With or without 8 I'm getting 16.1772hpm That's not right
I'll try adding 16 to the end or multiple instances
The linux client it should use all of the cores you have available. No need to run multiple instances.
What is the output of (syntax may very a bit by distro):
cat /proc/cpuinfo | grep -c "^processor"
cat /proc/cpuinfo | grep "^model name" | uniq
If I had to guess the 50% is because of the block time (start, stop, start etc). I will check with mine to confirm..
Edit: Hmm a lot of mine are doing the same. Using ~55% of the CPU per thread.