Jetson Nano support #2481
Replies: 2 comments 2 replies
-
I know this is not a real world performance metric, but NVIDIA itself states that the Orin Nano can deliver up to 67 TOPS only. I don't have a number for lower tier RTX GPUs, but NVIDIA has said that a 4090 offers more than 1,300 TOPS. I doubt the Nano has the processing power to handle anything above SDXL. And, I haven't begun to mention the VRAM situation. |
Beta Was this translation helpful? Give feedback.
-
In my experience the Jetson platform runs models fairly well, particularly the AGX Orin at 64GB of memory. The Orin diffuses as fast as my 3080Ti. Can't speak to how well exactly the Orin Nano would perform, particularly with larger models. I started work on a docker container port for jetpack 6.1, but CUDA 12.6 is incompatible with xformers v0.0.27 it would seem. |
Beta Was this translation helpful? Give feedback.
-
Hi there,
is there already any working "package" (fork) to support the (old and/or the new Orin Nano) nvidia jetson ?
i can only find it for A1111 webui but not for directly for forge webui, i just ask before i try to dev it for me by myself.. -> https://github.com/dusty-nv/jetson-containers/tree/master/packages/diffusion/stable-diffusion-webui#user-content-containers
Beta Was this translation helpful? Give feedback.
All reactions