This not only improves application performance but also reduces server CPUutilization by as much as 30 percent, which potentially reduces the number of servers needed.
Hyper-Q -- This enables multiple CPU cores to simultaneously use the CUDA architecture cores on a single Kepler GPU. This dramatically increases GPU utilization, slashing CPU idle times and advancing programmability.