Support for RTX 2060 Super 8GB VRAM?

[krigeta]

Does this support the second generation of Nvidia RTX cards, like the RTX 2060 8GB Super, which has 8GB VRAM? Will it work with less than 8GB VRAM?

[nitinmukesh]

We currently support only NVIDIA GPUs with architectures sm_86 (Ampere: RTX 3090, A6000), sm_89 (Ada: RTX 4090), and sm_80 (A100). See #1 for more details.

[krigeta]

We currently support only NVIDIA GPUs with architectures sm_86 (Ampere: RTX 3090, A6000), sm_89 (Ada: RTX 4090), and sm_80 (A100). See #1 for more details.

seems like again I miss a great opportunity to use flux.

[nitinmukesh]

Does your GPU support CUDA 12.6 drivers and torch 2.5.1+cu124.
If it does let me know.

[lmxyy]

I am not sure if 20-series GPUs work. It seems that they have INT4 tensor cores as in https://images.nvidia.com/aem-dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf. I will do a quick check.

[nitinmukesh]

You can also check using this repo. If it works for you (see if you are able to generate few images), I will integrate int4 and GGUF version of Flux.
https://github.com/newgenai79/sd-diffuser-webui

[krigeta]

Does your GPU support CUDA 12.6 drivers and torch 2.5.1+cu124. If it does let me know.

Yes, it supports this as I am using comfyUI and it shows me the same pair at startup.

[Dahvikiin]

Does your GPU support CUDA 12.6 drivers and torch 2.5.1+cu124. If it does let me know.

yes, RTX 2000 (Turing, sm7.5) are supported, even in the recent releases like NVIDIA CUDA 12.8 Update 1

[nitinmukesh]

Does your GPU support CUDA 12.6 drivers and torch 2.5.1+cu124. If it does let me know.

Yes, it supports this as I am using comfyUI and it shows me the same pair at startup.

If you are using Comfy you are set. There are many low VRAM workflows.

[krigeta]

Does your GPU support CUDA 12.6 drivers and torch 2.5.1+cu124. If it does let me know.

Yes, it supports this as I am using comfyUI and it shows me the same pair at startup.

If you are using Comfy you are set. There are many low VRAM workflows.

but my question was, how can I use nunchaku with RTX 2060