ComfyUI not using RAM when on integrated GPU

[mastermach50]

I am running a Ryzen 7 5800HS with integrated graphics. On windows ComfyUI will use half of my RAM (16 GB Total) for the integrated GPU but not on linux.

How can I change that?

Environment Variables

HSA_OVERRIDE_GFX_VERSION = 9.0.0

Torch installed with Rocm via

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.0

Output

amdgpu.ids: No such file or directory
Total VRAM 512 MB, total RAM 15404 MB
Trying to enable lowvram mode because your GPU seems to have 4GB or less. If you don't want this use: --normalvram
Set vram state to: LOW_VRAM
Device: cuda:0 AMD Radeon Graphics : native
VAE dtype: torch.float32
Using sub quadratic optimization for cross attention, if you have memory or speed issues try using: --use-split-cross-attention
Adding extra search path checkpoints ./../models/checkpoints/
Adding extra search path clip ./models/clip/
Adding extra search path clip_vision ./models/clip_vision/
Adding extra search path configs ./models/configs/
Adding extra search path controlnet ./models/controlnet/
Adding extra search path embeddings ./models/embeddings/
Adding extra search path loras ./models/loras/
Adding extra search path upscale_models ./models/upscale_models/
Adding extra search path vae ./models/vae/

Import times for custom nodes:
   0.0 seconds: /mnt/external/AI/StableDiffusion/ComfyUI/custom_nodes/websocket_image_save.py

Starting server

To see the GUI go to: http://127.0.0.1:8188