It has nothing to do with the CPUs on the clients. The CPUs on the relays are maxed out. We either need fewer clients on the network or more relays.