Hi, I'm trying Atomic Chat v1.1.70 on Windows and I have some questions about it.

1. Why it is so slow? I'm using gemma-4-E4B-it-IQ4_XS I have only 8t/s.
   With the official llama.cpp CUDA 13.1 I did 100t/s with Unsloth gemma-4-E4B-it-UD-Q4_K_XL.
   I leaved all the default settings in Atomic Chat and I installed only the CUDA 13 backend (as proposed).

2. Here my hardware specs. I have an RTX 5070, but no device found in Graphic cards. It's normal?

3) Gemma 4 MTP are supported on Windows? How to use it?

Thanks