Replies: 3 comments 4 replies
-
DirectMl is not really there at the moment to compete with Nvidia CUDA. AMD cards are just better value right now and SD DirectMl is the workaround to make them work for image generation. If you're buying the graphics card specifically for Stable Diffusion, Nvidia is a safer bet. |
Beta Was this translation helpful? Give feedback.
-
If someone read this. Is better use a 3050 than wherever shit of AMD, i have a 6750xt and is a crapy card :(, all here MUST have a NVIDIA card And 3060 12GB is very very faster than 6700xt, CUDA is really nice optmized, AMD dont have nothing, ROCm is other shit, less bugged than DirectML but its shit too and need linux for run it. In windows is impossible. |
Beta Was this translation helpful? Give feedback.
-
I must tell you now that for AMD, only from RX 7000 series onwards that they have designed their hardware for instructions best for neural networks. You should avoid RX 6000 series for this application. |
Beta Was this translation helpful? Give feedback.
-
Is better AMD 6700 XT 12GB or RTX 3060 12GB for generating? What is faster?
Now I have 6700 if I change it for nvidia 3060 12gb would it be better, faster?
Beta Was this translation helpful? Give feedback.
All reactions