We have a data import & processing process that often longer than expected, often delaying the time at which users can begin working.
I've determined that the CPU runs around 80% for the 1-2 hours the process runs (90 minutes is the SLA), but rarely hits 100 percent.
I've found some queries that look at DMV data to determine if CPU pressure exists, and results are negative. However DMV data these are based on is cumulative since the last time SQL service started, and CPU runs < 30% most of the day (this is also an Availability Group that fails over once a month, so each instance is mostly idle during alternating months except for AG replication)
Is 80 percent CPU during a time-critical process by itself justification to add additional cores, or is there more homework to do?