DarkAlchy

> Any update guys? 4070Ti here, even cannot get past cache latent.. Impossible to train now Yeah, we got sold a lemon while Jensen is laughing at us. I will...

Nope. That is about all you are going to get with a 4090. A 5090 I really don't know if it will be faster or not. Hardware wise the 4090...

> Have you already tried [disabling the sysmem fallback](https://nvidia.custhelp.com/app/answers/detail/a_id/5490/~/system-memory-fallback-for-stable-diffusion)? That has not one thing to do with this issue. Not a single thing. btw, it is now part of the...

It is faster in Linux but that is due to how Linux works.

> > > Have you already tried [disabling the sysmem fallback](https://nvidia.custhelp.com/app/answers/detail/a_id/5490/~/system-memory-fallback-for-stable-diffusion)? > > > > > > That has not one thing to do with this issue. Not a single...

> Same issue here, 4.61s/it training in Kohya_SS. NVIDIA DRIVER = Game Ready Driver 551.86 > > SYSTEM SPEC: RTX 4090 FE Ryzen 9 3900X MSI B450 MOBO 128GB DDR4...

On W10, with a 5600 3200mhz DD4 48GB at 7.5ish it/s in comfy and 1it/s slower in a1111.

> > AFTER = 1.78it/s > > It's very similar to mine. I don't think I have made any updates recently that would affect the training speed. I'm sure there...

This would be a great addition.

Qwen seems to be a bit better with realism images than Anime images as Cog is vs Intern with my usage of them.