There is a similar post which asks the question about scaling up hash rates with computational power here. My question is similar, however with the added question about a potential bottleneck.
I am running a self-contained node on an AMD EPYC 7402 on a Gigabyte MZ72-HBO motherboard (single CPU slot atm), with 2933MHz ECC RAM and blockchain located on an LVM-raid (mirror i.e. raid1 config) Samsung 870 QVO SATA SSD. If I mine with 24 threads, I get a hash rate of ~10,000H/s. If I double that thread count, then I only see an increase of ~1.5K-2KH/s. I see a limit to that increase at 36 threads. This increase is not substantial enough to warrant the extra power, it appears there is a sweet spot after which I see drastically diminishing returns.
Also, it appears that the Ryzen 9 5950X is hashing at almost twice this rate, but that doesn't make much sense. Even though the boost is an extra GHz, that is just boost. My cores are running at the max rate of 3.4 GHz consistently.
Is this a result of my total contribution to the global mining pool and inherent "limit" of what one person can contribute based on what is available? OR could this be the result of an unforeseen bottleneck somewhere in my system?
Would I benefit from placing the blockchain on an NVME SSD for example? Could the software raid be affecting read/write performance?
I'm not asking for specific hardware/config support, more curious if there are known limits other people have faced and things that have worked or not worked for their rigs.