The article is about the performance of the code generated by Clang/GCC, not the compile time one gets using them. For those interested in compile time scaling with number of cores, below are results from testing I did on Windows 10 with latest Chrome sources:
https://www.reddit.com/r/Amd/comments/99ckjv/threadripper_29...