Depends. On a large scale yes. On a small scale there are a TON of individual scientists writing code for research papers and such who run singlethreaded code (myself) because we can't be bothered to parallelize. You are writing code that will take maybe a couple of days to run (time is important because you are waiting on results) but considering the time and effort to parallelize code it simply isn't worth it.
AMD chips simply are not worth the bother for me. Low FP and singlethreaded performance.
Not the mention the
TON of problems that can occur when multithreading floating point code. Breaking code into threads and putting them back together will possibly give different results depending on rounding errors.
Generally FP calculations are not even commutative
(a+b)+c != a+(b+c)
due to rounding.
Thus if you need a ton or reproducible accuracy MT is pretty much impossible.
https://software.intel.com/sites/default/files/article/164389/fp-consistency-102511.pdf