[bug]: Inconsistent generation times with sdxl, incredibly slow

[vayoa]

Is there an existing issue for this problem?

• I have searched the existing issues

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

rtx 3070 mobile

GPU VRAM

8gb

Version number

5.5.0

Browser

Firefox

Python dependencies

No response

What happened

Usually the first generation I do is at normal speeds, any generation after that is double the speeds of the first one. This happens to me with no other programs running or potentially using up my resources. Only a pc restart seems to solve this, but then again I usually only get 1 normal generation before the speeds double.

What you expected to happen

The speeds should be comparable if not lower to account for the model load times.

How to reproduce the problem

No response

Additional context

I'm using sdxl, fp16 accuracy. Just to compare, this problem does not happen to me while using other generation tools like comfyui.

Discord username

No response