>>2831084>have you triedYes. Only thing where I noticed difference is --xformers. That doesn't exist on AMD GPUs. On newer UI version it autoloads I think.
>export ...This is for selecting a GPU. The Nvidia equivalent is CUDA_VISIBLE_DEVICES=# and yes this allows you to run multiple instances, each on whatever GPU you chose. 0 performance impact. I did this my system and they gen at almost same speeds as if only one is running.
>--listenThis lets you connect from other than localhost. 0 performance impact.
>--enable-insecureThis is a restriction for extension installation. 0 performance impact.
>restThese are all for compatibility or situations where you cannot prompt at all! There is no magic software switch for your transistor.