I have been always using raxml for my phylogenetic analyses. However, I had to do a computationally demanding analyses and I tried Iqtree. I was extremely surprised by the speed of iqtree in comparison with raxml and I was wondering if you had a similar opinion on that or as I am using it for the first time I'm not completely mastering iqtree, which makes it go super fast.
It goes back and forth as both groups make improvements to their programs, but as of about a year ago IQ-TREE was significantly faster than RAxML. This is something that can be stated objectively, yet that alone may or may not be the deciding factor. Ultimately, we want the trees to be reliable, even if it takes a bit longer. So now we are entering a subjective territory, where again I would give IQ-TREE a clear nod over RAxML. IQ-TREE has a model finder to determine reconstruction parameters which RAxML didn't until recently, and still may not have it. I also like better their partition and mixture models, and less complicated command-line structure.
At the very least I think it is worth giving both programs a try in a variety of settings. For about half a year I was doing all tree reconstructions with both methods to give them a fair comparison, and eventually decided to switch completely to IQ-TREE.