Tensor TFLOPS with FP32 Accumulate

Data source - page 44:
https://images.nvidia.com/aem-dam/en-zz … per-V1.pdf


The same GPU gives similar performance results for Flowframe for both FP16 and FP32 precision using the RIFE CUDA/PyTorch version. Also using the old RIFE ncnn/Vulkan version there is no clear performance difference between FP16 and FP32 precision.
https://www.svp-team.com/forum/viewtopi … 531#p79531

From this I conclude that both earlier versions of RIFE are based on the processing power of CUDA Cores.

According to the Ampere architecture based graphics cards specification I quoted above Tensor Cores should give better performance.

Of course, this is the theory. Practice should verify these assumptions. However, any performance increase will be welcomed smile