Glad to know its not just me who’s effected by performance regression in some models…
1 Like
Can’t download the stabilization model. Get an error message.
Error: Connection handling canceled.
3 Likes
Somehow the red Close Windows button is not working, i tried the exit command under File menu and it didn’t work either…
Is cpu really matter for topaz? I mean I have 8700K and I get 15 fps when converting 30 fps full hd video to 60 fps (apollo / not fast one)
gpu usage is %90 and cpu %30. I think im not cpu bottlenecked?
(rtx 5080 + 32 gb 3600 cl 16 ram)
The 8th series, though… brrrr… it’s cold in here!
For Apollo, CPU doesn’t matter that much which is why your GPU is nicely utilized.
If u do any 2X or 4X upscaling task, your CPU/RAM will limit your 5080 severely. But high end GPUs will always be limited by GPU/RAM in 2X/4X tasks because even the fastest CPU/RAM combo is not fast enough to provide the GPU with the next image to process.
Topaz would need to make better use of the CPU for this to change. Atm, Topaz performs best for most tasks when using only 8 cores without hyper threading on my 9950X3D (60% of CPU is unused). For scaling tasks, Topaz should create more CPU threads to upscale multiple images at once. This way they could possible almost double the scaling performance on high-end systems.
1 Like
power of the 8700k 
For me, Nyx saw about a 30% decrease in speed for 6.2 (compared to 6.0.2) for both the normal model and the fast model. 14900k, 4090.
2 Likes
Is that benchmark or regular video?
4090 is 29 times faster than 1060 =) 42.86/1.48
and 1060 is 3 times faster than UHD770 =)
btw 1060 is 10 times faster in games than UHD770 =)
Topaz Video AI v6.1.3
System Information
OS: Windows v10.22
CPU: 12th Gen Intel(R) Core(TM) i5-12600K 31.819 GB
GPU: NVIDIA GeForce GTX 1060 6GB 5.8877 GB
GPU: Intel(R) UHD Graphics 770 0.125 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 01.48 fps 2X: 01.04 fps 4X: ERR fps
Iris 1X: 01.68 fps 2X: 00.97 fps 4X: 00.10 fps
Proteus 1X: 01.49 fps 2X: 01.07 fps 4X: 00.18 fps
Gaia 1X: 00.11 fps 2X: 00.08 fps 4X: 00.05 fps
Nyx 1X: 00.24 fps 2X: 00.30 fps
Nyx Fast 1X: 00.80 fps
4X Slowmo Apollo: 01.42 fps APFast: 07.76 fps Chronos: 00.74 fps CHFast: 00.43 fps
16X Slowmo Aion: 02.61 fps
Topaz Video AI v6.1.3
System Information
OS: Windows v10.22
CPU: 12th Gen Intel(R) Core(TM) i5-12600K 31.819 GB
GPU: NVIDIA GeForce GTX 1060 6GB 5.8877 GB
GPU: Intel(R) UHD Graphics 770 0.125 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 720x480
Benchmark Results
Artemis 1X: 10.96 fps 2X: 07.62 fps 4X: 01.86 fps
Iris 1X: 12.00 fps 2X: 06.95 fps 4X: 01.15 fps
Proteus 1X: 10.85 fps 2X: 07.49 fps 4X: 02.74 fps
Gaia 1X: 01.20 fps 2X: 00.62 fps 4X: 00.51 fps
Nyx 1X: 02.95 fps 2X: 02.80 fps
Nyx Fast 1X: 07.30 fps
4X Slowmo Apollo: 07.67 fps APFast: 54.18 fps Chronos: 06.48 fps CHFast: 10.83 fps
16X Slowmo Aion: 18.10 fps
In Topaz Video Ai 6.2, when the original file is “WMV9 Advanced Profile” 29.97fps, even if interpolation is set to 59.94fps, the file is not correctly set to 59.94fps but is output at 29.97fps.
File loading and original file preview are functioning correctly, and upscaling and encoding also work as expected. However, even when setting 59.94fps, it’s not reflected correctly, and the original frame rate of 29.97fps is outputted.
1 Like
That’s what my post was trying to point out. The benchmark reports a low number, but processing a video happens at the same speed as before, if not better.
I thought I should mention that I get the same error that seems to be affecting multiple people when downloading the stabilization models:
“Error: Connection handling canceled.”"
I’ve tried the following, without success:
- setting model manager to Auto or RTX 4080
- fresh install
- firewall off and antivirus off
- auto-detect proxy settings on and off
- enable previous model versions on and off
Attempting to utilize Stabilization results in an “AI error” or an ffmepg crash, understandably. Luckily I rarely use Stabilization so for me this isn’t an urgent bug to fix, but since this seems like a common thing, I figured I should say something.
3 Likes
Which graphics cards does this apply to? It definitely doesn’t apply to the RTX 4000 series.
RTX 3080 ti. If you want, I can do a more intentional test. It would involve timing the same video with the same settings, before and after updating.
The speed regression in Iris benchmark holds true for processing on my 3070ti, with a 15-20% speed drop.
I’ve had similar issues for the last few updates. Using RTX4080. Driving me nuts that it’s not getting fixed and kicking myself I wasted money on paying for another year of updates that keep breaking things and slowing down my workflow which forces me to downgrade back to version 5.3.6. Really disappointed…once again.
4 Likes
ok this is even stranger. more fairly significant discrepancies between real world test and benchmark…
Nyx x1
BM: 13.06 fps
RW: ~14.9 fps
Nyx x2
BM: 9 fps
RW: ~11 fps
Nyx Fast 1x
BM: 31 fps
RW: ~15.1 fps
3 Likes
