Edit: Is a threadripper even recommended in this case?
i suppose, you do have a graphing solution running for your system!?
so just look at it - and learn to "read" / interpret it -, please ...
nobody can tell you a lot about your own server-systems performance / resource needs and therefore "best" areas of investment to achieve improvements than you as the person who runs the platform ... using decent (monitoring)tools.
if you don't have such a graphing-solution in place - install one right now() ... seriously ;))
... nobody really cares if its "good old" / (ancient) munin - very easy to install - or something "more modern" like prometheus plus a dashboard like grafana.
what counts is the insight you get from such a solution into your system / application / service(s) ...
just my 0.02€
ps. imho. in doubt invest in ram / io / cpu ... in that order and in a "balanced" fashion & new hardware also improves the availability of a system.
pps. and for example: reuse the old hardware as your monitoring system ... if you have none.