Man, you are "inventing" problem while there is no one. As I wrote, you can hardly load cpu of your server to max with nginx, because it depends a lot on the content served, disk i/o, memory, etc. Simply said, nginx is not cpu-stress tool! Try something else, i.e.:
cat /dev/zero > /dev/null
or (if you do not care about entropy):
cat /dev/random > /dev/null
This can load your cpu (or vcpu) to the max, and disk/memory does not play any role here. So if your assigned 4vcpu, start it 4 times in 4 VM-terminals and then have a look (in 5th terminal) at cpu-load, i.e. with "top". BTW, switch the output to "per cpu" mode (iirc press "1" in top-terminal. You will see cpu-load goes to the max, there will be no wait/idle, and also vsphere-client shows max % and MHz in "performance" tab...