I have a task which takes a long time on a PC: rebuilding a complete Linux system, from compiler bootstrap to Qt compilation and application compilation.
A bit like "buildroot", but doing the same thing for two targets, embedded ARM64 and amd64, for the application development team to be able to develop on Windows.

I have a correct tool for the job, "AMD® Ryzen threadripper 2950x 16-core processor x32", but did only have 32 GBytes RAM and would have liked to get the complete build quicker to do multiple tests. Using Ubuntu 21.04, most of the work done inside Docker (of Ubuntu 20.10) for "root" ownership problems.
So as reference, I have done the exact same job with the "time" command, on different PCs:

desktop, 32GB RAM: (32 cores, AMD Ryzen Threadripper 2950X 16-Core Processor):
real 756m37.391s -> 12h36m
user 3673m34.611s -> average 4.85 cores used
sys 245m40.424s

laptop, 32GB RAM: (12 cores, Intel(R) Core(TM) i7-8850H CPU @ 2.60GHz):
real 856m5.012s -> 14h16m
user 3815m59.492s -> average 4.45 cores used
sys 201m1.650s

remote PC, 128GB RAM: (48 cores, AMD Ryzen Threadripper 3960X 24-Core Processor):
real 557m18.318s -> 9h17min
user 2671m37.365s -> average 4.79 cores used
sys 164m36.397s

Then I upgraded my desktop PC with 32 GBytes more memory, so now have 64 Gbytes total.
Same specification, identical DRAM (going from 4 * 8 GBytes to 8 * 8 GBytes):

desktop, 64GB RAM: (32 cores, AMD Ryzen Threadripper 2950X 16-Core Processor):
real 758m45.548s -> 12h38m
user 3737m52.381s -> average 4.93 cores used
sys 241m52.829s

I lost 2 minutes by upgrading the destop from 32 GBytes to 64 GBytes...
Expensive downgrade!
Anybody would know why?