I want to upgrade my gpu from 1050ti to 2070super for learning tensorflow. However, I didn’t see 2070super on the cuda supported list, and I see someone in the forum say they have problem in tensorflow with 2070super. So I wonder if someone can tell me that if 2070super is suitable for tensorflow or not. sincerely thanks for your help!
The 2070 super shares the same CUDA compute capability (7.5) with the 2070 Ti, and other Turing-based GPUs. As such, it supports TensorFlow. I do not have a 2070 Super at hand to test with, but I can run tensorflow without issue on the Tesla T4 (which is based on the same TU104 chip as the 2070 Super).
thanks for your reply, nluehr. But I am still not very convinced, I am just afraid that I pay but I pay for nothing. Has somebody in this forum ever used tensorflow with 2070super? If so, please tell me. your advice is very important, thank you!
The 2070 (and all of the Turing GPUs) support TensorFlow. We test each of our NGC releases of TensorFlow (see https://ngc.nvidia.com/containers/nvidia:tensorflow ) against various Turing-based cards. And if you are looking for examples of models that are GPU-accelerated, I would suggest you look at https://github.com/nvidia/deeplearningexamples