The generation has become slow, and very slow

[Juranik]

I use a video card from NVIDIA I have a CUDA pwtorch, what's wrong? Please help

[Mikerhinos]

If I'm right, you're using shared GPU memory which means that you're over capaciting your GPU VRAM, it uses RAM which is way slower, but it works, because otherwise you'd get CUDA out of memory crashes :)
I'm experiencing the same with SDXL models on my 8GB RTX3070 too (using the --medvram parameter :/ ).

[Juranik]

yeah.. alright

[Juranik]

thanks bro