4070 GPU not being used

[Alpen1]

Checklist

• I have searched existing issues to avoid duplicates
• I am using the latest version of VideOCR

Describe the bug

Hello,
During part 2, I noticed my nvidia gpu is not being used, but the desktop amd one is. I have checked on use gpu in advanced settings. I also downloaded the cuda setup for my series.

Steps to reproduce

1. Choose video
2. Hit run

Expected behavior

For my nvidia gpu to be used

Screenshots / Logs

🖥 Environment

OS: Windows 11
VideOcr: 1.3.2

[timminator]

Are you using the CUDA 11.8 or CUDA 12.9 version? You did not list this in your environment.

[hoanghunglb]

Are you using the CUDA 11.8 or CUDA 12.9 version? You did not list this in your environment.
Can you add multiprocessing?
I’m using a win10, Xeon CPU, vga 1660s and VideOcr 1.3.2 CUDA 12.9 version.
When I run it, it looks like it only uses a single thread—occasionally 4 CPU threads—and that one thread is always at high. Meanwhile, in Task Manager, both the CPU and GPU graphs sit around 2–3%. I’m not sure whether the CPU is involved in processing when using the GPU version, but I think it’s slow because the Xeon has a low clock speed and the workload is mostly single-threaded.

[timminator]

Are you using the CUDA 11.8 or CUDA 12.9 version? You did not list this in your environment.
Can you add multiprocessing?
I’m using a win10, Xeon CPU, vga 1660s and VideOcr 1.3.2 CUDA 12.9 version.
When I run it, it looks like it only uses a single thread—occasionally 4 CPU threads—and that one thread is always at high. Meanwhile, in Task Manager, both the CPU and GPU graphs sit around 2–3%. I’m not sure whether the CPU is involved in processing when using the GPU version, but I think it’s slow because the Xeon has a low clock speed and the workload is mostly single-threaded.

@hoanghunglb Step1 is single threaded. There is nothing that I can do here, Step2 uses your GPU.

[hoanghunglb]

@hoanghunglb Step1 is single threaded. There is nothing that I can do here, Step2 uses your GPU.

In step 2, GPU usage is also very low. I’ve also tried the CUDA 11.8 version on a GTX 1080 on another computer, and it produced a similar result. As shown in the screenshot, you can see the GPU seems like it isn’t running.

• tried on laptop with vga 2050, vga also not working.

[Alpen1]

Are you using the CUDA 11.8 or CUDA 12.9 version? You did not list this in your environment.

I am using 12.9.