RTX 2080 vs GTX 1080ti


Choosing between RTX 2080 and GTX 1080TI, which one would help me more in deep learning? I am asking this question because rtx 2080 has 8 gb of memory and 1080ti has 11. Is this more important than the speed?

fastai has been recommending a minimum of 6GB of GPU memory. The more memory, the better. You can run larger models or batches with more memory. For most situations, the speed difference between a 2080 and 1080Ti isn’t worth considering compared to memory sizes.


In case if you don’t use fp16 training, it would be better to have more memory. As I can understand, fp16 training doesn’t help in all cases, like, RNNs training or something else different from conv nets. I am going to test the performance of cards more thoroughly soon to have some real estimates.


You may want to check this thread:https://forums.fast.ai/t/gtx-2080-2080ti-rtx-for-deep-learning/26783