Do neural networks train on processors?

Question

Neural networks

I wrote a simple version of the neural network in C ++, for training I chose the error back propagation algorithm / gradient of the error function.

Works for simple logical functions XOR, AND etc. But the training is very slow, +10 minutes suddenly . For logical functions, this is still tolerable, I have more time spent there creating / deleting two unfortunate threads, and other overhead costs of processor time.

In any case, it works too slowly, which is why the question is, on a regular (12 threads, frequency 3 GHz) processor, you can train which thread of a practical neural network without spending the rest of your life, or is it just using a GPU?

Link if interested to look at the code:
https://github.com/ASMcoder-Source/SimpleNeuro

PS: I understand that the code is just terrible, and if it were a little better, the results would be better, but still the question is the same.