Google: alphafold CASP, then you will find that in fact the best software is alphafold2, but you can't have it, then alphafold1, but you can't use it (at least for new proteins), so finally it turns out the best you can actually use is i-TASSER.
Secondary structure prediction is a mature field where the average 3-state prediction accuracy is in the range of 80-85% for most predictors. I could give you a long list, but some of the classics are listed here.
As to 3D predictions, it depends on whether you have a template or not. For template-based predictions, especially when the template and your protein of interest share at least 40-50% identical residues, there are many servers that will do the job. Here are some I have used: