I'd focus on who is sucking the cpu time. Use one of the sql tools to identify top cpu usage connections, or use this sqlhogs script during the heavy activity:
This are the results I get for the CpuPercent 4 4 3 3 3
I have a feeling it's related to temp tables, but the problem is not lack of memory or space with it. the user connections tho the temp db is 100% and doesn't go down also the transactions stays at 100% with an amount of 1500 transactions per second on average.
the active transactions is an average of 2 but climbs to 10
we have about a gb free of space in out tempdb the active temp tables is currently on 139 and is slowly climbing with a max of 150 now