The Forth books are good and interesting and prescribe a '''sane''' mode of writing Forth code: that you keep your definitions short and many, rather than ALGOL-shaped like you just want to write inside-out C code.
These books also talk about performance in the era of single-core machines much simpler than what we use today. I think the reality of Forth is that you need some kind of JIT compiler or the constant JMPing will tank your performance.
Also, Forth per se is terrible; you generally don't get so much as an array type out of the box, which might be okay for an exercise or for some chip that has <1k of RAM or something, but is a waste of time otherwise. Concatenative programming is great, and most of the good parts of Forth are applicable to Factor.