Inquiry regarding UniFab processing speed and GPU recognition

rockymountain

I am using UniFab’s “Video Upscaler AI”, but the AI processing speed is abnormally slow, taking about 6 hours or more to process a 10-second video.
I have checked and changed the PC and software settings as follows, but the problem is not resolved. Could you please tell me the cause and solution?

[PC Environment]
OS: Windows 11
CPU: AMD Ryzen
GPU 0: NVIDIA GeForce RTX 5050 Laptop GPU (VRAM: 8GB)
GPU 1: AMD Radeon 780M
Power: AC adapter connected to the outlet

[Settings Verified]
In UniFab “Settings > GPU > AI Processor”, “NVIDIA GeForce RTX 5050 Laptop GPU (Highest Performance)” is explicitly selected.
In Windows “Graphics Settings” and “NVIDIA Control Panel (Program Settings)”, the preferred graphics processor for UniFab has been set to “High-performance NVIDIA processor”.
I have also restarted the PC multiple times.

[Current Symptoms]
Previously, the integrated graphics (Radeon) was at 100%. However, after thoroughly applying high-performance settings in the OS and NVIDIA Control Panel, “GPU 0 (GeForce)” is now at 0% and “GPU 1 (Radeon)” is around 9%, meaning neither GPU is being used for AI processing anymore.
Instead, the load on the CPU (Ryzen) is constantly high in Task Manager, and it appears that emulation processing is being performed by the CPU instead of the GPU.
The AI model selected is “Equinox”.
It appears that the software side is not properly recognizing and utilizing the GPU for AI operations.

steelz1434419

Be happy your app is at least converting, though extremely slow due to not using your GPU. I haven’t been able to use UniFab since 4.x release. I try to convert a video (non-AI and AI version) and when I start the the process, the FPS spike like its doing something but then drop to 0.00. I’ve left it to see if it would just start up and after 24 hrs it never moves.. And what is worse is I pay for the Fab365 subscription and have since they first started the Fab365 subscription even through the price increases they impose, but with that extra I pay, the program has gone from functing to non-functional since 4.x first came out. I have had to revert to handbreak for emergencies but it certainly isn’t as good as what UniFab was prior to 4.x.

I hope for all of our sakes the Fab S&D team gets these issues resolve soon.

paulfletcher5705292304

steelz1434419

I have exactly the same issues. Paid for lifetime licences but can’t use them.

Ctrl+Z

@rockymountain

Thank you for detailing your environment setup and troubleshooting steps — this is very helpful for us in identifying the issue.

Regarding the GPU utilization showing 0% with all AI processing offloaded to the CPU:

We suspect the issue may be caused by an abnormal model or AI plugin that is failing to invoke the GPU. You can submit a Feedback report through the client. Your software version number and log files can be automatically packaged and uploaded via Client → Avatar (bottom left) → Feedback. We will investigate further.

—
steelz1434419
Regarding the FPS dropping to zero and video conversion stalling after version 4.x, as reported by steelz1434419:

This presents differently from the original poster’s issue and appears to be a separate regression. We have logged this feedback. We recommend submitting it through a Feedback ticket as well.

Ctrl+Z

paulfletcher5705292304

Which feature are you referring to in the perpetual license?

rockymountain

I have not yet received a response to this.

Adventurer

Dear Customer,

Thank you for contacting UniFab Support.

We sincerely apologize for the inconvenience caused by the abnormally slow processing speed in Video Upscaler AI, the GPU not being properly utilized, and the issue where conversion tasks drop to 0.00 FPS and make no progress after the 4.x release.

Based on the information you provided, you have already selected the high-performance NVIDIA GPU in UniFab, Windows Graphics Settings, and NVIDIA Control Panel. However, Task Manager still shows that the NVIDIA GPU is hardly being used, while the CPU remains under high load. This does require further investigation to confirm whether UniFab is correctly recognizing and utilizing the AI processor.

To help our technical team locate the cause, could you please provide the following information?

Your current UniFab version number
Your NVIDIA graphics driver version
The log files for the affected task
The format and codec information of the source video
Screenshots of the selected AI model and output settings
A screenshot or short screen recording showing the CPU/GPU usage after the task starts

If convenient, please also provide a short source video sample that can reliably reproduce the issue. You may upload it via Google Drive, OneDrive, Dropbox, or another cloud storage service, then send us the accessible download link. Please make sure the link permission is set so that we can access it.

We will forward the information to our technical team for further analysis, including whether the issue may be related to GPU recognition, AI model execution, driver environment, or compatibility with the current version.

Regarding the feedback from Fab365 users that UniFab has not been able to convert videos properly since the 4.x release, we will also record and report this to the relevant team for further checking.

Thank you for your cooperation and patience. Once we receive the requested information, we will continue to follow up and provide further assistance.
Best regards,
UniFab Support

ledang7842455663

If it helps, I noticed that GPU is only utilized when using H264 codec.

When using H265, GPU under 5% of load at all times.

P.S I have an RTX 5090 Laptop.